literature review on voice disorders

Applied Machine Learning Techniques to Diagnose Voice-Affecting Conditions and Disorders: Systematic Literature Review

Affiliations.

• 1 Department of Health, Blekinge Institute of Technology, Karslkrona, Sweden.
• 2 School of Health Sciences, University of Skövde, Skövde, Sweden.
• PMID: 37467031
• PMCID: PMC10398366
• DOI: 10.2196/46105

Background: Normal voice production depends on the synchronized cooperation of multiple physiological systems, which makes the voice sensitive to changes. Any systematic, neurological, and aerodigestive distortion is prone to affect voice production through reduced cognitive, pulmonary, and muscular functionality. This sensitivity inspired using voice as a biomarker to examine disorders that affect the voice. Technological improvements and emerging machine learning (ML) technologies have enabled possibilities of extracting digital vocal features from the voice for automated diagnosis and monitoring systems.

Objective: This study aims to summarize a comprehensive view of research on voice-affecting disorders that uses ML techniques for diagnosis and monitoring through voice samples where systematic conditions, nonlaryngeal aerodigestive disorders, and neurological disorders are specifically of interest.

Methods: This systematic literature review (SLR) investigated the state of the art of voice-based diagnostic and monitoring systems with ML technologies, targeting voice-affecting disorders without direct relation to the voice box from the point of view of applied health technology. Through a comprehensive search string, studies published from 2012 to 2022 from the databases Scopus, PubMed, and Web of Science were scanned and collected for assessment. To minimize bias, retrieval of the relevant references in other studies in the field was ensured, and 2 authors assessed the collected studies. Low-quality studies were removed through a quality assessment and relevant data were extracted through summary tables for analysis. The articles were checked for similarities between author groups to prevent cumulative redundancy bias during the screening process, where only 1 article was included from the same author group.

Results: In the analysis of the 145 included studies, support vector machines were the most utilized ML technique (51/145, 35.2%), with the most studied disease being Parkinson disease (PD; reported in 87/145, 60%, studies). After 2017, 16 additional voice-affecting disorders were examined, in contrast to the 3 investigated previously. Furthermore, an upsurge in the use of artificial neural network-based architectures was observed after 2017. Almost half of the included studies were published in last 2 years (2021 and 2022). A broad interest from many countries was observed. Notably, nearly one-half (n=75) of the studies relied on 10 distinct data sets, and 11/145 (7.6%) used demographic data as an input for ML models.

Conclusions: This SLR revealed considerable interest across multiple countries in using ML techniques for diagnosing and monitoring voice-affecting disorders, with PD being the most studied disorder. However, the review identified several gaps, including limited and unbalanced data set usage in studies, and a focus on diagnostic test rather than disorder-specific monitoring. Despite the limitations of being constrained by only peer-reviewed publications written in English, the SLR provides valuable insights into the current state of research on ML-based voice-affecting disorder diagnosis and monitoring and highlighting areas to address in future research.

Keywords: diagnosis; digital biomarkers; machine learning; monitoring; voice features; voice-affecting disorder.

©Alper Idrisoglu, Ana Luiza Dallora, Peter Anderberg, Johan Sanmartin Berglund. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 19.07.2023.

Publication types

• Systematic Review
• Research Support, Non-U.S. Gov't
• Machine Learning*
• Monitoring, Physiologic

REVIEW article

Investigating machine learning and natural language processing techniques applied for detecting eating disorders: a systematic literature review.

• Applied Machine Intelligence, Bern University of Applied Sciences, Biel/Bienne, Switzerland

Recent developments in the fields of natural language processing (NLP) and machine learning (ML) have shown significant improvements in automatic text processing. At the same time, the expression of human language plays a central role in the detection of mental health problems. Whereas spoken language is implicitly assessed during interviews with patients, written language can also provide interesting insights to clinical professionals. Existing work in the field often investigates mental health problems such as depression or anxiety. However, there is also work investigating how the diagnostics of eating disorders can benefit from these novel technologies. In this paper, we present a systematic overview of the latest research in this field. Our investigation encompasses four key areas: (a) an analysis of the metadata from published papers, (b) an examination of the sizes and specific topics of the datasets employed, (c) a review of the application of machine learning techniques in detecting eating disorders from text, and finally (d) an evaluation of the models used, focusing on their performance, limitations, and the potential risks associated with current methodologies.

1 Introduction

Recent reports in broad media about the latest conversational chatbots, which can generate human-like texts in response to user questions have made natural language processing (NLP) famous to the broad public. Yet the possibilities of this field go far beyond text generation and chatbots. Classifying texts into two (or more) groups and automatically extracting indicators that suggest that a text snippet belongs to either of the groups is also a common task. In particular, when using machine learning, this allows the identification of patterns that might differ from what a human might detect that are nonetheless effective in separating the two groups.

Meanwhile, in clinical practice in mental health, inventories with scaling questions are often used for diagnosis. Such inventories have limitations, including for example defensiveness (the denial of symptoms) or social bias that can influence the results of the questionnaires ( 1 ). In these cases, an automated text analysis applied to specific open questions or interview transcripts can provide further source of information indicating the patient’s condition that is more resistant to manipulations such as those arising from defensiveness.

Defensiveness is common amongst those afflicted with eating disorders (EDs). Respondents to a survey investigating the denial and concealment of EDs ( 2 ) reported a variety of attempts to hide the respective ED. Furthermore, the authors of the study state that such methods were described as deliberate strategies. This makes it challenging to use clinical instruments where an inventory item contains obvious indications for which options to choose in order to obtain a specific result.

EDs generally occur in the form of unhealthy eating habits, disturbances in behaviors, thoughts, and attitudes towards food, causing in some cases extreme weight loss or gain. These disorders not only impact mental health but also have physical effects ( 3 ). EDs are classified in the category F50 of the ICD-10 and can refer to different disorders including anorexia, bulimia or overeating 1 . A study conducted by Mohler-Kuo et al. ( 4 ) in Switzerland discovered that the lifetime prevalence for any ED is 3.5%. Another survey investigating the lifetime prevalence of EDs in English and French studies from 2000 to 2018 found that the weighted means were 8.4% for women, and 2.2% for men ( 5 ).

The power of natural language processing (NLP) has already been applied to the field of mental health, especially in research. Feelings and written expression are closely correlated: An analysis of student essays has shown that students suffering from depression use more negatively valenced 2 words and more frequently use the word “I” ( 6 ). Different approaches have been applied to explore how to use automated text analysis on tasks such as the detection of burnout ( 7 ), depression ( 8 , 9 ), the particular case of post-partum depression ( 10 , 11 ), anxiety ( 12 ), and suicide risk assessment ( 13 ), ( 14 ). Often, such methods are based on anonymized publicly available online data. Only little work makes use of clinical data. Furthermore, the English language has been the primary focus, even though these methods can be highly language-dependent, meaning that data and methods should be carefully reviewed when adapting to local languages. This is relevant, as it has been shown that adapting to the patient’s language is beneficial in mental health diagnostics and treatment ( 15 ). In our view, one aim of such technologies should be to explore ways to support clinical practitioners in their daily work, and provide them with additional sources of information to consider. Therefore, we often refer to such solutions as Augmented Intelligence 3 , rather than Artificial Intelligence, as they aim to empower humans rather than replacing them.

Despite existing work in the field of ML and NLP for depression, anxiety or suicide risk assessment, there has been a lack of a detailed systematic literature comparison on the automatic detection of EDs using NLP technologies for both clinical and non-clinical data. A recent survey ( 16 ) investigated the use of natural language processing applied to mental illness detection. The majority of the identified results (45%) had worked on depression, whereas only 2% were about eating disorders in general and 3% about anorexia. Whereas the broad scope of the survey provides a generous overview of the research landscape, it does not compare the case of eating disorders in detail.

In this paper, we have undertaken a systematic literature review to address this research gap, following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines ( 17 ) to ensure a well-structured and transparent methodology.

We contribute to the field by (a) analyzing the metadata of published papers to understand the current trends and methodologies, (b) examining the sizes and targeted topics of the datasets used in these studies, (c) reviewing how machine learning techniques are applied to detect eating disorders from textual data, and (d) evaluating the performance, limitations, and potential risks of the models deployed in this domain.

Our research is guided by specific questions, structured around four distinct perspectives, which collectively form the core of our investigative approach.

● Demographical Questions (DemRQ): Focus on metadata aspects of the paper:

● • DemRQ1: When was the paper published?

● • DemRQ2: From which countries were the contributors of the papers included in this study?

● Input Questions (InputRQ): Focus on the format and topic of the input data:

● • InputRQ1: Which languages were taken into consideration?

● • InputRQ2: What was the size of the dataset used?

● • InputRQ3: Which data sources were used for data collection in the case of both clinical and non-clinical data?

● • InputRQ4: What types of eating disorders were addressed in these studies?

● Architectural Questions (ArchRQ): Focus on the experimental architecture:

● • ArchRQ1: Which feature extraction technique was used?

● • ArchRQ2: Which machine learning techniques in the field of NLP have been used for ED detection?

● Evaluation Questions (EvalRQ): Focus on the evaluation aspects of the trained model:

● • EvalRQ1: How did the model perform?

● • EvalRQ2: What are the limitations and risks of the existing methods, and how can they be improved?

The article is structured as follows: First, we describe our methodology such as the study design and the paper selection process. We then describe the results of the literature search and describe the findings of our review. Finally, we summarize our results and describe perspectives for future research in the field.

2.1 Study design

To answer our research questions, we conducted a structured literature review (SLR) following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines ( 17 ). This includes standards for literature search strategies and setting criteria for the inclusion or exclusion of gathered works in the final review.

2.2 Literature search strategy

In accordance with PRISMA standards, we have set an 8-year time span for searching for documents (2014-2022) related to our research scope. We consider the year 2014 mainly because Bellows et al. ( 18 ) conducted a study on automatically detecting binge Eating disorder using clinical data, which we deem to be the initial research in the field. We then compiled a list of all databases to be searched. The list included the following databases:

● Google Scholar

● IEEE Xplore

● Pubmed

In addition, in order to efficiently conduct our database search we have compiled a list of keywords and conditions. These keywords are relevant to the research topic of EDs and their detection using NLP and machine learning techniques. Furthermore, the list included specific terms related to social media and online social networks in order to enable the identification of studies that explore the use of social media for the early detection of EDs, which is an ongoing research interest. The final query is presented below:

(eating disorder OR anorexia OR binge eating OR bulimia OR overeating) AND (natural language processing OR NLP OR text mining OR inventories OR machine learning OR artificial intelligence OR automatic detection OR early detection OR social media OR online social network OR clinical).

Using the aforementioned search keywords and conditions, we retrieved research articles where NLP techniques have been used for the detection of EDs from clinical and non-clinical data. The detailed workflow is depicted in Figure 1 , and the corresponding PRISMA flow diagram for this SLR is shown in Figure 2 .

Figure 1 Methodology for document collection.

Figure 2 PRISMA Flow diagram. Based on: Page et al., ( 17 ).

With the initially proposed search query, a large number of papers was identified. With manual analysis we explored options to define a more restrictive query, still making sure to capture the relevant papers, which turned out challenging. We therefore adapted our method to consider the first 100 elements returned by the search query on each database, sorted by relevance. This furthermore allowed to apply the same methodology for all three data sources, including especially Google Scholar, where the search functionalities are limited compared to databases like PubMed, and thus we had to make a selection on the number of items to be reviewed. Given the interidisciplinarity of our approach, we wanted to include Google Scholar to target a vast number of sources and ensure the most relevant work can be included.

A Python script was used to screen the articles for duplicates. As a result, 1 article was excluded from further consideration, leaving a total of 299 articles for further analysis (see Figure 2 ). To refine the results further, a manual title scan was performed to exclude articles that were not pertinent to the research topic. This resulted in the exclusion of 237 articles, leaving a total of 62 for further analysis. Additionally, a manual scan of the abstracts from the remaining 62 articles was performed to exclude any that were not relevant to the study. This process resulted in the exclusion of an additional 30 articles, leaving a total of 32 for inclusion in the final analysis. After thoroughly reading and evaluating 32 articles, 27 were selected as relevant for the researched topic (according to the criteria from Table 1 ). These chosen articles were deemed to possess high relevance and reliability for this SLR. Finally, we scanned the references section of the articles included in our survey and identified any relevant literature that may have been missed in the initial database search. This added n=18 articles to the studies that were finally included in the review (n=45). The process is illustrated in Figure 2 .

Table 1 SLR study selection of literature using inclusion and exclusion criteria.

2.3 Inclusion and exclusion criteria

Table 1 outlines the predefined exclusion and inclusion criteria that were used to guide the selection of related studies for the review. These criteria were established in advance to help simplify the process of identifying and selecting relevant papers. In particular, papers that focused solely on the psychological aspects of EDs and did not consider the use of automated text analysis technologies were excluded from the review. By adhering to these criteria, we were able to more effectively and efficiently select the relevant papers.

In this section, we provide a thorough review and analysis of the research studies included in this systematic literature review.

3.1 Terminology

● Bag of Words (BoW) is a fundamental technique used in NLP for text representation. It involves representing text data by counting the frequency of occurrence of each word in a document.

● Term Frequency-Inverse Document Frequency (TF-IDF) is a numerical statistic used to evaluate the importance of a word in a document within a collection or corpus. It combines two metrics: term frequency (TF), which measures the frequency of a word in a document, and inverse document frequency (IDF), which penalizes words that are common across the entire corpus.

● Bidirectional Encoder Representations from Transformers (BERT) ( 19 ) is a pretrained deep learning model introduced by Google in 2018. It belongs to the Transformer architecture and is designed to understand the context of words in a sentence by considering both left and right context simultaneously

● Word2Vec ( 20 ) is a technique for learning word embeddings. Word2Vec represents each word as a vector, with similar words having vectors that are closer together in the vector space.

● Global Vectors for Word Representation (GloVe) ( 21 ) is another technique for learning word embeddings. GloVe also generates vector representations of words based on their co-occurrence statistics in a corpus. However, GloVe considers the global context of the entire corpus to learn word embeddings, unlike Word2Vec, which focuses on local context.

● Embeddings from Language Models (ELMO) ( 22 ) is a deep contextualized word representation model. It generates word embeddings by considering the entire input sentence and capturing its contextual information.

● Doc2Vec ( 23 ) also known as Paragraph Vector, is an unsupervised learning algorithm to generate vector representations for pieces of texts like sentences and documents, it extends the Word2Vec methodology to larger blocks of text, capturing the context of words in a document.

● Bidirectional Long Short-Term Memory (Bi-LSTM) ( 24 ) is a type of Recurrent Neural Network (RNN) that processes data in both forward and backward directions. This architecture is particularly effective in understanding the context in sequence data like text or time series, as it captures information from both past (backward) and future (forward) states.

● Linguistic Inquiry and Word Count (LIWC) ( 25 ) is a text analysis program that counts words in psychologically meaningful categories.

3.2 Demographical research questions

Figure 3 shows the yearly distribution of the selected research work (DemRQ1). The data suggests a growing interest in this topic in recent years. This is in line with the findings of Zhang et al. ( 16 ) that found that there has been an upward trend over the last years in using NLP and machine learning methods to detect mental health problems. Notably, we highlight a prominent peak in 2018 and 2019, which coincides with the emergence of tasks related to EDs in eRisk competitions.

Figure 3 Yearly distribution of all research articles.

We also observed the geographical distribution of the authors’ affiliations of the selected studies (DemRQ2). As visualized in the heat-map in Figure 4 , 7 of the selected studies were from the USA and Spain, 5 from Mexico and France.

Figure 4 Geographic distribution of all institutions involved in the selected research articles.

From the 45 selected studies, 24 were results from the eRisk lab 4 , hosted by the CLEF Conference since 2017. This academic research competition focuses on the development and evaluation of text-based risk prediction models for social media. Each year, the lab provides a shared task framework where teams of participants are tasked with developing NLP techniques to automatically identify and predict the risk of different mental illness behaviors from social media data, including Eating Disorders. Participants are provided with a training dataset and a test dataset, and the performance of their models is evaluated based on two categories: performance and latency. The eRisk lab provides a unique opportunity for researchers to collaborate and innovate in the field of NLP and mental health, aiming to improve the detection and prevention of mental health issues in online communities. The datasets used in the eRisk lab are primarily sourced from the social media platform Reddit.

Since 2017, the challenge has included two tasks pertaining to the early detection of Eating Disorders. In both 2018 and 2019, the task involved the early detection of signs of anorexia [see e.g., Losada et al. ( 26 )]. In contrast, the 2022 iteration introduced a novel task centered on measuring the severity of eating disorders ( 27 ). This task diverged from the previous ones in that no labeled training data was supplied to participants, meaning that participants could not evaluate the quality of their models’ predictions until test time. The task objective was to assess a user’s level of eating disorder severity through analysis of their Reddit posting history. In order to achieve this, participants were required to predict users’ responses to a standard eating disorder questionnaire (EDE-Q) 5 ( 28 ).

3.3 Input research questions

Our first input research question (InputRQ1) investigates the different languages that are considered in the studies included in this SLR. Research has shown that only a small number of the over 7000 languages used worldwide are represented in recent technologies from the field of natural language processing ( 29 ). We wanted to investigate whether this is also the case for the detection of eating disorders. Text analysis, naturally, depends on the specific language and can typically not be transferred from one language to another without specific adaptions.

Table 2 gives indication about the language of data used, its size, its source, and the type of eating disorder that was investigated in the selected studies (excluding studies from eRisk). 18 of the 21 studies used English data, 2 used Polish and 1 Spanish data. The 24 papers from the eRisk lab challenges all relied on English data from the platform Reddit. Overall, only 3 out of 45 studies used a language other than English (7%). This confirms the need for further work in applying the latest technological developments to non-English texts.

Table 2 Datasets characteristics.

The dataset size is another crucial factor we took into account in our analysis ((InputRQ2). As depicted in Figure 5 , the distribution of dataset sizes used in the studies reveals that datasets ranging from 1k to 10k instances are the most frequently used.

Figure 5 Dataset sizes distribution based on Table 2 excluding articles from eRisk.

The distribution of dataset sizes across different research topics, as illustrated in Figure 6 , offers insightful perspectives. Notably, Anorexia research displays the most significant variance in dataset sizes, spanning from less than 1K to over 1 million data points. In contrast, binge eating research predominantly employs datasets within a narrower range of 1K to 10K data points. For broader Eating Disorders, 6 studies leverage datasets between 10K and 100K, while 3 others operate with datasets in the 100K to 1 million range. Finally, research on Mental Disorders encompasses datasets varying from 1K to more than 1 million data points.

Figure 6 Dataset sizes distribution by targeted ED based on Table 2 excluding articles from eRisk.

Table 2 also gives an overview of the data sources (InputRQ3). From the 45 studies, the used datasets can be classified as follows in four groups:

● eRisk lab datasets: 24 studies

● Other online forums and social media: 17

● Medical data: 3

● SMHD dataset ( 50 ): 1

The distribution of the primary focus of these studies is illustrated in Figure 7 (InputRQ4) The majority of the studies (n=29) we collected focused on anorexia, while 12 studies conducted a broader investigation of EDs in general rather than focusing on a specific type. Additionally, three studies had a more extensive scope, delving into various mental disorders, including but not limited to EDs, while one study focused on binge eating.

Figure 7 Research distribution of all research articles.

3.4 Architectural and evaluations research questions

3.4.1 erisk challenge.

Table 3 summarizes all the papers that we identified following our strategy, including the ones from eRisk. In 2018 and 2019, the eRisk papers focused on a text classification task aimed at developing an early detection system for eating disorders on social media using the history of users’ writings data. The aim was to train a text classifier that could effectively identify and flag potential cases of anorexia based on users’ social media content. For the eRisk challenge resulting in papers from 2022, the task was different. Participants were provided with the social media history of specific users and had to predict their answers to questions 1-12 and 19-28 from the Eating Disorder Examination Questionnaire (EDE-Q) 7 ( 28 ).

Table 3 Overview of machine learning methods and performance metrics of the studies included in this systematic literature review.

(ArchRQ1) The complexity of this task, along with the development in the field of NLP over the years 2019 to 2022, explains the choice of word2vec, GloVe ( 72 ) or transformer-based models ( 62 , 66 , 73 ) for vectorization/feature representation. For the remaining entries, very different approaches were used, ranging from anorexia specific vocabulary and LIWC ( 58 ) to more general approaches like Bag of Words (BoW) ( 52 , 53 ) or TF-IDF ( 51 , 57 ). (ArchRQ2) The choices of methods for prediction were also heterogeneous, ranging from cosine similarity ( 72 ) to linear models ( 52 , 54 , 58 , 66 , 71 ), to neural networks ( 51 , 53 , 56 ).

(EvalRQ1) For the 2018-2019 eRisk papers, we report F1 values corresponding to the binary classification task, whereas for the 2022 paper we report mean average error (MAE), corresponding to the average deviation between user’s predicted questionnaire responses and the ground truth responses.

3.4.2 Non-eRisk studies

Table 3 shows the feature representation, tasks studied, machine learning techniques, and performance metrics of all studies included in this SLR. In this section we focus on Non-eRisk studies. We grouped these studies into the following categories with regard to the feature extraction techniques they apply (ArchRQ1):

● Bag of Words (BoW)

● Word embeddings

● TF-IDF

● BERT representations

● and other feature representations

Furthermore, it is worth noting that the machine learning methods used in these studies span various categories (ArchRQ2), including:

● Classical machine learning (ML) methods such as Support Vector Machine (SVM), Naive Bayes, Logistic Regression, etc.

● Deep learning (DL) methods, e.g., recurrent neural networks.

● Combination of different methods from classical ML and DL.

● Large language models (LLMs), e.g., BERT.

● Other approaches.

Additionally, the tasks addressed in these studies can be broadly grouped into categories such as:

● Classification

● Topic modeling

● Sentiment analysis

In terms of feature extraction techniques employed across the 21 studies, a variety of methods were utilized. Among these, three studies ( 33 , 46 , 78 ) relied on TF-IDF. Four studies, including Zhang et al. ( 16 ) Benítez-Andrades et al. ( 38 ) Villegas et al. ( 48 ), and Jiang et al. ( 44 ), opted for BERT representations. Notably, Jiang et al. ( 44 ) combined BERT with LIWC.

Moreover, Bag of Words (BoW) and various types of Word Embeddings, including GloVe ( 35 , 48 ), FastText ( 35 ), and Word2Vec ( 35 , 36 ), were widely employed as feature extraction techniques in these studies.

It is pertinent to note that some studies, like Chancellor et al. ( 79 ) and Benítez-Andrades et al. ( 38 ), did not provide comprehensive details on this aspect in their papers. Conversely, other articles adopted a more personalized approach to construct their features. For instance, some represented each data point as a vector within certain categories ( 39 , 40 ), while others used rule-based methods ( 18 ) or leveraged algorithms like decision trees ( 41 ) and topic modeling ( 42 ) to determine feature selection.

Our results show that from the 21 studies, 8 make use of classical machine learning methods, 1 uses deep learning, 5 use a combination of classical ML and DL, 4 use large-language models and 3 use other approaches.

When using classical machine learning, some studies compare different methods. For example, López Úbeda et al. ( 33 ) apply 5 different supervised machine learning models: SVM, multilayer peceptron classifier, naive bayes, decision tree and logistic regression, and Villegas et al. ( 48 ) compare naive bayes, random forest, logistic regression and SVM. Along with the classical machine learning methods, the studies apply different feature representations ranging from Bag of Words (BoW) to TF-IDF ( 33 , 78 ), up to contextualized embeddings such as BERT ( 48 ).

Other studies compared both classical machine learning as well as deep learning methods. For example, in the case of Tébar and Gopalan ( 42 ), a so-called feature fusion model that includes both deep learning (a convolutional neural network (CNN) and a BiGRU model), as well as a classical machine learning model (logistic regression classifier with handcrafted features) is used.

For the studies using transformer-based large language models, different models including the BERT ( 19 ) model and its variations have been used. For example, Benítez-Andrades et al. ( 32 ) applied five variations of the BERT model. The paper from Dinu and Moldovan ( 43 ) uses BERT, RoBERTa and XLNET, whereas Jiang et al. ( 44 ) use BERT and REALM. The work from Zhang et al. ( 45 ) focusing on different mental illnesses used the BERT model, as well as the MBERT variation.

(EvalRQ1) The performance of each study is also reported in Table 3 .

(EvalRQ2) Finally, we investigated the limitations of the proposed studies (RQ4) in order to provide a structured outlook for future work in the field.

In many cases, there were limitations in terms of the datasets. For example, Yan et al. ( 78 ) cites the limited availability of labeled data. They used a dataset of 50 posts, which they expect to be labeled correctly. Also Zhou et al. ( 34 ) mention that their study is limited by the number of collected tweets, which may result in some irrelevant topics arising from noise for their topic modeling task.

In many studies, social media data is used. The nature of such data is seen as a potential limitation for the resulting methods ( 37 ). Other studies indicated as a limitation that only one social media platform was used to gather their data ( 38 , 42 ). For example, a study from ( 35 ) points out that their work did not take into account the potential biases in the data that may exist, such as underrepresented population or lack of diverse perspectives. In addition, one of the notable constraints arises from the fundamental disparity between social media data and traditional clinical text data, often used in healthcare and medical research. Clinical records encompass detailed information on patients’ medical histories, diagnoses, treatments, and outcomes, rendering them fundamentally distinct from the informal, user-generated content prevalent on social media platforms. Several studies point out that the involvement of clinical professionals would be beneficial. For example, Choudhury ( 30 ) states that their method could be more successful with the involvement of clinicians.

Different studies rely on anonymous data, which makes it difficult to ensure a good distribution within the training data over different populations and underrepresented groups. For example, Ragheb et al. ( 62 ) sees potential to optimize the model for different use cases and populations. Manual labeling by humans is also considered a source of bias since limited information about the users writing them is available to the annotators. This limited information may not encompass the full context of the users’ lives, beliefs, or backgrounds. Annotators may make subjective judgments based solely on the content of the post, which can be influenced by their own biases and interpretations. Thus, limited context can lead to misinterpretations or mislabeling, potentially distorting the research results ( 38 ).

In the limitations, it is also discussed how texts written by laypeople and ED promotional 8 and educational materials can be hard to classify ( 34 ). This can be partly explained by the short length of texts, for example in the case of tweets, and the semantic similarity of the two types of texts.

Whereas many studies achieved good performance in terms of accuracy or f1-scores, they see a potential limitation in this matter. For example, Wang et al. ( 40 ) discusses that the validation was done only with a small sample of the data, and thus further validation is required with larger samples. In another study, the authors were concerned about the problem of overfitting ( 52 ).

4 Discussion

In this systematic literature survey we have discussed the use of machine learning and natural language processing methods for the detection of eating disorders. Our survey was conducted using the PRISMA framework ( 17 ). Our results have shown that many studies focus on the detection of anorexia, or eating disorders in general (see Figure 7 ). We have also seen that there was more work over the last couple of years, indicating a growing interest in the topic (as shown in Figure 3 ). Whereas most publications were from institutions in the USA and Spain, work from other countries including Mexico, France and Canada was also identified, as shown in Figure 4 . Nevertheless, our work has shown that most research efforts have only been applied to the English language. Given the relevance of local languages for mental health diagnostics and treatment ( 15 ), it is thus necessary for future research to address other languages. With regard to the machine learning and feature extraction methods being applied, a comparison turned out to be challenging due to the diverse nature of the datasets and approaches used. The proposed approaches were classified into different categories, including classical machine learning, deep learning, a combination of classical and deep learning, the use of large language models, as well as other approaches. Several studies used f1-score as a common measure, reaching different performances ranging from 0.67 to 0.93. Overall, having a sufficient data quality and quantity was often seen as a major limitation of the approaches. Since 2017, the eRisk challenge has included two tasks pertaining to the early detection of Eating Disorders. In both 2018 and 2019, the task involved the early detection of signs of anorexia [see e.g., Losada et al. ( 26 )]. In contrast, the 2022 iteration introduced a novel task centered on measuring the severity of eating disorders ( 27 ). This task diverged from the previous ones in that no labeled training data was supplied to participants, meaning that participants could not evaluate the quality of their models’ predictions until test time. The objective task was to assess a user’s level of eating disorder severity through analysis of their Reddit posting history.

Given the composition of both the eRisk lab and the SMHD dataset ( 50 ) predominantly with social media data, it is notable that an overwhelming majority (93%) of the studies in our analysis employ this data type. This underscores the widespread reliance on social media sources in modern research methodologies. This finding confirms the results of Zhang et al. ( 16 ) who found that among 399 papers applying NLP methods for the identification of mental health problems, 81% consisted of social media data.

It is worth mentioning that we came across two types of use cases in the studies. Many studies focus on the individual’s expression of their behavior and feelings with regard to eating disorders. Some studies, namely Choudhury ( 30 ) and Chancellor et al. ( 49 ), investigate the wording of pro-anorexia or pro-eating disorders communities on social media and online forums. Such communities promote disordered eating habits as acceptable alternative lifestyles ( 49 ). Whereas in many of the studies the technologies target support for clinical professionals, in these cases other applications such as content moderation are in the foreground.

In the realm of data collection for eating disorder research, manual labeling of datasets has been a common approach, with various strategies employed. For instance, Zhang et al. ( 45 ) relied on the voluntary efforts of 31 individuals to meticulously annotate 8554 data points encompassing 38 symptoms related to MD (Mental Disorders). Other studies took different routes, combining expert knowledge with input from non-expert annotators 9 ( 38 ), or solely relying on domain experts ( 46 ). In some cases, researchers have employed machine learning algorithms to automatically annotate their datasets and subsequently validated the results with input from human labelers ( 44 ). The majority of datasets underwent annotation by non-expert human annotators, as seen in studies conducted by ( 79 , 40 , 34 , 41 ).

Our review revealed few instances of Large Language Models (LLMs) application ( 10 , 11 , 19 , 30 , 38 , 43 , 44 , 45 , 49 , 50 , 61 , 67 , 73 , 74 , 79 , 80 ). Despite this, the rising adoption of technologies like MentalBERT ( 77 ) and MentaLLama ( 81 ), alongside traditional machine and deep learning approaches, is notable. This trend, driven by the impressive efficacy of LLMs in natural language processing, is expected to continue on. As these technologies evolve and become more accessible, we anticipate their increased utilization in this field of research, enhancing computational model accuracy and efficiency.

Based on the identified limitations in the selected studies, we infer the following focus topics that we suggest for future work in the field of using natural language processing and machine learning in ED research:

● Data Quantity and Quality: how can more high-quality data be created and shared, while respecting the ethical and privacy limitations of such sensitive data?

● Involvement of Clinical Professionals: how can machine learning engineers and clinical professionals work together more closely?

● More Diversity in Data: How can the diversity of the population in the used datasets be increased to avoid bias in the classification?

● Local Languages: How can the proposed methods be extended to local languages other than English?

In conclusion, based on the studies investigated in this literature survey, there is potential for further development and in the long-term a novel tool support for clinical professionals based on text data.

Author contributions

GM: Formal analysis, Writing – review & editing, Writing – original draft, Visualization, Investigation, Data curation. AP: Formal analysis, Writing – review & editing, Writing – original draft, Validation, Supervision, Methodology, Conceptualization. MK-B: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing.

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. The authors gratefully acknowledge the support of the Inventus Bern Foundation for our research in the field of augmented intelligence for the detection of eating disorders.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

• ^ https://icd.who.int/browse10/2019/en#/F50
• ^ Valence is a measure of the emotional intensity or positivity/negativity associated with a word.
• ^ See e.g., https://digitalreality.ieee.org/publications/what-is-augmented-intelligence
• ^ https://erisk.irlab.org/
• ^ https://www.corc.uk.net/media/1273/ede-qquesionnaire.pdf
• ^ A content or an activity that promotes or encourages eating disorders (EDs).
• ^ individuals who lack specialized domain knowledge or expertise in the subject matter.

1. Williams CL, Butcher JN, Paulsen JA. 13 - overview of multidimensional inventories of psychopathology with a focus on the mmpi-2. In: Goldstein G, Allen DN, DeLuca J, editors. Handbook of Psychological Assessment , 4th ed. Academic Press, San Diego (2019). 397–417. doi: 10.1016/B978-0-12-802203-0.00013-4

CrossRef Full Text | Google Scholar

2. Vandereycken W, Van Humbeeck I. Denial and concealment of eating disorders: a retrospective survey. Eur Eating Disord Review: Prof J Eating Disord Assoc . (2008) 16:109–14.

Google Scholar

3. Smink FR, van Hoeken D, Hoek HW. Epidemiology, course, and outcome of eating disorders. Curr Opin Psychiatry . (2013) 26:543–8. doi: 10.1097/yco.0b013e328365a24f

PubMed Abstract | CrossRef Full Text | Google Scholar

4. Mohler-Kuo M, Schnyder U, Dermota P, Wei W, Milos G. The prevalence, correlates, and help-seeking of eating disorders in Switzerland. psychol Med . (2016) 46:2749–58. doi: 10.1017/S0033291716001136

5. Galmiche M, Déchelotte P, Lambert G, Tavolacci MP. Prevalence of eating disorders over the 2000–2018 period: a systematic literature review. Am J Clin Nutr . (2019) 109:1402–13.

PubMed Abstract | Google Scholar

6. Rude S, Gortner E-M, Pennebaker J. Language use of depressed and depression-vulnerable college students. Cogn Emotion . (2004) 18:1121–33.

7. Merhbene G, Nath S, Puttick AR, Kurpicz-Briki M. Burnoutensemble: Augmented intelligence to detect indications for burnout in clinical psychology. Front Big Data . (2022) 4.

8. Moreno MA, Jelenchick LA, Egan KG, Cox E, Young H, Gannon KE, et al. Feeling bad on facebook: Depression disclosures by college students on a social networking site. Depress. Anxiety . (2011) 28:447–55. doi: 10.1002/da.20805

9. Schwartz HA, Eichstaedt J, Kern ML, Park G, Sap M, Stillwell D, et al. (2014). “Towards assessing changes in degree of depression through facebook,” In: Resnik P, Resnik R, Mitchell M. editors. Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: from linguistic Signal to Clinical Reality, (Baltimore, Maryland, USA).

10. De Choudhury M, Counts S, Horvitz E. “Predicting postpartum changes in emotion and behavior via social media,” In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. New York, NY, USA: Association for Computing Machinery. (2013). pp. 3267–76. doi: 10.1145/2470654.2466447

11. De Choudhury M, Counts S, Horvitz EJ, Hoff A. “Characterizing and predicting postpartum depression from shared facebook data”. In: Proceedings of the 17th ACM Conference on Computer Supported Cooperative Work & Social Computing. New York, NY, USA: Association for Computing Machinery). 2014). pp. 626–38. doi: 10.1145/2531602.2531675

12. Tadesse MM, Lin H, Xu B, Yang L. Detection of depression-related posts in reddit social media forum. IEEE Access . (2019) 7:44883–93. doi: 10.1109/ACCESS.2019.2909180

13. Morales M, Dey P, Theisen T, Belitz D, Chernova N. “An investigation of deep learning systems for suicide risk assessment.” In: Niederhoffer K, Hollingshead K, Resnik P, Resnik R, Loveys K editors Proceedings of the Sixth Workshop on Computational Linguistics and Clinical Psychology . Minneapolis, Minnesota: Association for Computational Linguistics (2019). p. 177–81. doi: 10.18653/v1/W19-3023

14. Just MA, Pan L, Cherkassky VL, McMakin DL, Cha C, Nock MK, et al. Machine learning of neural representations of suicide and emotion concepts identifies suicidal youth. Nat Hum Behav . (2017) 1:911–9. doi: 10.1038/s41562-017-0234-y

15. Griner D, Smith TB. Culturally adapted mental health intervention: A meta-analytic review. Psychotherapy: Theory research practice Training . (2006) 43:531.

16. Zhang T, Schoene AM, Ji S, Ananiadou S. Natural language processing applied to mental illness detection: a narrative review. NPJ digital Med . (2022) 5:46.

17. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The prisma 2020 statement: An updated guideline for reporting systematic reviews. BMJ . (2021) 372. doi: 10.1136/BMJ.N71

18. Bellows BK, LaFleur J, Kamauu AWC, Ginter T, Forbush TB, Agbor S, et al. Automated identification of patients with a diagnosis of binge eating disorder from narrative electronic health records. J Am Med Inform Assoc . (2014) 21(e1):e163–8. doi: 10.1136/amiajnl-2013-001859

19. Devlin J, Chang M, Lee K, Toutanova K. “BERT: pre-training of deep bidirectional transformers for language understanding.” In: Burstein J, Doran C, Solorio T, editors. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2-7, 2019 , vol. 1 (Long and Short Papers). Association for Computational Linguistics (2019). p. 4171–86. doi: 10.18653/V1/N19-1423

20. Mikolov T, Chen K, Corrado G, Dean J. “Efficient estimation of word representations in vector space.” In: International Conference on Learning Representations . (2013).

21. Pennington J, Socher R, Manning C. GloVe: Global vectors for word representation. In: Moschitti A, Pang B, Daelemans W, editors. Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) . Association for Computational Linguistics, Doha, Qatar (2014). p. 1532–43. doi: 10.3115/v1/D14-1162

22. Peters ME, Neumann M, Iyyer M, Gardner M, Clark C, Lee K, et al. “Deep contextualized word representations.” In Walker M, Ji H, Stent A editors. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , Volume 1 (Long Papers). New Orleans, Louisiana: Association for Computational Linguistics (2018) 2227–37. doi: 10.18653/v1/N18-1202

23. Le QV, Mikolov T. “Distributed representations of sentences and documents.” In: Xing EP, Jebara T, editors. Proceedings of the 31st International Conference on Machine Learning. Proceedings of Machine Learning Research . Beijing, China: PMLR (2014) 32(2):1188–96. Available at: http://proceedings.mlr.press/v32/le14.pdf .

24. Schuster M, Paliwal K. Bidirectional recurrent neural networks. IEEE Trans Signal Process . (1997) 45:2673–81. doi: 10.1109/78.650093

25. Pennebaker JW, Francis ME, Booth RJ. Linguistic inquiry and word count: LIWC 2001. Mahway: Lawrence Erlbaum Associates (2001) 71(2001):2001.

26. Losada DE, Crestani F, Parapar J. “Overview of eRisk: Early Risk Prediction on the Internet.” In: Bellot P, Trabelsi C, Mothe J, Murtagh F, Nie JY, Soulier L, et al editors. Experimental IR Meets Multilinguality, Multimodality, and Interaction. Cham: Springer International Publishing. (2018) 343–61.

27. Parapar J, Martín-Rodilla P, Losada DE, Crestani F. Overview of eRisk 2022: Early risk prediction on the internet. In: Experimental IR Meets Multilinguality, Multimodality, and Interaction: 13th International Conference of the CLEF Association, CLEF 2022, Bologna, Italy, September 5–8, 2022, Proceedings . Springer (2022). p. 233–56.

28. Fairburn CG, Beglin SJ. Eating disorder examination questionnaire (ede-q) Database record, APA PsycTests. (1994).

29. Joshi P, Santy S, Budhiraja A, Bali K, Choudhury M. “The state and fate of linguistic diversity and inclusion in the nlp world.” In: Jurafsky D, Chai J, Schluter N, Tetreault J, editors. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (2020) 6282–6293. doi: 10.18653/v1/2020.acl-main.560

30. De Choudhury M. Anorexia on tumblr: A characterization study. In: Proceedings of the 5th International Conference on Digital Health 2015 . New York, NY, USA: Association for Computing Machinery (2015) 43–50. doi: 10.1145/2750511.2750515

31. Yan H, Fitzsimmons-Craft E, Goodman M, Krauss M, Das S, Cavazos-Rehg P. Automatic detection of eating disorder-related social media posts that could benefit from a mental health intervention. Int J Eating Disord . (2019) 52:1150–6. doi: 10.1002/eat.23148

32. Benítez-Andrades JA, Alija-Pérez JM, García-Rodríguez I, Benavides C, Alaiz-Moretón H, Vargas RP, et al. BERT model-based approach for detecting categories of tweets in the field of eating disorders (ED). In: 2021 IEEE 34th International Symposium on Computer-Based Medical Systems (CBMS) (New York, USA: IEEE) (2021). p. 586–90. Available at: https://api.semanticscholar.org/CorpusID:236095644 .

33. López Úbeda P, Plaza del Arco FM, Díaz Galiano MC, Urena Lopez LA, Martin M. “Detecting anorexia in Spanish tweets.” In: Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2019). Varna, Bulgaria: INCOMA Ltd. (2019) 655–63. doi: 10.26615/978-954-452-056-4077

34. Zhou S, Zhao Y, Bian J, Haynos AF, Zhang R. Exploring eating disorder topics on twitter: Machine learning approach. JMIR Med Inform. (2020) 8(10):e18273. doi: 10.2196/18273

35. Aguilera J, Hernández Farías DI, Ortega-Mendoza RM, Montes-y-Gómez M. Depression and anorexia detection in social media as a one-class classification problem. Applied Intelligence . (2021) 51:6088–103. doi: 10.1007/s10489-020-02131-2

36. Spinczyk D, Bas M, Dzieciatko M, Maćkowski M, Rojewska K, Maćkowska S. Computer-aided therapeutic diagnosis for anorexia. BioMed Eng OnLine (2020) 19:53. doi: 10.1186/s12938-020-00798-9

37. Aragón ME, López-Monroy AP, González-Gurrola LC, Montes-y-Gómez M. Detecting mental disorders in social media through emotional patterns-the case of anorexia and depression. EEE Transactions on Affective Computing . (2021) 14(1):211–22. doi: 10.1109/TAFFC.2021.3075638

38. Benítez-Andrades JA, Alija-Pérez J-M, Vidal M-E, Pastor-Vargas R, Vidal ME, García-Ordás T. Traditional machine learning models and bidirectional encoder representations from transformer (BERT)-based automatic classification of tweets about eating disorders: Algorithm development and validation study. JMIR Medical Informatics (2022) 10(2):e34492. doi: 10.2196/34492

39. Ramiandrisoa F, Mothe J. Early Detection of Depression and Anorexia from Social Media: A Machine Learning Approach. In: Cantador I, Chevalier M, Melucci M, Mothe J, editors. Circle 2020 , vol. 2621 . Proceedings of the Conference CIRCLE 2020, Samatan, France (2020).

40. Wang T, Brede M, Ianni A, Mentzakis E. “Detecting and characterizing eating-disorder communities on social media.” In: Proceedings of the Tenth ACM International Conference on Web Search and Data Mining (WSDM '17) . New York, NY, USA: Association for Computing Machinery (2017) 91–100. doi: 10.1145/3018661.3018706

41. He L, Luo J. “What makes a pro eating disorder hashtag: Using hashtags to identify pro eating disorder tumblr posts and Twitter users.” In: 2016 IEEE International Conference on Big Data (Big Data) . IEEE (2016). 3977–9. doi: 10.1109/BigData.2016.7841081

42. Tébar B, Gopalan A. “Early Detection of Eating Disorders using Social Media.” 2021 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) , Washington, DC, USA (2021). pp. 193–8. doi: 10.1109/CHASE52844.2021.00042.

43. Dinu A, Moldovan A-C. “Automatic detection and classification of mental illnesses from general social media texts.” In: Mitkov R, Angelova G Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021) . Held Online: INCOMA Ltd. (2021). p. 358–66. Available at: https://aclanthology.org/2021.ranlp-1.41

44. Jiang Z, Levitan SI, Zomick J, Hirschberg J. Detection of mental health from Reddit via deep contextualized representations. In: Holderness E, Jimeno Yepes A, Lavelli A, Minard A-L, Pustejovsky J, Rinaldi F, et al editors. Proceedings of the 11th International Workshop on Health Text Mining and Information Analysis . Online: Association for Computational Linguistics (2020) 147–56. doi: 10.18653/v1/2020.louhi-1.16

45. Zhang Z, Chen S, Wu M, Zhu KQ. Symptom identification for interpretable detection of multiple mental disorders. In: Goldberg Y, Kozareva Z, Zhang Y, editos. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing . Abu Dhabi, United Arab Emirates: Association for Computational Linguistics (2022) 9970–85. doi: 10.18653/v1/2022.emnlp-main.677

46. Hwang Y, Kim H, Choi H, Lee J. Exploring abnormal behavior patterns of online users with emotional eating behavior: Topic modeling study. J Med Internet Res . (2020) 22:e15700. doi: 10.2196/15700

47. Rojewska K, Maćkowska S, Maćkowski M, Różańska A, Barańska K, Dzieciatko M, et al. Natural language processing and machine learning supporting the work of a psychologist and its evaluation on the example of support for psychological diagnosis of anorexia. Appl Sci . (2022) 12. doi: 10.3390/app12094702

48. Villegas MP, Errecalde ML, Cagnina LC. “A comparison of text representation approaches for early detection of anorexia.” In: Memorias del Congreso Argentino en Ciencias de la Computación - CACIC 2021 , Workshop: WBDMD - Base de Datos y MinerÍa de Datos. (2021). 301–10.

49. Chancellor S, Pater JA, Clear T, Gilbert E, De Choudhury M. #thyghgapp: Instagram content moderation and lexical variation in pro-eating disorder communities (New York, NY, USA: Association for Computing Machinery). CSCW . (2016) 16:1201–13. doi: 10.1145/2818048.2819963

50. Cohan A, Desmet B, Yates A, Soldaini L, MacAvaney S, Goharian N. SMHD: a large-scale resource for exploring online language usage for multiple mental health conditions. In: Bender EM, Derczynski L, Isabelle P editors. Proceedings of the 27th International Conference on Computational Linguistics, Santa Fe, New Mexico, USA. Association for Computational Linguistics (2018). 1485–97. Available at: https://aclanthology.org/C18-1126

51. Wang Y-T, Huang H-H, Chen H-H. “A neural network approach to early risk detection of depression and anorexia on social media text.” In: Conference and Labs of the Evaluation Forum (CLEF) . Aachen, Germany: CEUR-WS.org (2018). Available at: https://api.semanticscholar.org/CorpusID:51940589

52. Paul S, Jandhyala SK, Basu T. Early detection of signs of anorexia and depression over social media using effective machine learning frameworks. In: Conference and Labs of the Evaluation Forum (CLEF) (2018). Available at: https://api.semanticscholar.org/CorpusID:51942457 .

53. Trotzek M, Koitka S, Friedrich CM. Word embeddings and linguistic metadata at the clef 2018 tasks for early detection of depression and anorexia. In: Conference and Labs of the Evaluation Forum (CLEF) . (2018). Available at: https://api.semanticscholar.org/CorpusID:51939971 .

54. Ramiandrisoa F, Mothe J, Benamara F, Moriceau V. IRIT at e-Risk 2018. In: 9th Conference and Labs of the Evaluation Forum, Living Labs (CLEF 2018), Avignon, France: CEUR-WS.org. (2018). pp. 1–12. Available at: https://hal.science/hal-02290007 .

55. Ortega-Mendoza RM, López-Monroy AP, Franco-Arcega A, Montes-y-Gómez M. PEIMEX at eRisk2018: Emphasizing personal information for depression and anorexia detection. In: Conference and Labs of the Evaluation Forum (CLEF) . (2018). Available at: https://api.semanticscholar.org/CorpusID:51939864 .

56. Ragheb W, Moulahi B, Azé J, Bringay S, Servajean M. Temporal mood variation: at the CLEF eRisk-2018 tasks for early risk detection on the internet. In: CLEF 2018 - Conference and Labs of the Evaluation Forum . Avignon, France. Aachen, Germany: CEUR Workshop Proceedings (2018) 2125(78). Available at: https://hal-lirmm.ccsd.cnrs.fr/lirmm-01989632/file/paper_78.pdf .

57. Liu N, Zhou Z, Xin K, Ren F. TUA1 at eRisk 2018. In: Cappellato L, Ferro N, Nie J, Soulier L, editors. Working Notes of CLEF 2018 - Conference and Labs of the Evaluation Forum, Avignon, France, September 10-14, 2018 Aachen, Germany: CEUR-WS.org. CEUR Workshop Proceedings (2018). Available at: https://ceur-ws.org/Vol-2125/paper_121.pdf .

58. Ramírez-Cifuentes D, Freire A. UPF’s participation at the clef eRisk 2018: Early risk prediction on the internet. In: Conference and Labs of the Evaluation Forum (CLEF) . (2018).

59. Funez DG, Ucelay MJG, Villegas MP, Burdisso SG, Cagnina LC, Montes-y-Gómez M, et al. UNSL’s participation at eRisk 2018 lab. In: Conference and Labs of the Evaluation Forum (CLEF) . Aachen, Germany: CEUR-WS.org (2018). Available at: https://api.semanticscholar.org/CorpusID:198489135 .

60. Aragón ME, Lopez-Monroy AP, Montes-y-Gómez M. (2019). “INAOE-CIMAT at eRisk 2019: Detecting Signs of Anorexia using Fine-Grained Emotions,” in: Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes. Aachen, Germany: CEUR-WS.org. Available at: https://api.semanticscholar.org/CorpusID:198489135 .

61. Burdisso SG, Errecalde ML, Montes-y-Gómez M. UNSL at eRisk 2019: a Unified Approach for Anorexia, Self-harm, and Depression Detection in Social Media. In: Cappellato L, Ferro N, Losada DE, Muller H. Conference and Labs of the Evaluation Forum (CLEF) . Aachen, Germany: CEUR-WS.org (2019). Available at: https://api.semanticscholar.org/CorpusID:198490018 .

62. Ragheb W, Azé J, Bringay S, Servajean M. Attentive Multi-stage Learning for Early Risk Detection of Signs of Anorexia and Self-harm on Social Media. In: Cappellato L, Ferro N, Losada DE, Müller H, editors. CLEF 2019 - Conference and Labs of the Evaluation Forum , vol. 2380 . CEUR Workshop Proceedings, Lugano, Switzerland (2019).

63. Fano E, Karlgren J, Nivre J. “Working Notes of CLEF 2019 - Conference and Labs of the Evaluation Forum, Lugano, Switzerland, September 9-12, 2019.” Vol. 2380. Cappellato L, Ferro N, Losada DE, Muller H. editors. Vol. 2380. Aachen, Germany: CEUR Workshop Proceedings (2019).

64. Masood R, Ramiandrisoa F, Aker A. “UDE at eRisk 2019: Early risk prediction on the internet.” In: Cappellato L, Ferro N, Losada DE, Müller H, editors. Working Notes of CLEF 2019 - Conference and Labs of the Evaluation Forum, Lugano, Switzerland, September 9-12, 2019 Aachen, Germany: CEUR Workshop Proceedings (2019). 2380.

65. Naderi N, Gobeill J, Teodoro D, Pasche E, Ruch P. A baseline approach for early detection of signs of anorexia and self-harm in reddit posts, in: Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes, Lugano, Switzerland: CEUR-WS.org. (2019). CEUR Workshop Proceedings.

66. Mohammadi E, Amini H, Kosseim L. Quick and (maybe not so) easy detection of anorexia in social media posts. Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes (2019). CEUR Workshop Proceedings.

67. Plaza del Arco FM, López-Úbeda P, Díaz-Galiano MC, Ureña López LA, Martín Valdivia MT. “Integrating UMLS for Early Detection of Signs of Anorexia,” In: Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes. Aachen, Germany: CEUR-WS.org. (2019). Available at: https://api.semanticscholar.org/CorpusID:198489706 .

68. Ranganathan A, Haritha A, Thenmozhi D, Aravindan C. “Early detection of anorexia using rnn-lstm and svm classifiers,” In: Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes. Aachen, Germany: CEUR-WS.org. (2019). Available at: https://api.semanticscholar.org/CorpusID:198488874 .

69. Ferdowsi S, Knafou J, Borissov N, Vicente Alvarez D, Mishra R, Amini P, et al. Deep learning-based risk prediction for interventional clinical trials based on protocol design: A retrospective study. Patterns . (2023) 4:100689. doi: 10.1016/j.patter.2023.100689

70. Trifan A, Oliveira JL. (2019). “BioInfo@UAVR at eRisk 2019: Delving into Social Media Texts for the Early Detection of Mental and Food Disorders,” In: Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes. Aachen, Germany: CEUR-WS.org. Available at: https://api.semanticscholar.org/CorpusID:198488663 .

71. Ortega-Mendoza RM, Irazú D, Farías H, Montes-Y-Gómez M. “LTL-INAOE's Participation at eRisk 2019: Detecting Anorexia in Social Media through Shared Personal Information.” In: Cappellato L, Ferro N, Losada DE, Müller H. editors. Working Notes of CLEF 2019 - Conference and Labs of the Evaluation Forum. CEUR Workshop Proceedings Vol. 2380. Lugano, Switzerland, September 9-12, 2019. CEUR-WS.org (2019). Available at: https://ceur-ws.org/Vol-2380/paper_75.pdf .

72. Hosseini Saravani SH, Normand L, Maupomé D, Rancourt F, Soulas T, Besharati S, et al. Measuring the severity of the signs of eating disorders using similarity-based models. CLEF (Working Notes) (2022). 936–46.

73. Mármol-Romero AM, Jiménez-Zafra SM, Plaza-Del-Arco FM, Molina-González MD, Martín-Valdivia M-T, Montejo-Ráez A. “SINAI at eRisk@CLEF 2022: Approaching Early Detection of Gambling and Eating Disorders with Natural Language Processing.“ In: Faggioli G, Ferro N, Hanbury A, Potthast M editors. Proceedings of the Working Notes of CLEF 2022 - Conference and Labs of the Evaluation Forum. CEUR Workshop Proceedings (2022) . Bologna, Italy. September 5th - to - 8th, 2022. CEUR-WS.org. 3180:961–71. Available at: https://ceur-ws.org/Vol-3180/paper-76.pdf .

74. Srivastava H, Lijin NS, Sruthi S, Basu T. “Exploring the potential of bag of words, document embeddings and transformer based framework for early prediction of eating disorder, depression and pathological gambling over social media.” In: Faggioli G, Ferro N, Hanbury A, Potthast M, editors. Proceedings of the Working Notes of CLEF 2022 - Conference and Labs of the Evaluation Forum, Bologna, Italy, September 5th - to - 8th Aachen, Germany: CEUR Workshop Proceedings (2022). p. 972–86.

75. Li W, Han J, Pei J. CMAR: accurate and efficient classification based on multiple classassociation rules, in: Proceedings 2001 IEEE International Conference on Data Mining. San Jose, CA, USA: IEEE. (2001). 369–76. doi: 10.1109/ICDM.2001.989541

76. Guu K, Lee K, Tung Z, Pasupat P, Chang M-W. REALM: retrieval-augmented language model pre-training. In: Proceedings of the 37th International Conference on Machine Learning. ICML'20. JMLR.org . (2020).

77. Ji S, Zhang T, Ansari L, Fu J, Tiwari P, Cambria E. MentalBERT: Publicly Available Pretrained Language Models for Mental Healthcare. In: Calzolari N, Béchet F, Blache P, Choukri K, Cieri C, Declerck T, et al editors. Proceedings of the Thirteenth Language Resources and Evaluation Conference . Marseille, France: European Language Resources Association (2022). 7184–90. Available at: https://aclanthology.org/2022.lrec-1.778 .

78. Yan H, Phd EEF-C, Goodman M, Krauss M, Das S, Cavazos-Rehg P. (2019). doi: 10.1002/eat.23148

79. Chancellor S, Kalantidis Y, Pater JA, De Choudhury MD, Shamma DA. Multimodal Classification of Moderated Online Pro-Eating Disorder Content, In: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. New York, NY, USA: Association for Computing Machinery. (2017). 3213–26. doi: 10.1145/3025453.3025985

80. Burdisso SG, Errecalde M, Montes-y-Gómez M. A text classification framework for simple and effective early depression detection over social media streams. Expert Syst Appl . (2019) 133:182–97. doi: 10.1016/j.eswa.2019.05.023

81. Yang K, Zhang T, Kuang Z, Xie Q, Ananiadou S. Mentalllama: Interpretable mental health analysis on social media with large language models. arXiv . (2023) arXiv preprint arXiv:2309.13567.

Keywords: natural language processing, machine learning, eating disorders, mental health, artificial intelligence, anorexia, bulimia, binge eating

Citation: Merhbene G, Puttick A and Kurpicz-Briki M (2024) Investigating machine learning and natural language processing techniques applied for detecting eating disorders: a systematic literature review. Front. Psychiatry 15:1319522. doi: 10.3389/fpsyt.2024.1319522

Received: 11 October 2023; Accepted: 05 March 2024; Published: 26 March 2024.

Reviewed by:

Copyright © 2024 Merhbene, Puttick and Kurpicz-Briki. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Mascha Kurpicz-Briki, [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Table of Contents

Introduction, different clinical presentations of pvd, most commonly found psychological traits, how frequent is pvd in singers, personal experience with opera singers, case report, acknowledgement.

Clarós P, Karlikowska A, Clarós-Pujol A, Clarós A, Pujol C (2019) Psychogenic Voice Disorders Literature Review, Personal Experiences with Opera Singers and Case Report of Psychogenic Dysphonia in Opera Singer. Int J Depress Anxiety 2:015. doi.org/10.23937/2643-4059/1710015

REVIEW ARTICLE | OPEN ACCESS DOI: 10.23937/2643-4059/1710015

Psychogenic voice disorders literature review, personal experiences with opera singers and case report of psychogenic dysphonia in opera singer, pedro clarós 1* , agata karlikowska 1,2 , astrid clarós-pujol 1 , andrés clarós 1 and carmen pujol 1.

1 Clarós Clinic Barcelona, Spain

2 Scholarship Clarós Clinic, Cracow, Poland

The point of this article is to make a diagnosis of psychological voice disorders easier by reviewing germane to the subject literature. Current view on terminology, classification, clinical manifestation and underlying psychological background of this rare condition is given. Secondly our aim is to asses prevalence ratio of psychological voice disorders in a group of 1520 professional opera singers-people with the most challenging voice effort among professional voice users. Our findings contradict common belief of high occurrence rate of this disorder among opera singers. Characteristics of this professional group are discussed and a short example case report is described. Based on literature findings and authors personal experience we propose an outline of an assessment protocol when psychological voice disorder is suspected. The importance of multidisciplinary diagnosis involving laryngologist, psychologist, speech therapist and possibly a psychiatrist is pointed out.

Psychological voice disorders, Opera singers, Psychogenic dysphonia, Psychogenic aphonia, Multidisciplinary diagnosis

A voice disorder is diagnosed when quality of voice, loudness or pitch of sound is insufficient for communication or inappropriate for person's age, gender or cultural background [ 1 ].

According to American Speech-Language-Hearing Association (ASHA) voice disorder is also present when the patient is having concerns about his voice even if others do not perceive it as abnormal [ 2 ].

Organic speech or voice disorder has structural or neurological components that cause the speech disturbance (e.g. vocal nodules, polyps, hematoma of vocal folds, structural changes in the larynx due to aging, vocal tremor, spasmodic dysphonia, or paralysis of vocal folds, among others).

On the contrary, a functional speech disorder is a voice impairment that is caused by underlying psychological process with no organic pathology (or a non-severe one which doesn't justify the intensity of the symptoms) of the vocal folds not associated with neurological illness, or if the vocal folds lesions are secondary to the functional disorder.

This group could be divided into two subgroups of pathologies: Muscle tension voice disorder (MVD) and psychogenic voice disorder (PVD) [ 3 ].

Muscle tension voice disorder (also called hyperfunctional voice disorder) is a phonation impairment which develops over time due to psychological process which causes disbalance of phonatory muscles. Examination reveals: Excessively effortful, forced voice during all types of sound, as well as rare phonatory breaks and involvement of the false vocal folds which can in extreme cases lead to ventricular phonation. There are two subtypes of MTVD: with no pathology to the vocal folds (MTVD1) or with secondary, developed overtime organic changes (MTVD2) such as vocal nodules, contact ulcers or polyps due to muscle hyperactivity. This article focus is on the psychogenic voice disorders group. Psychogenic voice disorder (PVD) is a voice impairment (foremost dysphonia or aphonia) caused by pathological psychological process with absence of organic structural or neurological pathologies. It's main characteristics include: Existence of symptom incongruity as the patient would manifest severe dysphonia but preserve non-verbal sounds of vegetative behavior, such as coughing, laughing, or crying with normal glottic closure and reversibility of the voice pathology [ 4 ]. We present review of literature alongside with an interesting case of psychogenic dysphonia in opera singer.

Presentations of psychogenic voice disorders may include different symptoms and with myriad of underlying psychological disorders causing them they form a very heterogeneous group.

One of the most common manifestations of PVD is dysphonia or aphonia. The disorder can be explained that an individual who doesn't have sufficient ways of coping with stress or difficult life situations may lose voluntary voice control because of intrinsic and extrinsic laryngeal muscles imbalance.

Recent belief is that this disequilibrium could be caused by either a conversion reaction where psychological problem is 'converted' into a somatic manifestation, or high level of emotional distress due to difficult life events or psychological conflicts for example over fear of expressing negative emotions [ 5 ].

In psychogenic dysphonia, voice quality is deteriorated, voice may be rough, breathy, asthenic, low or high pitched and hoarse or strained, less commonly diplophonia may be present (two different tones). Aphonia usually presents itself as an asthenic or normal whisper or rarely as a total inability to produce sound [ 6 ]. Patient usually preserves ability of normal glottic closure during cough. Individual may also present normal or better phonation accidentally, involuntarily when he or she doesn't realize it. Like mentioned before this may occur during coughing or making other vocal, non-verbal sounds for example crying or even gurgling water [ 7 ]. Patient may also find counting or singing easier than conducting a conversation.

The course of PVD can be highly variable, usually the onset is sudden, but it can also develop gradually over hours or days. In this type of disorder episodes of normal voice can occur intermittent with dysphonia depending on the patient's emotional attitude [ 8 ].

On examination of the larynx by an ENT (with video-laryngoscopy or laryngo-stroboscopy) PVD may present itself as anterior or posterior constriction of the aryepiglottic folds and/or hyper or hypoactivity of vocal folds. Also, paradoxical vocal folds dysfunction can occur or false vocal folds (vestibular vocal folds) overactivity which in extreme cases may lead to ventricular phonation. Muscle misuse may lead to quick vocal fatigue and weak voice [ 1 ]. Patient often presents himself as very anxious, agitated with untypical facial expression, excessive face movements. Typically he or she may be experiencing increased stress level or suppressed anger, sometimes poor sex identification or suffer from dysthymia/depression [ 9 ]. Other associated somatoform pathologies can be: Habitual cough or hyperventilation syndrome. Globus sensation is very common and can be reported by the patient as inability of swallowing, foreign object sensation or even dyspnea. The diagnosis should be confirmed by psychological and speech therapy assessments. Occasionally PVD may resolve spontaneously, it more commonly remains for longer periods, sometimes for months, or even for years. In most cases it is alleviated with one voice therapy session; however, it is important to include longer psychological treatment. Even after successful therapy reoccurrences are common and may appear couple of times in the future [ 10 ].

Other type of PVD is puberphonia or mutational falsetto which can appear either in an adult male or adolescent boy for an unusually long period of time. In this condition the voice characteristics are: High-pitched (usually a full octave or more above the normal pitch) asthenic, breathy, falsetto voice which can be intermittent by a normal voice [ 1 ].

Some authors believe that other rare manifestations may be: A foreign accent syndrome which means developing foreign speech prosody without actually changing geographical surroundings; and childish voice in adults, where voice is usually of higher pitch, and childlike pattern, with no other than psychological reasons [ 11 ].

Mutism or selective mutism disorder is characterized by the fact that the patient usually just mouths words without actually making attempt of using voice when patient with elective mutism consciously chooses not to speak in some domains but can normally speak in others [ 12 ].

Another type of pathology, marked out only by some of the authors, is psychogenic adductor spasmodic dysphonia (ADSD). Voice presentation resembles neurological ADSA and could be strained or asthenic, with frequent or irregular phonatory breaks, sometimes a voice tremor may be present [ 1 ].

Studies about psychogenic implication in the pathogenesis of psychogenic voice disorders reveal significant depressive symptoms, elevated stress level, anxiety and adjustment disorders as well as eating disorders, hypochondriasis or hysteria.

In his study Willinger, et al. [ 13 ] tested severity of depression and anxiety in 61 patients with functional dysphonia, and they concluded that as much as 33% of the patients showed mood disorder symptoms and 20% of them were overly anxious compared to the non-dysphonic control group matched by age, sex and occupation. This study also showed increased concern regarding health issues than other reasons for anxiety. Interesting is the fact that those symptoms seem to worsen after appearance of voice disorder, therefore we could argue: What is the cause and what is the effect.

The same author conducted a study where he assessed personality aspects as: novelty seeking, harm avoidance, reward dependence and persistence (based on Cloninger's personality model) and their connection to psychological voice disorders. The results showed that harm avoidance scores were significantly elevated in comparison to control group but no meaningful differences in novelty seeking, reward dependence or persistence appeared [ 14 ].

Different comparative study made by Kotby, et al. [ 15 ] reviewed degree of severity of anxiety symptoms in 100 patients diagnosed with functional voice disorders and 50 normal individuals (control group). The result in patients with functional disorders reached 43% of individuals but only 6% had elevated anxiety level in healthy control group. It also concluded that other often associated psychological traits are: Hypochondriasis (especially in puberphonia), somatization, hysteria, distress, and psychogenic pain disorder.

Martins, et al. [ 16 ] gives some important clinical characteristics in his case series report. His findings included profession (the most common was a housekeeper) type of onset (most common was sudden onset) and the most common manifestation (which was conversion aphonia alongside with muscle tension disorder and intermittent voicing). The prevalence rate of women was 26: 2 in a not a very numerous group of 28 patients. Tezcanez, et al. when examined the factor of patient's employment found the most common was also housewife 31% before retired 18% [ 17 ].

On the other hand Reiter, et al. assessed that the most common profession to develop this malady are business related, highly stressful positions like businessman/businesswoman [ 18 ].

Another interesting study about prevalence of orthorexia nervosa among the performance artists (in the State Opera and Ballet and in the Bilkent University Symphony Orchestra) was submitted by Aksoydan, et al. [ 19 ]. This study showed increased frequency in artists of orthorexia than control group with highest prevalence of eating disorders recorded among opera singers (81.8%).

Etiology of this disorder was described by Andersson, et al. [ 20 ] she examined the most common type of underlying reasons to this pathology. Her findings revealed that interpersonal conflicts related firstly to family and secondly to work are the most common causes of distress. She also believes that cognitive and behavioral therapy alone is not sufficient for PVD patients as it should be accompanied by voice therapy and therapeutic counselling.

The therapeutic approach have changed a lot during years. The outdated methods were quite violent and aimed at regaining voice at first voice therapy session. Worth mentioning are historic methods like obstructing the larynx until patient would feel a need to scream, applying electrical impulses or irritating the larynx by powder or fluid or cocaine solution [ 21 ].

Big longitudinal study on a group of 500 patients took place between year 1972 and 2004. Patients were treated for psychogenic aphonia with more suitable methods of treatment: Respiration, relaxation and phonetic exercise, inhalation phonation, gargle and chewing. However, in the study authors strongly point out the importance of quick resolution of symptoms preferably during the first voice therapy session [ 22 ], which seems not to be accurate. At present recommended approach is psychodynamic-systemic therapy helping the patient understand the essence of the underlying problem. It is no longer believed that aphonia should be alleviated as soon as possible but rather in the patient's own time [ 21 ].

Sixty-eight patients took part in an uncontrolled study conducted by Martinez, et al. [ 23 ]. Its aim was to determine how voice therapy affects not only voice parameters but also what is the change in anxiety and depression symptoms before and after voice therapy. The result was that not only it improves quality of the voice but also alleviates (in patients self-evaluation) symptoms of anxiety and depression.

The data from literature put prevalence rate of all voice disorders in general adult population between 6.6% at the time of the study [ 24 ] to 7.6% in the last 12 months [ 25 ]. It is believed that voice disorder's lifetime occurrence rate oscillates around 29.1% [ 26 ]. Some researchers estimate that the prevalence rate of PVD is 0.4% in general population [ 21 ] which corresponds to reports where 10-40% of all voice disorders have some kind of underlying or coexisting functional cause [ 2 ]. Psychological voice disorders are more often in women with a predominance varying between different studies. Baker, et al. estimated that ratio at 8: 1 by reviewing associated literature [ 27 ]. It was also conducted that the period of life with the highest professional voice activity (30-50 years old) would be the most endangered with developing this illness [ 24 ].

Singers were reported 2.43% among all voice patients in a study of occupational risks conducted by Verdolini, et al. [ 28 ], but Titze, et al. [ 29 ] found that 11.5% of the clinical voice population was composed of singers, and there was a high representation of non-classical singers.

In other study, Pestana, et al. [ 30 ] conducted a meta-analysis of the prevalence of self-reported voice disorders in singers. The result went as high as 46%, far superior to 18.8% in control group. Like the author says it can be suspected that singers have more concerns about their voice than the rest of the population, still the result is surprisingly high.

Within professional voice users, opera singers are concerned the ones with the most challenging use of voice. Furthermore, the stress level in this group is estimated high above average. The common belief is that they are more susceptible to voice disorders. However, our findings contradict this idea.

Clarós Clinic has been the reference center for voice care for many opera singers of several opera houses all over the world since more than 40 years. One of them is the Barcelona's opera house, the Gran Teatre del Liceu. During that time more than 1520 singers were treated in our clinic due to different voice impairments both organic and functional. Within this large group only 34 singers were, to our knowledge, diagnosed with a psychogenic voice disorder: 26 with psychogenic dysphonia (with various types of voice impairment) and 8 with psychogenic aphonia.

Among these 34 patients: 20 were sopranos (58.8%) 5 were mezzo-sopranos (14.8%), 3 were countertenors (8.8%), 3 were tenors (8.8%), 2 were baritones (5.9%) and 1 was contralto (2.9%). No bass singer was diagnosed with this disorder.

Number of women was 26 singers (76.5%) and there was 8 male singers (23.5%). From this data it can be concluded that female to male prevalence ratio was 3.25: 1 and that prevalence rate of psychogenic voice disorder, in the group of opera singers with all of the voice disorders, over the last 40 years was 2.2%. Also, more susceptible to developing PVD are singers with high tessitura as than with low.

There are no medical records to report vocal fatigue, alcohol, drugs or tobacco use in any of the 34 singers. During their examinations, no substantial organic vocal pathology was discovered and neither were any symptoms of the inflammatory disease of the larynx.

Patients did not use routinely any medications, which may alter voice emission, for example, medication for thyroid function, hormones, and hypertension or diabetes medicaments. No history of relevant neurological disease (for example Parkinson disease or myasthenia gravis) was reported.

However, there are data that 24 of the 34 were taking antipsychotic drugs to treat the anxiety, in a routine way (Diazepam, Fluoxetine, Sertraline, Paroxetine, Citalopram, Escitalopram, Lorazepam, Alprazolam or Clonazepam).

A 52-years-old baritone opera singer, from Germany, expert on Wagner's operas, suddenly presented loss of power of his voice, just at the end of the second act of the performance. He was the lead male singer in Siegfried opera written by Wagner and played at the Opera House in Barcelona. During overture after the second act (which takes 15 min) his symptoms intensified, and he presented with weak, strained and breathy voice. The famous opera singer was terrified and wanted to cancel the rest of his performance.

Medical examination of the larynx showed no organic lesions on his vocal folds, no secretion was present. The color of the mucosa was normal. Normal movement of vocal folds, their closure and phonatory function was correct. Basic test of coughing was performed-patients cough showed no pathological features. On further examination slight muscular tenderness was present in the back of the neck and the perilaryngeal area.

Patient presented difficulty in communicating, spontaneous anomalous movements of the eye lids and lips, and autonomic symptoms of forehead sweating and face flushing. When asked, he admitted to having dry mouth and feeling weak. Time was running out, so the next question was if it ever had happened to him before. His answer was yes, 2 years previous to the incident. His personal situation at the time was difficult as he had been in the process of getting a divorce with his first wife at that time.

When asked about his present marital situation he admitted to having a crisis with his second wife and that the word "divorce" was used on that very day. After having realized that fact his voice improved and after a few voice and phonatory exercises he resumed with his professional activities and gave a good performance in the third act.

A provisional diagnosis of psychogenic voice disorder was made in the form of mild dysphonia. On the subsequent day he started psychological treatment. His evaluation confirmed the diagnosis. No previous psychiatric problems were referred before. Patient denied use of tobacco, alcohol, drugs or any new medication. His voice use (the number of hours a day practicing singing) was not significantly different from standard at the time. There were no changes in work conditions or significant deterioration of atmosphere in his work environment found.

More detailed history revealed that he was a son of a protestant priest who very strictly condemned his first divorce causing a lot of distress to the patient. His second wife was much younger than him and a very attractive woman and many differences occurred between them in everyday life.

His anxiety level was assessed using Hamilton Anxiety rating scale HAM-A with a score pointing to moderate/severe anxiety level. Alongside with the long-term psychological treatment he received help from the speech therapist and there were no recurrences so far.

His family life also improved, and he is a father of a one-year-old girl at present. He modified his professional career and now he is running his master class sessions as a voice conduct teacher.

Interesting is also the fact that his dysphonia begun in the third scene of the second act of the Siegfried opera where a female character (Brunhilde) appears.

Making a diagnosis of patient with a PVD is a difficult process, especially with professional singers. As shown in the example above sometimes it is essential to keep in mind other than organic causes of voice disorders. Authors would suggest that the diagnosis process should be conducted in the following way: every patient with a voice complaint should undergo a voice assessment protocol. Primarily she or he should be examined by an ENT specialist using 70 or 90 degree rigid endoscope or fiberscope with HD camera with the possibility of recording. Examination should be conducted during articulation of/i/a and singing of low and high-pitched sounds. As doctor performs videolaryngoscopy (VLS) or video fiberolaryngoscopy he can detect most organic causes of voice disorders.

Coexisting medical conditions should be diagnosed and treated (for example GERD or rhinosinusitis).

Next a videostroboscopy or high-speed digital imaging (HSDI) should be performed to asses if the mobility of the vocal cords is symmetric, their amplitudes and mucosal waves normal. Vocal folds closure should be complete and their movement regular. Videostroboscopy together with a manual examination of neck in the larynx area (which can reveal muscle tension and pain) allows determining and excluding MVD group of disorders. Those patients should be referred to speech therapist and closely monitored by an ENT specialist for any occurring vocal folds lesions.

The remaining group of patients is likely to have a psychogenic voice disorder. A simple test of cough can help detect the hysteric origin of the process. Before referring patient to psychologist, the speaking voice should be assessed in regard to grade of dysphonia (severity of the disorder) roughness, breathiness, asthenia, strain (GRBAS scale and/or VHI scale) Pitch and voice range should be defined (using a Bruel & Kjaer stroboscope or High-Speed Digital Imaging-HSDI). High definition recordings of the voice should be made.

Breathing patterns should be assessed and if necessary improved by learning to activate and loosen muscles breathing to improve resonance of the voice.

Thorough psychological evaluation should be conducted, aimed to asses stress and anxiety levels using standardized scales (e.g. Hamilton Anxiety rating scale HAM-A Hamilton Anxiety Rating Scale for Depression HAM-D, Beck depression Inventory BDI, State-Trait Anxiety Inventory STAI, Hospital Anxiety and Depression Scale HADS scale) and also psychoanalysis should define if any stressful life event occurred or if the patient is suffering from conversion reaction or any other psychological conflict (e.g. over speaking out). If necessary also a thorough psychiatric examination should be performed.

In contrast to the common belief prevalence rate of voice disorders in opera singers in not higher than other professional voice users: It was 2.2% in the group of 1520 opera singers over the last 40 years. In authors opinion they owe it to increased (compared to other people) voice care. It is important to point out that professional opera singers are a very specific group of people, very reluctant to give out any personal detail, especially regarding their voice disorders or difficulties in private life. For this reason, we believe, some pathologies may not be routinely communicated to the general practitioners.

Psychological voice disorder is a very rare and complicated illness which is often misdiagnosed and therefore inadequately treated. Both the diagnosis and the treatment of this condition should be multidisciplinary involving speech therapist, psychologist sometimes psychiatrist and a laryngologist, as each of the specialists has an important role in the process. Establishing diagnostic criteria not only for voice assessment but also for psychological diagnosis would help to improve the process. Making a correct diagnosis of PVD is a task for a group of specialists but also a disorder which mustn't be forgotten by general practitioners.

The authors report no conflicts of interest. The authors report no financial and material support for the research and the work reported in the manuscript.

• Baker J, Ben-Tovim DI, Butcher A, Esterman A, McLaughlin K (2007) Development of a modified diagnostic classification system for voice disorders with inter-rater reliability study. Logoped Phoniatr Vocol 32: 99-112.
• Andrea M, Andrea M, Figueira ML (2018) Self-perception of quality of life in patients with functional voice disorders: The effects of psychological and vocal acoustic variables. Eur Arch Otorhinolaryngol 275: 2745-2754.
• Ludlow CL, Domangue R, Sharma D, Jinnah HA, Perlmutter JS, et al. (2018) Consensus-based attributes for identifying patients with spasmodic dysphonia and other voice disorders. JAMA Otolaryngol Head Neck Surg 144: 657-665.
• Schalén L, Andersson K, Eliasson I (1992) Diagnosis of psychogenic dysphonia. Acta Otolaryngol Suppl 112: 110-112.
• Baker J (2016) Functional voice disorders: Clinical presentations and differential diagnosis. Handb Clin Neurol 139: 389-405.
• Mathieson L (2013) Greene and mathieson's the voice and its disorders, Hoboken.
• House A, Andrews HB (1987) The psychiatric and social characteristics of patients with functional dysphonia. J Psychosom Res 31: 483-490.
• Butcher P (1995) Psychological processes in psychogenic voice disorder. Eur J Disord Commun 30: 467-474.
• Baker J (2008) The role of psychogenic and psychosocial factors in the development of functional voice disorders. Int J Speech Lang Pathol 10: 210-230.
• Roy N, Bless DM (2000) Personality traits and psychological factors in voice pathology: A foundation for future research. J Speech Lang Hear Res 43: 737-748.
• Keulen S, Verhoeven J, De Witte E, De Page L, Bastiaanse R, et al. (2016) Foreign accent syndrome as a psychogenic disorder: A review. Front Hum Neurosci 10: 168.
• Rubin J (2016) The professional voice user. In: Declan Costello, Guri Sandhu, Practical Laryngology, CRC Press, Boca Raton, FL, 19-26.
• Willinger U, Völkl-Kernstock S, Aschauer HN (2005) Marked depression and anxiety in patients with functional dysphonia. Psychiatry Res 134: 85-91.
• Willinger U, Aschauer, HN (2005) Personality, anxiety and functional dysphonia. Personality and Individual Differences 39: 1441-1449.
• Kotby MN, Baraka M, El Sady SR, Ghanem M, Shoeib R (2003) Psychogenic stress as a possible etiological factor in non-organic dysphonia. Int Congr Ser 1240: 1251-1256.
• Martins RH, Tavares EL, Ranalli PF, Branco A, Pessin AB (2014) Psychogenic dysphonia: Diversity of clinical and vocal manifestations in a case series. Braz J Otorhinolaryngol 80: 497-502.
• Tezcaner ZÇ, Gökmen MF, Yıldırım S, Dursun G (2017) Clinical features of psychogenic voice disorder and the efficiency of voice therapy and psychological evaluation. J Voice 33: 250-254.
• Reiter R, Rommel D, Brosch S (2013) Long term outcome of psychogenic voice disorders. Auris Nasus Larynx 40: 470-475.
• Aksoydan E, Camci N (2009) Prevalence of orthorexia nervosa among Turkish performance artists. Eat Weight Disord 14: 33-37.
• Andersson K, Schalén L (1998) Etiology and treatment of psychogenic voice disorder: Results of a follow-up study of thirty patients. J Voice 12: 96-106.
• Kolbrunner J, Menet AD, Seifert E (2010) Psychogenic aphonia: No fixation even after a lengthy period of aphonia. Swiss med Wkly 140: 12-17.
• Maniecka-Aleksandrowicz B, Domeracka-Kołodziej A, Rózak-Komorowska A, Szeptycka-Adamus A (2006) Management and therapy in functional aphonia: Analysis of 500 cases. Otolaryngol Pol 60: 191-197.
• Martinez CC, Cassol M (2015) Measurement of voice quality, anxiety and depression symptoms after speech therapy. J Voice 29: 446-449.
• Roy N, Merrill RM, Gray SD, Smith EM (2005) Voice disorders in the general population: Prevalence, risk factors, and occupational impact. Laryngoscope 115: 1988-1995.
• Bhattacharyya N (2014) The prevalence of voice problems among adults in the United States. Laryngoscope 124: 2359-2362.
• Cohen SM (2010) Self‐reported impact of dysphonia in a primary care population: An epidemiological study. Laryngoscope 120: 2022-2032.
• Baker J (2002) Psychogenic voice disorders-heroes or hysterics? A brief overview with questions and discussion. Logoped Phoniatr Vocol 27: 84-91.
• Verdolini K, Ramig LO (2001) Review: Occupational risks for voice problems. Logoped Phoniatr Vocol 26: 37-46.
• Titze IR, Lemke J, Montequin D (1997) Populations in the US workforce who rely on voice as a primary tool of trade: A preliminary report. J Voice 11: 254-259.
• Pestana PM, Vaz-Freitas S, Manso MC (2017) Prevalence of voice disorders in singers: Systematic review and meta-analysis. J Voice 31: 722-727.

• Open access
• Published: 09 January 2024

Recurrence of post-traumatic stress disorder: systematic review of definitions, prevalence and predictors

• Samantha K Brooks 1 &
• Neil Greenberg 1  

BMC Psychiatry volume  24 , Article number:  37 ( 2024 ) Cite this article

1499 Accesses

3 Altmetric

Metrics details

Many people will experience a potentially traumatic event in their lifetime and a minority will go on to develop post-traumatic stress disorder (PTSD). A wealth of literature explores different trajectories of PTSD, focusing mostly on resilient, chronic, recovered and delayed-onset trajectories. Less is known about other potential trajectories such as recurring episodes of PTSD after initial recovery, and to date there has been no estimate of what percentage of those who initially recover from PTSD later go on to experience a recurrence. This systematic review aimed to synthesise existing literature to identify (i) how ‘recurrence’ of PTSD is defined in the literature; (ii) the prevalence of recurrent episodes of PTSD; and (iii) factors associated with recurrence.

A literature search of five electronic databases identified primary, quantitative studies relevant to the research aims. Reference lists of studies meeting pre-defined inclusion criteria were also hand-searched. Relevant data were extracted systematically from the included studies and results are reported narratively.

Searches identified 5,398 studies, and 35 were deemed relevant to the aims of the review. Results showed there is little consensus in the terminology or definitions used to refer to recurrence of PTSD. Because recurrence was defined and measured in different ways across the literature, and prevalence rates were reported in numerous different ways, it was not possible to perform meta-analysis to estimate the prevalence of recurrence. We also found no consistent evidence regarding predictors of PTSD recurrence.

A clear and consistent evidence-based definition of recurrence is urgently needed before the prevalence and predictors of recurrence can be truly understood.

Peer Review reports

Potentially traumatic events are common. Research suggests that over 70% of people will experience a potentially traumatic event (such as witnessing death or serious injury, automobile accident, life-threatening illness or injury, or violent encounter) in their lifetime [ 1 ]. Understandably, these events can be very distressing in the short-term and many people will experience acute post-traumatic symptoms in the immediate aftermath of a traumatic event, including intrusive symptoms (e.g. recurrent unwanted thoughts, nightmares); avoidance symptoms (e.g. emotional numbing, social withdrawing); hyperarousal (e.g. easily startled, feeling ‘on edge’); and physical symptoms (e.g. chest pain, dizziness) [ 2 ]. For the majority, these symptoms will decline naturally without intervention [ 3 ], typically within the first four weeks [ 2 ]. An important minority will find their symptoms persist for longer than a month. Those who continue to experience persistent re-experiencing of the traumatic event; avoidance of stimuli associated with the event; negative alterations in cognitions and mood and alterations in arousal and reactivity, causing clinical distress or functional impairment and not attributable to any other medical condition, are likely to be diagnosed with post-traumatic stress disorder (PTSD) [ 4 ]. Although only a minority of people who experience potentially traumatic events will go on to develop PTSD, it remains one of the most common mental disorders with lifetime prevalence estimated to be between 8% [ 5 ] and 12% [ 6 ]. PTSD is associated with reduced health-related quality of life and physical comorbidities, as well as major socio-economic costs [ 7 ].

The early 2000s saw a shift from studying PTSD itself as an outcome to studying change in symptoms as an outcome [ 8 ], with a wealth of studies using modelling approaches such as latent class growth analysis and latent growth mixture modelling to identify different trajectories of PTSD. Most of this literature identifies four trajectories, two of which are relatively stable trajectories ( chronic , a stable trajectory of post-traumatic stress symptoms, and resilient , a stable trajectory of healthy functioning after an adverse event), and two which display dynamic symptom patterns ( recovered , i.e. decreasing symptoms after an initial diagnosis of PTSD, and delayed-onset , i.e. increasing symptoms not meeting the diagnostic criteria for PTSD until potentially months or even years after traumatic exposure) [ 9 ]. Van de Schoot et al. [ 10 ] suggest that the two trajectories which typically occur less often (chronic and delayed-onset) are at risk of being overlooked by researchers or overwhelmed within the data by the larger trajectories. There may also be other less-researched or less-understood trajectories overlooked to an even greater extent. For example, one previous review [ 11 ] identified limited evidence of another, smaller trajectory referred to as a ‘relapsing’ or ‘recurring’ PTSD trajectory, in which individuals develop PTSD, are free from symptoms for long enough to be considered ‘recovered’, and then experience a recurrence of symptoms.

Recurrence is given relatively little attention in the PTSD literature, perhaps due to limitations of study methodologies and the complexities of studying recurrence. For example, Santiago et al. [ 11 ] note that few studies of PTSD follow participants for more than a year or with more than two assessments. Clearly, it would not be possible for researchers to identify recurrence of PTSD if data is only collected for two time-points: the only possible outcomes would be low symptom levels at each time-point (‘resilience’), high symptoms at each time-point (‘chronic’), or low level of symptoms at one time-point and a high level at the other (either ‘recovery’ or ‘delayed-onset’ depending on time-point at which symptoms were experienced). Additionally, studies which only follow up participants for a year or less are unlikely to clearly identify a recurrent trajectory of PTSD given the time needed to both recover and to experience a recurrent episode. The timing of PTSD assessment is also important: identification of PTSD recurrence relies on studies capturing the presence of symptoms during the recurrence, rather than before it occurs or after recurring symptoms have subsided. Therefore, it is perhaps unsurprising that the majority of the literature does not identify a ‘recurring’ trajectory of PTSD. Even studies which do identify recurrences often group these in with other trajectories: for example, Mota et al. [ 12 ] identified ‘recurrent’ cases of PTSD (individuals who had a lifetime diagnosis in 2002 and another post-2002 diagnosis reported in 2018), but grouped ‘persistent’ and ‘recurrent’ cases of PTSD together. Magruder et al. [ 13 ] identified a group of recurrent cases of PTSD – individuals who had lifetime PTSD pre-1992 but not a current diagnosis in 2002, who then had a diagnosis again in 2021, but these were grouped with ‘chronic’ cases. Karamustafalioglu et al. [ 14 ] simply include an ‘other’ group constituting both recurrent cases (individuals who met the criteria for PTSD diagnosis 1–3 months post-trauma and at the third follow-up 18–20 months post-trauma, but not at the second follow-up 6–10 months post-trauma) and others with delayed-onset PTSD which resolved. Boe et al. [ 15 ] identified a group of individuals with ‘reactivated’ PTSD who reported remission from PTSD in the first five years after the North Sea oil rig disaster of 1980 and a new episode at any point between 1985 and 2007. However, the authors suggest that there are blurred boundaries between delayed-onset and ‘reactivated’ PTSD, going on to include ‘possible delayed cases’ in their analysis of reactivated PTSD.

It is important to note that even the definitions of the more well-established trajectories of PTSD are not without their controversies. For example, Andrews et al. [ 16 ] point out the ambiguity in the criterion for delayed-onset PTSD, questioning whether ‘the onset of symptoms’ refers to any symptoms which might eventually lead to PTSD or only to full-blown PTSD itself. North et al. [ 17 ] comment on the ambiguities involved in the term remission (i.e. whether remission should be symptom-based or threshold-based) as well as the term onset (i.e. whether onset refers to first symptoms or first meeting diagnostic criteria). Definition of recovery also appears to differ from study to study, with some authors considering recovery to be symptom-based (i.e. no symptoms of the disorder remain) and others considering it to be threshold-based (i.e. some symptoms may remain, but they are beneath the diagnostic threshold) [ 18 ].

To date, several systematic reviews have been published which focus solely on only one PTSD trajectory. For example, previous reviews have focused on the delayed-onset trajectory [ 16 , 19 ]; the recovery trajectory [ 20 ]; and the resilient trajectory [ 21 ]. To date there has not been a literature review examining evidence of a recurrent trajectory of PTSD. Berge et al. [ 22 ] aimed to systematically review research on relapse in veterans but found no studies reporting actual rates of relapse or recurrence. Reviews have also explored the risk of relapse of various anxiety disorders, including PTSD, after discontinuation of antidepressants [ 23 ] and after cognitive behavioural therapy [ 24 ]. However, there have been no reviews attempting to quantify the risk of PTSD recurring, establish the predictors of recurrence, or quantify how much each predictive factor contributes to the risk of recurrence. The current review aimed to fill this gap in the literature by synthesising existing published data on how researchers define ‘recurrence’ of PTSD, recurrence rates of PTSD, and predictive factors of recurrence.

Having an appropriate understanding of recurrence is important as the concept needs to be properly understood in order to take steps to mitigate the risks of recurrent PTSD episodes. Mitigating the risk of PTSD recurring could benefit the health and wellbeing of trauma-exposed individuals and could reduce the socio-economic costs to the wider society [ 7 ]. The prevalence of recurrence is of particular importance to occupational medicine: regularly trauma-exposed organisations, for example, are often faced with decisions about when (and if) staff who have had and recovered from PTSD should return to the frontline duties. Understanding the risk of recurrent episodes may therefore have implications for those in charge of making such decisions. The present time is also a particularly relevant time to develop our understanding of recurrence of PTSD, as it is possible that the COVID-19 pandemic could contribute to recurrence. The pandemic has been declared a potential traumatic stressor, with research suggesting that COVID-19 survivors are at elevated risk of experiencing PTSD [ 25 ] and that PTSD symptoms may also develop due to quarantine [ 26 ], concerns about the health of loved ones, or economic loss as a result of the pandemic [ 27 ]. Hori et al. [ 28 ] suggest that the daily television updates regarding COVID-19 could trigger memories of surviving a previous traumatic situation, and exacerbate subthreshold PTSD symptoms. Therefore, experiencing the pandemic could potentially cause a recurrence of symptoms in people who have previously been diagnosed with PTSD.

The aim of this review was to collate literature which provides evidence of the lesser-studied ‘recurrent’ trajectory of PTSD and to identify: (i) the definitions of ‘recurrence’ used throughout the literature; (ii) prevalence of recurrence; and (iii) risk and protective factors for the recurrent trajectory of PTSD.

This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [ 29 ]. Our population of interest were people who had been diagnosed with, recovered from, and experienced a recurrence of PTSD (as diagnosed by a clinician or validated PTSD assessment tool). For the aim relating to prevalence of recurrent episodes, studies needed to involve a suitable design allowing prevalence to be assessed: for example, studies involving a population of people who had recovered from PTSD, followed over time to show how many had a recurrent episode and how many did not. For the other aims (i.e., definitions of recurrence and factors associated with recurrence), a comparison group was not necessary.

Registering the review

A protocol for the current review was developed and registered with PROSPERO on March 9th 2023 (registration number CRD42023405752). The only deviation from the protocol was the addition of another quality appraisal tool, due to finding a study design (retrospective analysis of existing health data) which we had not anticipated.

Eligibility criteria

To be included in the review, studies needed to (1) be published in peer-reviewed journals, (2) be published in the English language, (3) use quantitative methodology, (4) use a standardised tool to assess PTSD and (5) present data on recurrence rates of PTSD and/or factors associated with PTSD recurrence. There were no limitations relating to publication date or location of the studies. Case studies were excluded but there were no other exclusion criteria relating to population size.

Data searching and screening

A systematic literature search was carried out to examine definitions, prevalence rates and predictors of PTSD recurrence. Four electronic databases (Embase, PsycInfo, Medline and Web of Science) were searched on 24th November 2022, using a combination of search terms relating to PTSD, recurrence, and prevalence/predictors which were combined using Boolean operators. The full list of search terms is presented in Appendix 1 . The US Department of Veterans Affairs National Center for Post-Traumatic Stress Disorder’s PTSDPubs database (formerly PILOTS) was searched separately on the same date using the individual terms ‘recurrence’ and ‘recurrent’ and limited to peer-reviewed articles. Reference lists of articles deemed to meet the inclusion criteria were also hand-searched.

All citations resulting from the literature searches were downloaded to an EndNote library where duplicates were removed. The titles of all citations were then screened for relevance to the review, with any clearly not relevant being excluded. Abstracts were then screened for eligibility and the full texts of all remaining citations after abstract screening were located and read in their entirety to identify studies meeting all inclusion criteria. The literature searches and screening were carried out by the first author. The two authors met regularly throughout the screening process to discuss any uncertainties about inclusion or exclusion until a decision was reached.

Data extraction

The first author carried out data extraction of all citations deemed to meet the inclusion criteria. Data were extracted to a Microsoft Excel spreadsheet with the following headings: authors, year of publication, country, study design, sampling method, inclusion/exclusion criteria, study population size, socio-demographic characteristics of participants, type of trauma exposure, time-points at which PTSD was assessed, tools for assessing PTSD, definitions of recovery and recurrence, whether any PTSD treatment was received, prevalence rates of recurrence, and factors examined as potential predictors of recurrence.

Data synthesis

For the first aim of the review (relating to definitions of recurrence), we designed a table to present data relating to how ‘recurrence’ was understood and defined in each study. The tools used to diagnose and measure PTSD symptoms in the first place are important in understanding how PTSD is defined, so first the assessment tools used in each study were extracted into the table. Given that we wanted to understand the length of time an individual needs to be free of PTSD in order to be considered ‘recovered’, for each study we also included the time-points of PTSD assessment in the table. Next, we included the definitions of recovery and recurrence from each study, explained narratively in the table. We also added information to this table to report whether participants had received PTSD treatment during each study, as some studies focusing on interventions used ‘response to treatment’ in their definitions of recovery. We compared the different definitions used within the studies to establish whether there was consensus within the literature around (i) whether recovery and recurrence are symptom-based or threshold-based and (ii) how long the recovery period between initial diagnosis and recurrent episodes needs to be in order to be considered recurrent rather than chronic PTSD.

The second aim related to prevalence of PTSD recurrence. Due to the various research designs and definitions of ‘recurrence’ in the literature, as well as the different ways in which prevalence was reported, meta-analytic techniques could not be used. Rather, we presented the prevalence data as it was reported in each study. This sometimes meant presenting the prevalence of PTSD recurrence within an entire trauma-exposed population, including those who never experienced PTSD at any time. Other times, this meant presenting the prevalence of PTSD within a population who all had PTSD at one time-point, and other times this meant presenting the prevalence of PTSD within a group who had recovered from PTSD.

Finally, in order to explore factors associated with PTSD recurrence, all variables considered as potential covariates were recorded individually for each study. Each potential predictive factor was descriptively reported in a table, and any found to be significantly associated with experiencing PTSD recurrence were bolded to differentiate between non-significant and significant findings. Factors are also described narratively within the results section. Insights from thematic analysis [ 30 ] were used to group similar data together. For example, data relating to gender or age as predictors of recurrence were coded ‘socio-demographic’ and discussed together within the results.

Quality appraisal

We appraised the quality of studies using National Institutes of Health (NIH) tools: either the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies or the Quality Assessment of Controlled Intervention Studies tool, depending on study design. Concurrent with other reviews [e.g. 31 ] we rated quality as ‘poor’ if studies scored 0–4/14, ‘fair’ if they scored 5–10/14 and ‘good’ if they scored 11–14/14. One study used retrospective analysis of existing health data, and for this study we used the MetaQAT Critical Appraisal Tool [ 32 ]. To keep the ratings consistent with our rating system for the studies appraised by NIH tools, we defined ‘poor’ quality as a score of 0–34%, ‘fair’ quality as a score of 35–72% and ‘good’ quality as a score of 78% or higher.

Literature searches yielded 5,398 citations of which 1,083 were duplicates. After title and abstract screening, 4,210 citations were excluded leaving 105 citations for full-text screening. After reading full texts of the remaining citations, 75 were excluded and an additional five studies were added after hand-searching reference lists. A total of 35 citations were included in the review [ 15 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 ]. Figure  1 illustrates the screening process in a PRISMA flow diagram.

PRISMA flow diagram of screening process

Table  1 provides an overview of key characteristics of all included studies. Studies originated from the United States of America (n = 13), Denmark (n = 5), Israel (n = 4), China (n = 4), Norway (n = 2), the United Kingdom (n = 2), Japan (n = 1), the Netherlands (n = 1), Switzerland (n = 1), and Turkey (n = 1). The remaining study included participants in multiple different countries across Europe and Asia. Study populations ranged from 35 to 7,918 and included military personnel (n = 15), civilian adults (n = 14), children or adolescents (n = 4) or a combination of military and civilian adults (n = 2). Only three studies were rated as ‘good’ quality; the majority were rated ‘fair’.

Definitions of recurrence

Table  2 reports, for each study, the tools used to assess PTSD; time-points at which PTSD was assessed; definitions of recovery and recurrence; and whether the participants received PTSD treatment or not.

Terminology

The first aim of the review was to explore how ‘recurrence’ is defined in the literature. We found no consensus in terms of how this is defined. In fact, the studies used a variety of different terms to describe the emergence of new PTSD episodes after initial ‘recovery’, including ‘recurrence’ [ 33 , 37 , 44 , 47 , 64 , 65 ]; ‘relapse’ [ 35 , 36 , 40 , 49 , 50 , 52 , 53 , 57 ]; ‘reactivation’ [ 15 , 60 , 62 ]; ‘exacerbation/reactivation’ [ 61 ]; ‘relieved-worsening PTSD’ [ 34 , 48 , 51 , 63 ]; ‘response-remit’ trajectory [ 54 ]; ‘fluctuating course’ [ 58 ]; ‘intermittent cases’ [ 43 ]; ‘delayed increase in symptoms’ [ 46 ]; and the ‘relapsing/remitting’ trajectory [ 42 , 55 ]. Many others simply described recurrence as ‘symptom increase’ [ 38 ], ‘initial declines followed by symptom increases’ [ 56 ] or ‘exacerbation of symptoms’ [ 41 , 60 ]. Some studies did not name the trajectory at all; rather, they presented tables or flow charts showing the number of participants with PTSD at each time-point, from which it was possible for us to identify a sub-group of participants who were described as having PTSD at one time-point, not having it at least one follow-up, and then having it again at subsequent time-points [ 39 , 59 ]. Similarly, Hansen et al. [ 45 ] identified and commented on a sub-group of participants who met the criteria for PTSD, did not meet the criteria at a subsequent time-point, and then met the criteria again later, but they did not give this a name.

Criteria for recurrence

Several studies defined recurrence (or equivalent terminology such as relapse) as meeting diagnostic criteria for PTSD at a follow-up time-point after an initial ‘recovery’ period where they did not meet the cut-off for PTSD [ 33 , 35 , 37 , 39 , 43 , 45 , 46 , 58 , 65 ]. Holliday et al. [ 47 ] referred to ‘clinically meaningful change in PTSD symptoms’, which was also assumed to refer to clinical cut-off scores. Markowitz et al. [ 52 ] based the definition of relapse on similarity to baseline scores. Sungur and Kaya [ 64 ] defined recovery and recurrence as being asymptomatic and then symptomatic again, but it is not clear whether this referred to clinical cut-offs. One study defined ‘reactivation’ of PTSD as meeting full diagnostic criteria or being a sub-syndromal case [ 15 ]. Others were more vague and did not mention cut-offs, instead referring to dramatic or steep symptom increases [ 34 , 38 , 56 , 63 ], fluctuating symptoms [ 42 , 55 ], returning to pre-treatment levels of PTSD [ 54 ], symptoms which ‘decreased somewhat and increased drastically’ [ 48 ], symptoms which ‘decreased to a low level and increased again’ [ 49 , 50 ] or ‘steadily worsening’ symptoms [ 36 ]. DenVelde et al. [ 41 ] simply asked participants to self-report whether they had ‘experienced remissions and exacerbations’. Martenyi et al. [ 53 ] had multiple definitions of relapse, including increases in scores on their PTSD measures or ‘the clinical judgement of the investigator’. Others labelled the trajectory but did not specify the parameters of their definitions [ 51 , 60 , 61 , 62 , 66 ]. One study [ 57 ] used ‘being hospitalised’ as a proxy measure of PTSD recurrence, although this way of defining recurrence would obviously not capture individuals who developed recurring symptoms which were not severe enough to warrant hospitalisation; additionally, no criteria for hospitalisation were described. Similarly, Davidson et al. [ 40 ] described ‘relapse’ as PTSD scores reverting back to baseline or worse, or experiencing an ‘untoward clinical event’ including suicidality, hospitalisation, or dropping out of the study due to feeling progress was not being made.

We found little consensus as to how long participants needed to be symptom-free (or have reduced symptoms) in order to be considered ‘recovered’ prior to recurrence. The majority of studies simply based their definitions on the time-points of the study, suggesting that recurrence was identified if participants had PTSD at baseline, did not have PTSD during at least one follow-up, and then had PTSD again at a later follow-up. The time-points of follow-ups ranged from weeks to months to years. Only four studies suggested specific timeframes: three studies claimed that participants needed to be ‘recovered’ for eight weeks in order for later reports of PTSD to count as ‘recurrence’ rather than symptom fluctuation [ 35 , 37 , 66 ] whereas Zanarini et al. [ 65 ] reported that participants needed to be not meeting the PTSD criteria for at least two years in order to be considered ‘recovered’. Similarly, most studies did not clarify a time-scale for how long symptoms needed to be experienced in order to be considered a ‘recurrence’. Most studies again simply based their diagnosis on the scores participants happened to report on the days they were assessed. Few studies specified a time-frame: three [ 35 , 43 , 65 ] suggested a duration of four consecutive weeks of meeting their criteria for PTSD, while Benítez et al. [ 37 ] suggested two weeks of symptoms was sufficient to identify a recurrent episode.

Prevalence of recurrence

The review’s second aim was to explore PTSD recurrence rates. Table  3 presents data on the prevalence of recurrence of PTSD for each study. The second column of Table  3 presents the data that is reported in the original studies. The findings reported in this column are not easily comparable because studies reported recurrence rates in different ways. Some reported the percentage of the entire trauma-exposed sample who experienced PTSD recurrence (column 3 of Table  3 ). Others reported the percentage of those with PTSD who experienced recurrence (column 4 of Table  3 ) and the remaining studies reported the percentage of those who recovered from PTSD who experienced recurrence (column 5 of Table  3 ). Three studies [ 44 , 47 , 57 ] did not report the prevalence of recurrence, but were still included in the review as they included definitions and/or predictors of recurrence. One study [ 60 ] deliberately chose a sample who had all experienced recurrence; therefore, recurrence prevalence data for this study was not recorded in Table  3 as it would, by design, be 100%.

Most studies (19/35) reported the prevalence of recurrence within the entire trauma-exposed population. We would therefore expect prevalence rates to be extremely small, given that the majority of trauma-exposed people will not develop PTSD in the first place [ 3 ], let alone have recurrent episodes. However, in several studies this was not the case. Prevalence of recurrence ranged from 0.2% (for a sub-set of participants who did not directly witness the disaster in question) [ 45 ] to 57% of 63 women newly-diagnosed with ovarian cancer [ 43 ]. The latter study was carried out over 27 weeks and identified ‘intermittent cases’ who had PTSD at one time-point, no PTSD at a later time-point, and then PTSD again later on. We note that 27 weeks is a fairly short period of time for both recovery and recurrence to occur, and it is therefore possible that the data reflects symptom fluctuations rather than true recovery or recurrence. Overall, the mean prevalence of recurrent PTSD in trauma-exposed populations was 13.1%, and the median was 3.8%.

Five studies presented the prevalence of recurrence within populations diagnosed with PTSD. We would expect these prevalence rates to be higher than the prevalence rates of recurrence within full trauma-exposed samples, as they are based on populations who developed PTSD only. The rates were 4.9% [ 39 ], 15.4% [ 66 ], 24.5% [ 36 ], 28% [ 46 ] and 49.6% [ 41 ]. Mean and median prevalence of recurrent PTSD were both 24.5%.

Seven studies presented data on the prevalence of recurrence within sub-sets of study populations who had recovered from PTSD; therefore, the only possible trajectories for these participants would be recurrence or maintenance of recovery. Recurrence rates ranged from 5.8% (for a sub-set of participants treated with fluoxetine) [ 53 ] to 50% (for a sub-group treated with a placebo) [ 40 ]. Mean prevalence of recurrent PTSD was 25.4% and the median was 22.2%.

The three studies rated highest in quality [ 34 , 47 , 55 ] did not report similar findings relating to prevalence. Holliday et al. [ 47 ] did not present prevalence data at all. Andersen et al. [ 34 ] reported that 2% of participants followed the ‘relieved-worsening’ trajectory, whereas Osenbach et al. [ 55 ] reported that 35% of participants followed the ‘relapsing-remitting’ trajectory. Notably, Andersen et al.’s [ 34 ] participants were military personnel, whilst Osenbach et al.’s [ 55 ] participants were civilian trauma survivors. For this reason, we decided to look separately at recurrence rates in military and civilian participants. We also decided to look separately at data on children as children’s experiences during and after potentially traumatic events are likely to be distinct from those of adults [ 67 ]. Table  4 presents the mean and median recurrence rates for different populations.

Prevalence of PTSD recurrence in military populations

Fifteen studies focused on military personnel and veterans, three of which did not provide prevalence data and one of which included only participants with PTSD recurrence. Military studies which presented rates of recurrence in trauma-exposed populations (rather than focusing on people diagnosed with PTSD only) typically found low prevalence of recurrence: seven studies found prevalence rates under 4% [ 34 , 48 , 51 , 54 , 61 , 62 ]. Another study found a prevalence rate of 6% [ 38 ]. The only higher prevalence rates were reported by Solomon & Mikulincer [ 59 ], who reported recurrence rates of 24.4% for those with combat stress reactions (people referred for psychiatric intervention during the war) and 13.2% for participants who participated in combat in the same units but without need for psychiatric intervention during the war. This study assessed participants over twenty years, which may explain its higher prevalence rate than the majority of studies which were completed within two-and-a-half years or less. However, the study period was shorter than the forty-seven years of Solomon et al.’s [ 62 ] study, which reported only a 1.6% rate of recurrence. It is unclear why Solomon and Mikulincer [ 59 ] found much higher rates of recurrence.

Two military studies reported recurrence rates for PTSD-populations. These were 24.5% [ 36 ] and 49.6% [ 41 ]. We note that all of Armenta et al.’s [ 36 ] participants had comorbid depression at baseline. We also note some concerns about the reliability of DenVelde et al.’s study [ 41 ], which was a retrospective study asking participants to give complete life-history data at one time-point only.

One military study reported on the prevalence of recurrence in a sub-group of participants who had recovered. Solomon et al. [ 62 ], who reported a prevalence rate of 1.6% (out of the entire trauma-exposed sample) over the first forty-two years of the study, found in a follow-up at forty-seven years that 16.7% of those who had initially recovered experienced recurrence of PTSD during the COVID-19 pandemic.

Prevalence of PTSD recurrence in civilian adult populations

Fourteen studies focused on civilian adults. Findings relating to recurrence prevalence in entire trauma-exposed samples were mixed. Two studies reported rates of under 5% [ 45 , 58 ] in survivors of a terrorist attack and an earthquake respectively. Sungur and Kaya [ 64 ] reported a recurrence rate of 8.9% in survivors of the Sivas disaster, a religious fundamentalist protest which resulted in civilian deaths. Higher rates of recurrence were reported for survivors of an oil rig disaster (18.8%) [ 15 ], survivors of an oil spill (32%) [ 56 ], acutely injured trauma survivors (35%) [ 55 ] and women recently diagnosed with ovarian cancer (57%) [ 43 ].

For populations of civilians with PTSD only, recurrence rates were 4.9% [ 39 ] (type of trauma not reported), 15.4% [ 66 ] (trauma type varied), and 28% [ 46 ] (participants severely injured in accidents). Four studies reported data on the prevalence of recurrence in populations who had previously recovered from PTSD. Reported rates were 14% [ 52 ] (trauma type varied), 29.5% [ 37 ] (trauma type varied), 34% [ 35 ] (trauma type not reported) and 40% [ 65 ] (trauma type varied).

Prevalence of PTSD recurrence in children

Four studies focused on recurrence in adolescents / children, with mixed findings. Fan et al. [ 42 ] found that 3.3% of 1,573 earthquake survivors experienced ‘relapsing/remitting’ PTSD. Liang et al. [ 49 , 50 ] found that 17.7% of 301 earthquake survivors experienced the ‘relapsing’ trajectory of PTSD. An et al. [ 33 ] found that 37% of 246 adolescents experienced ‘recurrent dysfunction’ after experiencing an earthquake.

Prevalence of PTSD recurrence in combined military and civilian populations

Finally, two studies included both military and civilian participants; both of these studies were trials comparing fluoxetine to placebo treatment in people with PTSD. Davidson et al. [ 40 ] found that half of the placebo group relapsed after recovery, compared to 22.2% of the fluoxetine group. Martenyi et al. [ 53 ] reported lower rates of ‘relapse’: 16.1% of the placebo group and 5.8% of the fluoxetine group. The latter study followed up participants after 36 months, while Davidson et al. [ 40 ] followed up participants for a year after treatment.

Predictors of PTSD recurrence

The third and final aim of the present review was to identify factors associated with PTSD recurrence. Firstly, we note that (as shown in Table  2 ), participants in a number of studies had received some type of intervention during the study period, which was typically not accounted for in analyses of predictors. Many other studies did not report whether participants received treatment or not. Having treatment, whether it be medication, therapy, or a combination, is likely to be an important factor influencing PTSD trajectory, given that there are evidence-based treatments for the condition [ 68 ], but this was typically not explored.

Table  5 shows the factors considered as predictors in each study, with significant associations presented in bold. The majority of included studies (22/35) explored at least one covariate; the remaining studies either did not explore covariates or combined recurrent trajectories with other trajectories in their analyses of predictors. Of those studies which did explore covariates of recurrence, we found little consensus.

Sociodemographic factors

Gender was considered as a potential covariate by six studies; one [ 33 ] found that recurrent PTSD was associated with female gender while five studies (including two based on the same data-set) [ 49 , 50 ] found no significant gender association [ 35 , 36 , 42 , 49 , 50 ]. None of the three studies testing age as a covariate found a significant association [ 35 , 36 , 57 ]. One study of school-aged children found that children in a higher grade (i.e. older in age) were more likely to experience PTSD recurrence [ 33 ], while three studies of two cohorts [ 42 , 49 , 50 ] found no significant association between recurrence and school grade. Three studies considered race as a covariate, finding no significant association between PTSD recurrence and race [ 36 , 44 , 55 ]. Other socio-demographic characteristics considered included number of children in the family [ 42 ], marital status and level of education [ 36 ], none of which were found to be associated with PTSD recurrence. For military participants, there were no significant differences in service branch, service component or pay grade between the recurrent and rapid recovery groups [ 36 ].

Psychiatric history

Seven studies considered psychiatric history and concurrent diagnoses as potential covariates of PTSD recurrence, again with mixed findings. Recurrence was not found to be associated with other anxiety syndromes [ 36 ], baseline levels of anxiety [ 54 ], depressive symptoms [ 55 ], baseline levels of depression [ 54 ] or psychiatric history [ 55 ]. Ansell et al. [ 35 ] found that diagnoses of a number of co-morbid mental health disorders such as major depressive disorder and personality disorders such as schizotypal personality disorder, avoidant personality disorder and borderline personality disorder were not associated with recurrence, but participants with a baseline diagnosis of obsessive-compulsive personality disorder were significantly less likely to experience PTSD recurrence. Conversely, Perconte et al. [ 57 ] found that those who experienced recurrence were significantly more likely to report obsessive-compulsive symptoms than those whose symptoms improved without recurrence. Sakuma et al. [ 58 ] found that pre-disaster treatment for mental illness was significantly associated with PTSD recurrence, but note that the results should be interpreted carefully due to the very small number of participants in the ‘fluctuating symptoms’ group who appeared to have experienced recurrent episodes. Perconte et al. [ 57 ] found that, versus the improved symptoms group, those with PTSD recurrence were more likely to report depression, anxiety, hostility, phobic anxiety, somaticism and psychoticism; however, previous psychiatric hospitalisations and pre-treatment ratings of global pathology on a psychiatric scale did not predict recurrence. Finally, Madsen et al. [ 51 ] found that suicidal ideation was significantly higher in the ‘relieved-worsening PTSD’ group than the ‘low-stable’ group and that suicidal ideation was in fact highest in the recurrent (termed ‘relieved-worsening’) group than any other. However, it should be noted that suicidality was not assessed at baseline in this study, therefore it is not clear whether suicidal ideation is a cause or a consequence of PTSD recurrence.

Physical health

Fewer studies considered physical health as a potential predictor of PTSD recurrence. One study found no association between recurrence and disabling injury/illness, somatic symptoms or bodily pain [ 36 ] and another found no association between recurrence and prior treatment for physical illness [ 57 ]. However, obesity was a significant predictor of PTSD recurrence [ 36 ]. In terms of health-related behaviours, Armenta et al. [ 36 ] found no association between PTSD recurrence and smoking status, alcohol problems or sleep duration. However, Perconte et al. [ 57 ] found that higher weekly alcohol intake both before and at termination of PTSD treatment predicted recurrence.

Cognitive ability

Only one study [ 63 ] explored cognitive ability as a potential covariate, finding that the participants who were in the recurrent (termed ‘relieved-worsening PTSD’) group had significantly lower cognitive ability scores than those in the ‘low-stable’ group.

Trauma history and pre-trauma experiences

The review also found mixed evidence for trauma history as a predictor of PTSD recurrence. Liang et al. [ 49 , 50 ] found no association between pre-disaster traumatic experience and PTSD recurrence. Armenta et al. [ 36 ] found no association between recurrence and childhood sexual abuse, childhood verbal abuse, childhood neglect, sexual assault, physical assault, or ‘other life events’, but did find that participants reporting a history of childhood physical abuse were significantly more likely to experience PTSD recurrence. Holliday et al. [ 47 ] found that veterans who had experienced military sexual trauma (MST) had greater initial reductions in PTSD symptoms than those who had not experienced MST, but also experienced a ‘modestly greater’ recurrence of symptoms than those without MST, although this difference did not appear to reach statistical significance. Zanarini et al. [ 65 ] found that the presence of childhood sexual abuse history did not significantly predict time-to-recurrence, but severity of childhood sexual abuse, adult rape history, combination of childhood sexual abuse history and adult rape history, and experiencing sexual assault during study follow-up were associated with less time-to-recurrence. Osofsky et al. [ 56 ] found that abuse, emotional abuse, domestic violence, and greater number of traumas experienced were associated with recurrence of PTSD, and Osenbach et al. [ 55 ] found that recurrent life stressors significantly increased the odds of membership in chronic, relapsing or recovery groups rather than the resilient group. For military participants, one study found combat deployment was significantly associated with recurrent PTSD [ 36 ] while others found combat exposure was not associated with recurrence [ 54 , 57 ]. Finally, Fan et al. [ 42 ] found that compared to the recovery group, relapsing participants experienced significantly fewer negative life events 6-months post-disaster, but significantly more such events at the 24-month follow-up.

Few other pre-trauma experiences were considered. An et al. [ 33 ] found that those with recurrent PTSD were significantly more likely to have experienced academic burnout than those in the recovery trajectory, although there was no difference between the recurrent and delayed trajectories.

Experiences during and immediately after the traumatic experience

The review also found mixed evidence for an association between peri-traumatic experiences and PTSD recurrence. The most consistent finding related to how stressful the traumatic experience was perceived to be at the time. For example, risk of recurrence was significantly higher in those with combat stress reactions [ 59 ] and in those with higher stress relating to the disaster they had experienced [ 56 ], as well as with greater trauma severity [ 49 , 50 ]. However, recurrence was not found to be associated with subjective fear during the event [ 33 ]; directly witnessing a disaster [ 42 ]; property loss during the event [ 33 , 42 ]; property damage [ 42 ]; displacement due to property damage [ 58 ]; near-death experience [ 58 ]; or having a family member injured, killed or missing [ 42 , 58 ].

There was some evidence that initial post-traumatic stress symptoms immediately after the traumatic event could predict PTSD trajectory. Liang et al. [ 49 , 50 ], in a study of PTSD in children from two schools affected by an earthquake, found that children from one of the two schools (‘School 2’) were significantly more likely to experience PTSD recurrence than children from the other school (‘School 1’). Further investigations revealed that after adjusting for immediate post-traumatic stress symptoms the school no longer predicted relapse; those from School 2 had significantly greater post-traumatic stress symptoms immediately after the disaster, which the authors suggest might be due to School 1 providing sufficient psychological services as well as having the same students and teachers before and after the earthquake (therefore perhaps greater social support available), whereas School 2 had insufficient psychological services and consisted of teachers and students from several different schools which could not be reconstructed after the earthquake.

One study [ 58 ] considered occupational-related covariates of PTSD recurrence for disaster recovery workers. They found that having mainly disaster-related occupational duties and lack of rest due to occupational duties were not associated with recurrence, but perceived poor workplace communication did predict recurrence.

Post-trauma experiences and symptoms

An et al. [ 33 ] found that, compared to the delayed PTSD trajectory, those who experienced recurrence were less likely to have experienced post-traumatic growth after the traumatic event; however, there were no differences in post-traumatic growth between the recurrent and recovery groups. Fan et al. [ 42 ] found that neither positive coping nor negative coping six months post-disaster were associated with PTSD recurrence. In a military study, Karstoft et al. [ 48 ] found that poor adjustment to civilian life (i.e. difficulties with community reintegration after deployment) was significantly higher for the recurrent (‘relieved-worsening PTSD’) group than all other groups. However, it is not clear whether poor adjustment was a cause or an effect of PTSD symptoms worsening after initial improvement.

Two studies explored specific cluster symptoms. Murphy and Smith [ 54 ] found PTSD recurrence was not predicted by the magnitude of re-experiencing, avoidance, or hyperarousal symptoms. Boe et al. [ 15 ] found that the number of intrusion and avoidance symptoms five-and-a-half months post-trauma did not predict recurrence, but the number of intrusion and avoidance symptoms both fourteen months and five years after the disaster did predict recurrence.

Social support

Only three studies directly considered social support as a potential covariate. Armenta et al. [ 36 ] found no association between social support and PTSD recurrence, and Perconte et al. [ 57 ] found that family support did not predict recurrence. Fan et al. [ 42 ] found that level of social support six months after experiencing an earthquake was not associated with PTSD recurrence, but those in the ‘relapsing’ group reported significantly less social support 24 months after the earthquake than those in the ‘recovery’ group.

PTSD treatment

Most of the studies investigating treatment for PTSD found that not receiving interventions, or discontinuing treatment, were associated with PTSD recurrence. For example, Osenbach et al. [ 55 ] found that those who received ‘usual care’ only were significantly more likely to experience recurrence than those who received interventions designed to reduce post-traumatic symptoms. Davidson et al. [ 40 ] found that those who received placebo treatment were significantly more likely to experience recurrence than those who received fluoxetine. Martenyi et al. [ 53 ] found that those who discontinued fluoxetine treatment were significantly more likely to experience recurrence, especially for those with combat-related PTSD. However, Perconte et al. [ 57 ] found that number of weeks enrolled in treatment and number of treatment sessions attended did not significantly affect risk of recurrence. In this study, though, being hospitalised at least once since the termination of treatment was used as a proxy measure of ‘recurrence’ and so the findings are arguably not truly representative of actual recurrent episodes of PTSD. Overall, our findings indicated some evidence that treatment helped to avoid recurrent episodes.

In this study, we systematically reviewed 35 studies to identify definitions and prevalence of recurrent PTSD and factors associated with recurrence. It is important to define and operationalise recurrence as the concept needs to be understood in order to make prevention efforts. The health-related, social and economic costs of PTSD can be substantial. PTSD negatively affects individuals’ emotional wellbeing and physical health [ 7 ], impedes social relationships [ 69 ], limits productivity at work and increases sickness absence [ 70 ]. The direct costs (e.g., medical care costs) and indirect costs (e.g., costs of unemployment or reduced productivity) of PTSD can create substantial economic burden [ 7 , 71 ]. Determining the predictors of recurrence of PTSD (which can only be properly understood if ‘recurrence’ itself has a clear definition) is important for prevention efforts: identifying those most at risk for recurrent episodes would allow for the subsequent investigation of ways of mitigating or preventing the risk. However, we found little consensus as to how recurrence is defined, mixed evidence on the prevalence of recurrence and inconsistent findings relating to predictors of recurrence. This lack of clarity about what relapse or recurrence is, and is not, is a major barrier to understanding this important topic.

In a previous review exploring PTSD recurrence in veterans, Berge et al. [ 22 ] acknowledge that there is no generally accepted or used definition of recovery relating to psychological trauma. The definition of recurrence used in their review was the return of symptoms following a period of complete recovery, representing the start of a new and separate episode . However, it is not clear what length of time is covered by ‘a period of complete recovery’ nor what ‘complete recovery’ means. How many days, weeks, or months does an individual need to be free of symptoms of PTSD in order to be considered truly recovered? Is ‘symptom-free’ the only definition of recovery, or is ‘not meeting the criteria for PTSD’ enough? Our own review revealed that there is little consensus as to what recurrence means and the parameters for its definition. Even the terminology used varied across studies, with ‘relapse’, ‘recurrence’, ‘reactivation’ and numerous other terms often used to describe what essentially appeared to be the same concept. There was no consensus as to how long an individual needed to be free of symptoms in order to be considered recovered, nor for how long symptoms needed to recur in order to be considered a recurrent episode. Most studies simply defined recurrence as a change in symptoms between assessments, meaning that whether or not an individual was defined as having a recurrent episode or not very much depended on the scores they reported at arbitrary time-points. Even minor symptom fluctuations could cause someone to change from being identified as a ‘case’ to ‘recovered’ and vice versa. Because PTSD tended to be examined using prospective studies where symptoms were assessed at predetermined assessment points, it is possible that individuals may have onsets of PTSD after one assessment and then remit before the next. With no retrospective assessment between time-points, it is difficult to assess the true prevalence of recurrence. Andrews et al. [ 16 ] make a similar point in relation to delayed onset PTSD, suggesting the absence of information about symptoms outside of the predetermined time-points of studies means that estimates of delayed onset PTSD may be unreliable.

The second aim of the review was to examine the prevalence of PTSD recurrence in existing literature. Given the numerous different ways of assessing PTSD, defining initial recovery and defining recurrence, as well as the differing time-points at which PTSD was assessed across studies, we suggest that the current data on recurrence prevalence is not especially meaningful. We found very different prevalence rates reported within the literature, with data suggesting that anywhere between 0.2% and 57% of trauma-exposed populations might experience recurrent episodes of PTSD. Some of the higher percentages we found seem greater than we would expect, given that only a minority of trauma-exposed people are likely to develop PTSD in the first place – let alone suffer from it, recover from it, and experience a recurrent episode. We would expect that studies carried out over a longer period of time would find higher recurrence rates, simply because in these studies there is more time for recurrent episodes to occur. However, the highest prevalence rate (57%) was found in a study which took place over only 27 weeks [ 43 ]; the authors labelled these participants as ‘intermittent cases’ and it appears likely that symptom fluctuation, rather than true recovery and recurrence, occurred in this study – and potentially many others. Additionally, studies did not typically control for exposure to subsequent trauma, meaning that ‘recurrences’ of PTSD identified may actually be new episodes, rather than a relapse. Further research studies, especially research involving assessments over a number of years, are needed to establish the true prevalence of recurrent PTSD which also needs to be clearly defined with an agreed time period between remission and relapse.

It has been proposed that recurrence rates might increase with old age. Murray [ 72 ] suggests that PTSD can be ‘reactivated’ in older age because physical illnesses become more common, which can reactivate traumatic memories; increased dependence on others due to ageing can reactivate feelings of helplessness; and loss of structure and identity caused by retirement can similarly reactivate traumatic symptoms. Other factors relating to ageing such as decline of cognitive function, difficulty controlling ruminations, reminiscing, and late-life stressors such as serious illness, surgical procedures and death of spouses, siblings or close friends can either directly remind the person of their previous traumatic experience(s) or can induce similar feelings of vulnerability [ 73 ]. Three studies of adults in this review did not find age predicted recurrence [ 35 , 36 , 57 ]; however, the populations trended young overall, with each of the three studies reporting the mean age of participants was under 40. We suggest, then, that more studies of older adults with lifetime PTSD are needed to establish whether this group are at increased risk of recurrence.

The third aim of this review was to understand factors associated with PTSD recurrence. Although a number of potential covariates were considered, most were not investigated by more than a few studies, and findings were varied and inconsistent. Of the covariates investigated by multiple studies, none were found to have significant associations with recurrence across all studies. It was therefore not possible to quantify the extent to which potential risk factors contribute to the risk of recurrence. One reason for the inconsistent findings might be the relatively small numbers of participants with recurrent PTSD in many of the studies. We note also that most studies did not consider either subsequent trauma or treatment impact in their analysis of predictors of recurrence.

We did not find strong evidence of an association between PTSD recurrence and comorbid psychiatric conditions. Recurrence of other mental health disorders, such as anxiety, is reportedly associated with comorbid psychiatric conditions including major depression, alcohol and substance use disorders [ 74 ]. Additionally, comorbid disorders have been found to be associated with an ‘unfavourable long-term course’ of PTSD [ 18 ]. However, in a review of predictors of developing PTSD, Brewin et al. [ 75 ] found that while psychiatric history was associated with development of PTSD, it was not a strong risk factor – factors operating during or after the traumatic exposure had greater effects than the pre-trauma factors. Many studies in this review found no evidence of a relationship between PTSD recurrence and other mental health conditions; in those that did find a relationship, it was not always clear whether the other conditions pre-dated the recurrent PTSD episode or not. Overall, the most consistent evidence we found indicated that recurrence of PTSD was associated with greater stress and traumatic response at the time of the traumatic experience.

We did not find evidence to suggest that trauma type may affect recurrence. Many studies examined PTSD trajectories after a single traumatic event. Those that did include participants who had experienced various different types of trauma did not consider trauma type as a potential predictor of recurrence. Given the wide variations in methodology, it was not appropriate for us to compare recurrence rates for different trauma types within the review. Future research should include participants who have experienced different types of trauma and should consider trauma type as a potential predictor of PTSD trajectory.

Only one study assessed PTSD during the COVID-19 pandemic, with Solomon et al. [ 62 ] reporting that 16.7% of initially-recovered participants experienced recurrence during the pandemic. However, it is not clear how many of this cohort may also have experienced recurrence before the pandemic, and without being able to make that comparison, we cannot ascertain the extent to which recurrence was exacerbated by the pandemic. Additionally, the percentage (16.7%) is similar to recurrence rates in several other, non-COVID studies. Ideally, future studies will present data on PTSD recurrence rates for one cohort at regular intervals, including data collected during or after the COVID-19 pandemic, to ascertain whether the pandemic did affect recurrence rates.

In their review, Steinert et al. [ 18 ] identified older age, higher education, greater trauma severity, higher baseline symptoms, more physical/functional impairments, and poorer social support as predictors of ‘unfavourable’ long-term course of PTSD. These were identified as predictors due to being reported in at least two studies within their review. The current review did not find consistent evidence that age, education, trauma severity, baseline symptoms, impairments or social support predicted recurrence – although age was only considered in studies of young people. We found some evidence from treatment studies that fluoxetine reduced the risk of recurrence, as did participation in an intervention involving a combination of motivational interviewing, behavioural activation and pharmacotherapy. It is therefore difficult to make recommendations relevant to occupational health, as we had hoped to do. Managers of trauma-exposed employees who have developed PTSD may have questions around whether recovered individuals can go back to frontline work, or whether they risk experiencing a recurrence of PTSD. Our findings tentatively suggest that recurrence might be relatively rare (rates of recurrence ranged from 0.2 − 57% in full trauma-exposed samples, mean 13.1%; 4.9 − 49.6% in PTSD-only subgroups, mean 24.5%; and 5.8 − 50% for recovered subgroups, mean 25.4%) but clearer definitions and assessments of recurrence are needed to substantiate that claim. As we found no consistent evidence of predictors of recurrence, it was therefore not possible to identify which sub-groups of people might be more likely to have their PTSD recur. We did find evidence from two studies that recurrence was more prevalent in groups of PTSD patients treated with placebos compared to PTSD patients treated with fluoxetine, suggesting that medication appears at least somewhat effective in reducing the risk of recurrence. However, we found no studies looking at the impact of first-line treatments on relapse (i.e. trauma-focused cognitive behavioural therapy [ 76 ] or eye movement desensitisation and reprocessing [ 77 , 78 ]) which is a major gap in the literature. Whilst more, high-quality studies are carried out, employers should ensure that workers get evidence-based treatments and have an occupational mental health assessment on completion of potentially traumatic work to provide an expert judgement, given that we cannot identify any clear risk factors from the literature.

The key limitation of the literature on PTSD recurrence is that it is not always easy to differentiate between recurrence and symptom fluctuation, and it is also difficult to know what ‘recovery’ truly means. It is not clear how many of the so-called ‘recovered’ participants within the reviewed studies may have been close to clinical thresholds for PTSD at the assessment points. Rather than moving from distinct ‘recovered’ to ‘recurrent episodes’, it may be that individuals only experienced small fluctuations in PTSD symptoms, moving them above and below the symptom thresholds. Indeed, the authors of several of the included studies remarked on the difficulties in identifying PTSD trajectories. In Boe et al.’s [ 15 ] study, clinical interviews were conducted by two clinical psychologists who were trained and supervised by an experienced clinician and trauma researcher and even these experienced individuals had difficulties identifying recurrence of PTSD, with one case being recategorised from ‘full-blown PTSD reactivation’ to ‘sub-syndromal reactivation’ after discussion between the researchers. Markowitz et al. [ 52 ] pointed out that, as they defined relapse as ‘loss of response (to treatment) status’, relapse might reflect barely crossing that threshold: indeed, more in-depth analysis of their six ‘relapsers’ showed that all but one still showed some, albeit more modest, treatment benefit relative to their baseline PTSD severity.

Sakuma et al. [ 58 ] discussed their finding of a ‘fluctuating’ trajectory (and lack of a delayed-onset trajectory), differing from the typical four trajectories widely accepted within the PTSD literature. They suggested the difference may be due to variations in the duration of study periods and characteristics of the study samples. The majority of studies which produce the typical four trajectories are conducted over short periods between a few months and two years [ 9 ], compared to the longer (54-month) period of Sakuma et al.’s [ 58 ] study: the trajectory commonly identified as ‘delayed onset’ could really be a fluctuating trajectory if examined over a longer period. Or, it could reflect a gradual accumulation of symptoms resulting in a delayed presentation of PTSD, rather than delayed onset.

The time-points of assessments could also affect reported prevalence rates. For example, Sungur & Kaya [ 64 ] pointed out that some of their ‘recurrent’ cases would have been considered ‘recovered’ if the study period had been shorter or if participants had not been reassessed at the particular time-points chosen. They also noted that symptoms across the entire participant population seemed to be higher at particular times during the study (namely, at the anniversary of the event and at the time of a disappointing result of a court hearing for compensation), suggesting that the nature and course of PTSD might be influenced by particular events which might trigger unwanted memories of the traumatic event. In the current review, most studies assessed participants for at least a year, but not all: five [ 38 , 39 , 43 , 52 , 53 ] followed participants for less than a year. Additionally, two studies [ 44 , 47 ] reported assessing participants pre-treatment and four months post-treatment but it was not clear how long treatment lasted.

We suggest that PTSD recurrence may not have been adequately assessed in many of the included studies. For example, Chopra et al. [ 39 ] described how, in order to minimise respondent burden, assessors were expected to stop inquiring about PTSD symptoms if participants were unlikely to meet the criteria and if they answered no to particular questions on the assessment tool. This could mean that some individuals who did have recurrent episodes of PTSD were not identified as they did not complete the full measures. Additionally, we found that a number of studies had very vague definitions of recurrence, such as ‘increasing symptoms’, where it was unclear what exactly this meant. Others used hospitalisation as a proxy measure for recurrence, or simply asked participants whether they perceived their symptoms had been exacerbated and in one case used the investigator’s own judgements as a way of determining recurrence. It is therefore likely that some recurrent cases may have been missed while others who never truly ‘recovered’ at all may have been reported to have experienced recurrence. Overall, the vague and inconsistent ways of assessing recurrence mean it is currently impossible to ascertain true recurrence rates within existing literature.

It is also possible that recurrent trajectories of PTSD appear in studies which do not identify them as such. For example, in Andrews et al.’s [ 16 ] review, the authors note that some cases of ‘delayed-onset PTSD’ in veterans of relatively old age with long intervals to first onset may in fact have had episodes of PTSD soon after their traumatic experiences which were undisclosed or forgotten. In other words, some cases of supposedly ‘delayed-onset’ PTSD might actually be recurrent cases. Andrews et al. [ 16 ] also point out that many of the studies included in their review of delayed-onset PTSD did not assess whether respondents could have had onsets of PTSD and then remitted before the next assessment point – which could lead to both over- and under-estimates of delayed-onset rates of PTSD. Indeed, the studies included in our own review tended to focus only on the scores at the various time-points and did not explore participants’ perceptions of symptom fluctuations outside of the time-points set by the study.

Limitations

There are a number of limitations of the literature reviewed. Many did not collect data on whether participants had undergone any intervention or not, and those that did tended not to include this as a potential confounding variable. The majority of studies did not assess whether participants experienced additional potentially traumatic experiences between PTSD assessments. Many did not define the parameters of ‘recovery’ and ‘recurrence’ and it is not clear whether recurrent episodes identified were truly recurrent episodes or merely symptom fluctuations. Many did not collect data on whether or not participants received any treatment for PTSD between data collection time-points, and many of those which did ask participants whether they had received any treatment did not distinguish between types of treatment. It is therefore unclear if, and how many, participants in many studies received any evidence-based PTSD treatment or not. Additionally, the majority of studies did not collect data on the time period of any treatment received. Some studies had extremely long gaps (e.g., decades) between assessments which could mean that recurrences were missed.

There are also limitations of the review process itself. Firstly, the screening, data extraction and quality appraisal were carried out by one author. Although decisions about exclusion or inclusion were discussed with the second author, it would have been preferable to have multiple screeners. We limited the review to English-language studies only, meaning that important studies published in other languages would have been missed. We included only studies which identified ‘recurrent episodes’ (or equivalent terminology e.g. relapse, reactivation); studies which identified no recurrent trajectory were not reviewed. It may be that these studies did not include a sufficient number of assessments to pick up on recurrent episodes, but it may also be that no participants in these studies experienced recurrence and therefore the true prevalence of recurrence may be lower than this review suggests.

Conclusions and implications

The main conclusion that can be drawn from the current review is that, moving forward, better clarity and consensus regarding the definition and identification of recurrent PTSD are urgently needed. Berge et al. [ 22 ] suggest that consistent definitions of relapse-related terms, supported by empirical research, are required in order to make studies of PTSD trajectories more robust. The findings of this review support this suggestion. Experts in the field should agree on an appropriate definition of recurrence (i.e. symptom-based or threshold-based) and should agree how long an individual needs to be ‘better’ for in order to be considered recovered as well as how long an individual needs to experience symptoms for in order to be considered as having a recurrent episode. Recurrence is arguably better-defined for recurrent depressive disorder, with the ICD-11 stating that recurrence is characterised by a history of depressive episodes separated by at least several months without significant mood disturbance [ 79 ]. However, further clarity is still needed. How many months is ‘several’? What are ‘significant’ symptoms? Still, we suggest this might be a useful starting point for a working definition of recurrent PTSD: a history of episodes of PTSD separated by at least several (i.e., three) months without significant (i.e., meeting diagnostic criteria) PTSD symptoms . However, further research is necessary to clarify whether these parameters (i.e. three months as a time period, symptom thresholds as a diagnostic tool) are the most appropriate to use. Using consistent terminology within the literature would make it easier to researchers in the future to understand true prevalence rates of PTSD recurrence and to compare them across studies. Further research allowing for the identification of recurrent PTSD episodes is needed. We believe the gold standard for assessing PTSD and properly identifying its trajectories, including recurrent trajectories, would be using the Clinician Administered PTSD Scale (CAPS) [ 80 ], or other validated questionnaires, at multiple specific time points over a long period of time. Figure  2 summarises the findings of the review and the proposed next steps based on our findings.

Summary of review and suggested next steps

It is important to understand recurrence in order to take steps towards reducing the risk of PTSD recurring. However, due to the inconsistent findings relating to predictors of recurrence, it is difficult to draw conclusions about the best ways of preventing or minimising recurrence. We suggest that ensuring that people who develop PTSD are provided with timely, evidence-based treatments is a logical first step [ 68 ]. Second, awareness of ‘early warning sign’ symptoms and ‘triggers’ might be useful, as well as awareness of effective coping strategies and how to access support. That is, if people with PTSD are able to recognise when they are struggling more and acknowledge that they need to be proactive in ensuring symptoms do not develop into full-blown PTSD again, they may be able to draw on their coping skills or reach out for formal or informal support when a recurrent episode seems imminent and may be able to stave off the recurrent episode. We also suggest that reframing the re-emergence of symptoms in a more positive way might be useful: instead of feeling defeated that symptoms have recurred, people could remind themselves that they have recovered once and therefore know that they are capable of doing so again. Within organisational settings, it is also important to foster an environment in which people who have any mental health condition, including PTSD, feel confident that asking for help will not lead to stigmatisation or increase the likelihood of inappropriate job loss. It may also be helpful to incorporate relapse prevention, understanding ‘warning signs’ of recurrent episodes and positive reframing into PTSD treatment programmes.

Data availability

All data generated or analysed during this study are included in this published article.

Abbreviations

Clinician-Administered PTSD Scale

Military sexual trauma

National Institutes for Health

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

Post-Traumatic Stress Disorder

Benjet C, Bromet E, Karam EG, Kessler RC, McLaughlin KA, Ruscio AM, et al. The epidemiology of traumatic event exposure worldwide: results from the World Mental Health Survey Consortium. Psychol Med. 2016;46(2):327–43.

Article   CAS   PubMed   Google Scholar  

Brooks SK, Greenberg N. Preventing and treating trauma-related mental health problems. In: Lax P, editor. Textbook of acute trauma care. Switzerland: Springer Cham; 2022. pp. 829–46.

Chapter   Google Scholar  

Bonanno GA. Loss, trauma, and human resilience: have we underestimated the human capacity to thrive after extremely aversive events? Am Psychol. 2004;59(1):20–8.

Article   PubMed   Google Scholar  

American Psychiatric Association. Diagnostic and statistical manual of mental disorders (fifth edition). ; 2013. https://doi.org/10.1176/appi.books.9780890425596 .

Messman-Moore TL, Cook NK. Posttraumatic stress disorder. In: H. S. Friedman, editor. Encyclopedia of mental health (second edition). San Diego: Academic Press; 2016. pp. 308 – 12.

Spottswood M, Davydow DS, Huang H. The prevalence of posttraumatic stress disorder in primary care: a systematic review. Harv Rev Psychiatry. 2017;25(4):159–69. https://doi.org/10.1097/HRP.0000000000000136 .

Article   PubMed   PubMed Central   Google Scholar  

Kapfhammer HP. Acute and long-term mental and physical sequelae in the aftermath of traumatic exposure - some remarks on the body keeps the score. Psychiatr Danub. 2018;30(3):254–72. https://doi.org/10.24869/psyd.2018.254 .

Peleg T, Shalev AY. Longitudinal studies of PTSD: overview of findings and methods. CNS Spectr. 2006;11(8):589–602. https://doi.org/10.1017/s109285290001364x .

Galatzer-Levy IR, Huang SH, Bonanno GA. Trajectories of resilience and dysfunction following potential trauma: a review and statistical evaluation. Clin Psychol Rev. 2018;63:41–55. https://doi.org/10.1016/j.cpr.2018.05.008 .

van de Schoot R, Sijbrandij M, Depaoli S, Winter SD, Olff M, van Loey NE. Bayesian PTSD-trajectory analysis with informed priors based on a systematic literature search and expert elicitation. Multivar Behav Res. 2018;53(2):267–91. https://doi.org/10.1080/00273171.2017.1412293 .

Article   Google Scholar  

Santiago PN, Ursano RJ, Gray CL, Pynoos RS, Spiegel D, Lewis-Fernandez R, et al. A systematic review of PTSD prevalence and trajectories in DSM-5 defined trauma exposed populations: intentional and non-intentional traumatic events. PLoS ONE. 2013;8(4):5. https://doi.org/10.1371/journal.pone.0059236 .

Article   CAS   Google Scholar  

Mota N, Bolton SL, Enns MW, Afifi TO, El-Gabalawy R, Sommer JL, et al. Course and predictors of posttraumatic stress disorder in the Canadian Armed forces: a nationally representative, 16-year follow-up study. Can J Psychiatry. 2021;66(11):982–95. https://doi.org/10.1177/0706743721989167 .

Magruder KM, Goldberg J, Forsberg CW, Friedman MJ, Litz BT, Vaccarino V, et al. Long-term trajectories of PTSD in Vietnam-era veterans: the course and consequences of PTSD in twins. J Trauma Stress. 2016;29(1):5–16. https://doi.org/10.1002/jts.22075 .

Karamustafalioglu OK, Zohar J, Guveli M, Gal G, Bakirn B, Fostick L, et al. Natural course of posttraumatic stress disorder: a 20-month prospective study of Turkish Earthquake survivors. J Clin Psychiatry. 2006;67(6):882–9. https://doi.org/10.4088/JCP.v67n0604 .

Boe HJ, Holgersen KH, Holen A. Reactivation of posttraumatic stress in male Disaster survivors: the role of residual symptoms. J Anxiety Disord. 2010;24(4):397–402doi. https://doi.org/10.1016/j.janxdis.2010.02.003 .

Andrews B, Brewin CR, Philpott R, Stewart L. Delayed-onset posttraumatic stress disorder: a systematic review of the evidence. Am J Psychiatry. 2007;164(9):1319–26. https://doi.org/10.1176/appi.ajp.2007.06091491 .

North CS, Oliver J. Analysis of the longitudinal course of PTSD in 716 survivors of 10 Disasters. Soc Psychiatry Psychiatr Epidemiol. 2013;48(8):1189–97. https://doi.org/10.1007/s00127-012-0639-x .

Steinert C, Hofmann M, Leichsenring F, Kruse J. The course of PTSD in naturalistic long-term studies: high variability of outcomes. A systematic review. Nord J Psychiatry. 2015;69(7):483–96. https://doi.org/10.3109/08039488.2015.1005023 .

Bonde JPE, Jensen JH, Smid GE, Flachs EM, Elklit A, Mors O, Videbech P. Time course of symptoms in posttraumatic stress disorder with delayed expression: a systematic review. Acta Psychiatr Scand. 2022;145(2):116–31. https://doi.org/10.1111/acps.13372 .

Morina N, Wicherts JM, Lobbrecht J, Priebe S. Remission from post-traumatic stress disorder in adults: a systematic review of long term outcome studies. Clin Psychol Rev. 2014;34(3):249–55. https://doi.org/10.1016/j.cpr.2014.03.002 .

Pavlacic JM, Buchanan EM, McCaslin SE, Schulenberg SE, Young JN. A systematic review of posttraumatic stress and resilience trajectories: identifying predictors for future treatment of veterans and service members. Prof Psychol Res Pr. 2022;3266–75. https://doi.org/10.1037/pro0000451 .

Berge EE, Hagen R, Halvorsen JO. PTSD relapse in veterans of Iraq and Afghanistan: a systematic review. Mil Psychol. 2020;32(4):300–12. https://doi.org/10.1080/08995605.2020.1754123 .

Article   PubMed Central   Google Scholar  

Batelaan NM, Bosman RC, Muntingh A, Scholten WD, Huijbregts KM, van Balkom AJLM. Risk of relapse after antidepressant discontinuation in anxiety disorders, obsessive-compulsive disorder, and post-traumatic stress disorder: systematic review and meta-analysis of relapse prevention trials. BMJ. 2017;358:j3927. https://doi.org/10.1136/bmj.j3927 .

Levy HC, O’Bryan EM, Tolin DF. A meta-analysis of relapse rates in cognitive-behavioural therapy for anxiety disorders. J Anxiety Disord. 2021;81:102407. https://doi.org/10.1016/j.janxdis.2021.102407 .

Nagarajan R, Krishnamoorthy Y, Basavarachar V, Dakshinamoorthy R. Prevalence of post-traumatic stress disorder among survivors of severe COVID-19 Infections: a systematic review and meta-analysis. J Affect Disord. 2022;299:52–9.

Brooks SK, Webster RK, Smith LE, Woodland L, Wessely S, Greenberg N, Rubin GJ. The psychological impact of quarantine and how to reduce it: Rapid review of the evidence. Lancet. 2020;395(10227):912–20.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Bonsaksen T, Heir T, Schou-Bredal I, Ekeberg Ø, Skogstad L, Grimholt TK. Post-traumatic stress disorder and associated factors during the early stage of the COVID-19 pandemic in Norway. Int J Environ Res Public Health. 2020;17(24):9210.

Hori A, Sawano T, Ozaki A, Tsubokura M. Exacerbation of subthreshold PTSD symptoms in a Great East Japan Earthquake survivor in the context of the COVID-19 pandemic. Case Rep Psychiatry. 2021;2021:6699775. https://doi.org/10.1155/2021/6699775 .

Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ. 2009;339:b2535.

Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006;3(2):77–101.

Bagias C, Sukumar N, Weldeselassie Y, Oyebode O, Saravanan P. Cord blood adipocytokines and body composition in early childhood: a systematic review and meta-analysis. Int J Environ Res Public Health. 2021;18(4):1897. https://doi.org/10.3390/ijerph18041897 .

Rosella L, Bowman C, Pach B, Morgan S, Fitzpatrick T, Goel V. The development and validation of a meta-tool for quality appraisal of public health evidence: Meta Quality Appraisal Tool (MetaQAT). Public Health. 2016;136:57–65.

An YY, Huang JL, Yeung ETF, Hou WK. Academic burnout and posttraumatic growth predict trajectories of posttraumatic stress disorder symptoms of adolescents following Yancheng Tornado in China. Int J Stress Manage. 2022;29(2):143–53. https://doi.org/10.1037/str0000240 .

Andersen SB, Karstoft KI, Bertelsen M, Madsen T. Latent trajectories of trauma symptoms and resilience: the 3-year longitudinal prospective USPER study of Danish veterans deployed in Afghanistan. J Clin Psychiatry. 2014;75(9):1001–8. https://doi.org/10.4088/JCP.13m08914 .

Ansell EB, Pinto A, Edelen MO, Markowitz JC, Sanislow CA, Yen S, et al. The association of personality disorders with the prospective 7-year course of anxiety disorders. Psychol Med. 2011;41(5):1019–28.

Armenta RF, Walter KH, Geronimo-Hara TR, Porter B, Stander VA, LeardMann CA, et al. Longitudinal trajectories of comorbid PTSD and depression symptoms among US service members and veterans. BMC Psychiatry. 2019;19(1):12. https://doi.org/10.1186/s12888-019-2375-1 .

Benítez CIP, Zlotnick C, Stout RI, Lou FJ, Dyck I, Weisberg R, Keller M. (2012). A 5-year longitudinal study of posttraumatic stress disorder in primary care patients. Psychopathol. 2012;45(5):286 – 93. https://doi.org/10.1159/000331595 .

Berntsen D, Johannessen KB, Thomsen YD, Bertelsen M, Hoyle RH, Rubin DC. Peace and War: trajectories of posttraumatic stress disorder symptoms before, during, and after military deployment in Afghanistan. Psychol Sci. 2012;23(12):1557–65. https://doi.org/10.1177/0956797612457389 .

Chopra MP, Zhang H, Kaiser AP, Moye JA, Llorente MD, Oslin DW, Spiro IA. PTSD is a chronic, fluctuating disorder affecting the mental quality of life in older adults. Am J Geriatr Psychiatry. 2014;22(1):86–97. https://doi.org/10.1016/j.jagp.2013.01.064 .

Davidson JRT, Connor KM, Hertzberg MA, Weisler RH, Wilson WH, Payne VM. Maintenance therapy with fluoxetine in posttraumatic stress disorder: a placebo-controlled discontinuation study. J Clin Pharmacol. 2005;25(2):166–9. https://doi.org/10.1097/01.jcp.0000155817.21467.6c .

DenVelde WO, Hovens JE, Aarts PGH, FreyWouters E, Falger PRJ, VanDuijn H, et al. Prevalence and course of posttraumatic stress disorder in Dutch veterans of the civilian resistance during World War II: an overview. Psychol Rep. 1996;78(2):519–29. https://doi.org/10.2466/pr0.1996.78.2.519 .

Fan F, Long K, Zhou Y, Zheng Y, Liu X. Longitudinal trajectories of post-traumatic stress disorder symptoms among adolescents after the Wenchuan Earthquake in China. Psychol Med. 2015;45(13):2885–96. https://doi.org/10.1017/s0033291715000884 .

Gonçalves V, Jayson G, Tarrier N. A longitudinal investigation of posttraumatic stress disorder in patients with Ovarian cancer. J Psychosom Res. 2011;70(5):422–31. https://doi.org/10.1016/j.jpsychores.2010.09.017 .

Gross GM, Smith N, Holliday R, Rozek DC, Hoff R, Harpaz-Rotem I. Racial disparities in clinical outcomes of Veterans affairs residential PTSD treatment between Black and White veterans. Psychiatr Serv. 2022;73(2):126–32. https://doi.org/10.1176/appi.ps.202000783 .

Hansen MB, Birkeland MS, Nissen A, Blix I, Solberg O, Heir T. Prevalence and course of symptom-defined PTSD in individuals directly or indirectly exposed to terror: a longitudinal study. Psychiatry. 2017;80(2):171–83. https://doi.org/10.1080/00332747.2016.1230983 .

Hepp U, Moergeli H, Buchi S, Bruchhaus-Steinert H, Kraemer B, Sensky T, et al. Post-traumatic stress disorder in serious accidental injury: 3-year follow-up study. Br J Psychiatry. 2008;192(5):376–83. https://doi.org/10.1192/bjp.bp.106.030569 .

Holliday R, Smith NB, Holder N, Gross GM, Monteith LL, Maguen S, et al. Comparing the effectiveness of VA residential PTSD treatment for veterans who do and do not report a history of MST: a national investigation. J Psychiatr Res. 2020;122:42. https://doi.org/10.1016/j.jpsychires.2019.12.012

Karstoft KI, Armour C, Andersen SB, Bertelsen M, Madsen T. Community integration after deployment to Afghanistan: a longitudinal investigation of Danish soldiers. Soc Psychiatry Psychiatr Epidemiol. 2015;50(4):653–60. https://doi.org/10.1007/s00127-014-0973-2 .

Liang YM, Cheng J, Zhou YY, Liu ZK. Trajectories of posttraumatic stress disorders among children after the Wenchuan Earthquake: a four-year longitudinal study. Eur J Psychotraumatology. 2019;10(1):11. https://doi.org/10.1080/20008198.2019.1586266 .

Liang YM, Zhou YY, Liu ZK. Consistencies and differences in posttraumatic stress disorder and depression trajectories from the Wenchuan Earthquake among children over a 4 year period. J Affect Disord. 2021;279:9–16. https://doi.org/10.1016/j.jad.2020.09.107 .

Madsen T, Karstoft K-I, Bertelsen M, Andersen SB. Postdeployment suicidal ideations and trajectories of posttraumatic stress disorder in Danish soldiers: a 3-year follow-up of the USPER study. J Clin Psychiatry. 2014;75(9):994–1000. https://doi.org/10.4088/JCP.13m08910 .

Markowitz JC, Choo T-H, Neria Y. Do acute benefits of interpersonal psychotherapy for posttraumatic stress disorder endure? Can J Psychiatry. 2018;63(1):37–43. https://doi.org/10.1177/0706743717720690 .

Martenyi F, Brown EB, Zhang H, Koke SC, Prakash A. Fluoxetine v placebo in prevention of relapse in post-traumatic stress disorder. Br J Psychiatry. 2002;181(4):315 – 20. doi:10/1192/bjp.181.4.315.

Murphy D, Smith KV. Treatment efficacy for veterans with posttraumatic stress disorder: latent class trajectories of treatment response and their predictors. J Trauma Stress. 2018;31(5):753–63. https://doi.org/10.1002/jts.22333 .

Osenbach JE, Lewis C, Rosenfeld B, Russo J, Ingraham LM, Peterson R, et al. Exploring the longitudinal trajectories of posttraumatic stress disorder in injured trauma survivors. Psychiatry. 2014;77(4):386–97. https://doi.org/10.1521/psyc.2014.77.4.386 .

Osofsky HJ, Weems CF, Hansel TC, Speier AH, Osofsky JD, Graham R, et al. Identifying trajectories of change to improve understanding of integrated health care outcomes on PTSD symptoms post Disaster. Fam Syst Health. 2017;35(2):155–66. https://doi.org/10.1037/fsh0000274 .

Perconte ST, Griger ML. Comparison of successful, unsuccessful, and relapsed Vietnam veterans treated for posttraumatic stress disorder. J Nerv Ment Dis. 1991;179(9):558–62.

Sakuma A, Ueda I, Shoji W, Tomita H, Matsuoka H, Matsumoto K. Trajectories for post-traumatic stress disorder symptoms among local Disaster recovery workers following the Great East Japan Earthquake: Group-based trajectory modeling. J Affect Dis. 2020;274:742-51. https://doi.org/10.1016/j.jad.2020.05.152 .

Solomon Z, Mikulincer M. Trajectories of PTSD: a 20-year longitudinal study. Am J Psychiatry. 2006;163(4):659–66. https://doi.org/10.1176/ajp.2006.163.4.659 .

Solomon Z, Garb R, Bleich A, Grupper D. Reactivation of combat-related posttraumatic stress disorder. Am J Psychiatry. 1987;144(1):51–5.

Solomon Z, Bachem R, Levin Y, Crompton L, Ginzburg K. Long-term trajectories of posttraumatic stress disorder: categorical versus continuous assessment. Psychiatry. 2018;81(4):376–90. https://doi.org/10.1080/00332747.2018.1485369 .

Solomon Z, Mikulincer M, Ohry A, Ginzburg K. Prior trauma, PTSD long-term trajectories, and risk for PTSD during the COVID-19 pandemic: a 29-year longitudinal study. J Psychiatr Res. 2021;141:140–5. https://doi.org/10.1016/j.jpsychires.2021.06.031 .

Sørensen HJ, Andersen SB, Karstoft KI, Madsen T. The influence of pre-deployment cognitive ability on post-traumatic stress disorder symptoms and trajectories: the Danish USPER follow-up study of Afghanistan veterans. J Affect Disord. 2016;196:148–53. https://doi.org/10.1016/j.jad.2016.02.037 .

Sungur M, Kaya B. The onset and longitudinal course of a man-made post-traumatic morbidity: survivors of the Sivas Disaster. Int J Psychiatry Clin Pract. 2001;5(3):195–202. https://doi.org/10.1080/136515001317021662 .

Zanarini MC, Horz S, Frankenburg FR, Weingeroff J, Reich DB, Fitzmaurice G. The 10-year course of PTSD in borderline patients and axis II comparison subjects. Acta Psychiatr Scand. 2011;124(5):349–56. https://doi.org/10.1111/j.1600-0447.2011.01717.x .

Zlotnick C, Warshaw M, Shea MT, Allsworth J, Pearlstein T, Keller MB. Chronicity in posttraumatic stress disorder (PTSD) and predictors of course of comorbid PTSD in patients with anxiety disorders. J Trauma Stress. 1999;12(1):89–100. https://doi.org/10.1023/A:1024746316245 .

Lai BS, Lewis R, Livings MS, La Greca AM, Esnard AM. Posttraumatic stress symptom trajectories among children after Disaster exposure: a review. J Trauma Stress. 2017;30(6):571–82. https://doi.org/10.1002/jts.22242 .

National Institute for Health and Care Excellence. Post-traumatic stress disorder. 2018. https://www.nice.org.uk/guidance/ng116

Campbell SB, Renshaw KD. Posttraumatic stress disorder and relationship functioning: a comprehensive review and organizational framework. Clin Psychol Rev. 2018;65:152–62. https://doi.org/10.1016/j.cpr.2018.08.003 .

Stergiopoulos E, Cimo A, Cheng C, Bonato S, Dewa CS. Interventions to improve work outcomes in work-related PTSD: a systematic review. BMC Public Health. 2011;11:838. https://doi.org/10.1186/1471-2458-11-838

Davis LL, Schein J, Cloutier M, Gagnon-Sanschagrin P, Maitland J, Urganus A, Guerin A, Lefebvre P, Houle CR. The economic burden of posttraumatic stress disorder in the United States from a societal perspective. J Clin Psychiatry. 2022;83(3):21m14116. https://doi.org/10.4088/JCP.21m14116

Murray A. Recurrence of post traumatic stress disorder. Nurs Older People. 2005;17(6):24–30.

Floyd M, Rice J, Black SR. Recurrence of posttraumatic stress disorder in late life: a cognitive aging perspective. J Clin Geropsychol. 2002;8(4):303–11. https://doi.org/10.1023/A:1019679307628 .

Bruce SE, Yonkers KA, Otto MW, Eisen JL, Weisberg RB, Pagano M, et al. Influence of psychiatric comorbidity on recovery and recurrence in generalized anxiety disorder, social phobia, and panic disorder: a 12-year prospective study. Am J Psychiatry. 2005;162(6):1179–87.

Brewin CR, Andrews B, Valentine JD. Meta-analysis of risk factors for posttraumatic stress disorder in trauma-exposed adults. J Consult Clin Psychol. 2000;68(5):748–66.

Cohen JA, Mannarino AP, Deblinger E. Treating trauma and traumatic grief in children and adolescents. New York: Guildford; 2010.

Google Scholar  

Shapiro F. Efficacy of the eye movement desensitization procedure in the treatment of traumatic memories. J Trauma Stress. 1989;2(2):199–223.

Shapiro F. Eye movement desensitization and reprocessing (EMDR): evaluation of controlled PTSD research. J Behav Ther Exp Psychiatry. 1996;27(3):209–18.

World Health Organization. 6A71 Recurrent depressive disorder. In International statistical classification of diseases and related health problems (11th ed.). ; 2019. https://icd.who.int/browse11/l-m/en#/http%3A%2F%2Fid.who.int%2Ficd%2Fentity%2F1194756772 .

Weathers FW, Blake DD, Schnurr PP, Kaloupek DG, Marx BP, Keane TM. The Clinician-Administered PTSD Scale for DSM-5 (CAPS-5). [Assessment]. 2013. Available from www.ptsd.va.gov .

Download references

This study was funded by the National Institute for Health and Care Research Health Protection Research Unit (NIHR HPRU) in Emergency Preparedness and Response, a partnership between the UK Health Security Agency, King’s College London and the University of East Anglia. The views expressed are those of the author(s) and not necessarily those of the NIHR, UKHSA or the Department of Health and Social Care. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising. The funders had no role in carrying out the review or preparing the manuscript for publication.

Author information

Authors and affiliations.

Department of Psychological Medicine, King’s College London, Weston Education Centre, SE5 9RJ, London, United Kingdom

Samantha K Brooks & Neil Greenberg

You can also search for this author in PubMed   Google Scholar

Contributions

NG and SKB conceptualised the review. SKB carried out searches, screening, data extraction and analysis. NG contributed to analysis. SKB wrote the first draft of the manuscript and NG edited the manuscript. Both authors reviewed the final manuscript.

Corresponding author

Correspondence to Samantha K Brooks .

Ethics declarations

Ethics approval and consent to participate.

Not applicable.

Consent for publication

Competing interests.

NG runs March on Stress Ltd which is a psychological health consultancy and carries out expert witness work which includes PTSD cases.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Brooks, S.K., Greenberg, N. Recurrence of post-traumatic stress disorder: systematic review of definitions, prevalence and predictors. BMC Psychiatry 24 , 37 (2024). https://doi.org/10.1186/s12888-023-05460-x

Download citation

Received : 04 September 2023

Accepted : 13 December 2023

Published : 09 January 2024

DOI : https://doi.org/10.1186/s12888-023-05460-x

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Post-traumatic stress disorder

BMC Psychiatry

ISSN: 1471-244X

• Submission enquiries: [email protected]
• General enquiries: [email protected]

An official website of the United States government

The .gov means it's official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you're on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• Browse Titles

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

The voice and its disorders in teachers

INSERM Collective Expertise Centre .

• Copyright and Permissions

This document presents the review and recommendations of the expert group convened by Inserm in the context of the collective expert procedure to reply to the request of the French National Education General Mutualist Association (MGEN) relating to the voice and its disorders in teachers. The document is based on the scientific data available as at the second half of 2005. Some 570 articles constituted the document base for this expert review.

The Inserm collective expert review center ensured coordination of this collective expert review.

The voice is essential to the exercise of the teaching profession. The voice, which may be qualified as a `professional instrument', is frequently put to rude use. Teachers express themselves in a noisy and frequently poorly sound-proofed environment for several hours every working day.

A voice disorder may have mental, physical, emotional and communicational repercussions and thus have an impact on professional and social life. Teachers' awareness of the importance of the voice as a tool with which to capture their listeners' attention and communicate effectively should sensitize them to the acoustic conditions of their working environment and encourage them to consult as of the initial signs of vocal fatigue.

The French National Education General Mutualist Association (MGEN) asked Inserm for a critical and synthetic analysis of all the international scientific data available on the various disorders encountered by professionals using their voices in their work, particularly teachers. In order to respond to that request, Inserm convened a pluridisciplinary group of experts in the field of ENT medicine, phoniatrics, wave and signal physics, speech therapy and spoken communication.

The group's scientific expert review addressed the following questions:

• What concepts underlie analysis of speech? How do the various methods adopted, types of measurement and progress achieved in elucidating the fundamental processes of speech evolve over time?
• What mechanisms are involved in the production of sound, voice and speech?
• What forms of disease are associated with voice disorders? What is vocal fatigue? What is a strained voice?
• What prevalence and incidence data on voice disorders in teachers in France and abroad are available? Do prevalence and incidence vary with age or gender?
• What exacerbating etiological factors are linked to the subject him/herself and which are linked to the conditions of voice use?
• How can risk populations among teachers be defined? What screening methods are available? What instruments are available to measure the quality of the voice?
• How is a vocal assessment to be conducted? What methods are implemented to diagnose voice disorders?
• What medical and surgical treatments are used for voice disorders?
• How can the voice be rehabilitated? What is the proportion of relapses after voice rehabilitation? What are the roles of occupational physicians and speech therapists in teacher information?
• What preventive programs are available? How effective are those programs?

The international bibliographic database search led to selective retrieval of over 570 articles. During eight sessions held between March 2005 and April 2006, the experts presented an analysis and synthesis of the international published work, each in his/her field of expertise. In addition to the analytical proceedings, hearings were scheduled during which presentations on the psychological management of voice disorders and on pilot studies in the field of acoustics were delivered.

• Group of experts and authors

Denis autesserre , Institute of spoken communication, Grenoble

Nicole charpy , Unit of phoniatry, Institute Arthur Vernes, Paris

Lise crevier - buchman , Laboratory voice, speech and swallowing, European Hospital Georges Pompidou, CNRS-UMR 7018, Paris

Philippe dejonckere , Institute of Phoniatrics, Surgical Speciality Division, University Medical Center, Utrecht, The Netherlands and Institute of Occupational Diseases, Brussels, Belgium

Frédéric eluard , Occupational physician, Perpignan

Élisabeth fresnel , Laboratory of voice ESPACE (Exploration and care of voice, hearing, singing and recording), Paris

Antoine giovanni , Laboratory of clinical hearing phonology, UPRES-EA 2668, Federation ORL University hospital center Timone, Marseille

Jean schoentgen , Research unit waves and signals, Independant University of Brussels, Brussels, Belgium

Bernadette timmermans , Research institute for theater and screen, Erasmushogeschool Brussels, Belgium

Paul-Laurent assoun , Clinical laboratory of body and psychanalytic anthropology, University Paris 7

Rose Marie bourgault , Laboratory of voice ESPACE (Exploration and care of voice, hearing, singing and recording), Paris

Guy cornut , Phoniatrician, Lyon

Line guerra , Sooch san souci , Hearing, Intelligibility, Acoustic, Boulogne

Malte kob , Chair of pediatric phoniatry and audiology, High technical training school of Rhénanie Westphalie, University of Aix-la-Chapelle

• Scientific and editorial coordination

Élisabeth alimi , Expert associate, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

Fabienne bonnin , Scientific associate, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

Catherine chenu , Scientific associate, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

Jeanne etiemble , Director, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

Anne-Laure pellier , Scientific associate, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

• Bibliographic assistance

Chantal rondet - grellier , Documentalist, Inserm collective expertise center, Faculty of medicine Xavier-Bichat, Paris

Voice disorders are frequent in the teaching profession. A third of teachers consider that the disorders interfere with their ability to teach. A historical epistemological approach to speech and the voice was thus a natural component of the literature review.

Voice disorders are the consequences - experienced, perceptible or audible - of an anomaly or organic lesion, acquired or congenital, of the vocal cords. Deficient control of the respiration, or laryngeal articulation or a psychological difficulty may also be involved. Acquired organic lesions may reflect straining or overworking the voice. Physical, chemical and infectious agents present in the environment may play an important role in the good or bad functioning of the vocal cords.

Voice diseases are subject to medical or surgical treatment and usually necessitate discontinuation of vocal activity. Rehabilitation, an integral part of treatment, is to be adapted to the subject's personal situation and the context of the dysphonia. The preventive approach involves information, awareness promotion and training exercises. Prevention may also address the acoustic environmental conditions, particularly for teachers, and act on the environment to enhance the effectiveness of communication.

Analyses of speech and the voice have a long history

How can speaking man who has the impression of being at one with his language detach himself for its use in order to make it an instrument for knowledge or for acting on others? Did reflection precede the action or did action precede reflection? The need to answer those questions and take into account, in the present day, the simultaneous existence of original scientific analyses and know-how transmitted through preceding generations fully legitimizes this historical epistemological approach. What should the subject be?

Speech, through the material references associated with it, would appear to be an appropriate Ariadne's thread. However, a material analysis of the linguistic phenomena by appropriate methods does not dispense one from resituating speech studies in the underlying theoretical contexts. In fact, the generic term `speech' should be better defined in order to reflect the early historical differentiation of speech and voice.

The first human graves, clothing and cave paintings reflect the emergence of symbolic thought in man and may perhaps be related, indirectly, to the existence of language. With the advent of writing, at least 25,000 years later, the first direct evidence of reflection and/or action on language emerged. That thinking became clearer, a few centuries before the Christian era, with the articulatory analyses conducted on Sanskrit by Hindu grammarians with a view to preserving their language with its sacred function (reviewed in the 4th century before the Christian era by Panini). It was also in a religious context, that of the advent of the first religion of the Book, with the Bible, as of the first verses of Genesis, that the concept of language as nomenclature emerged: the world was created by the word of God.

The Greek philosophers addressed language in depth but their curiosity was restricted to a single language. We are indebted to those Greek philosophers for the first definitions of the voice and speech. Thus, Aristotle (384-322 BC) clearly distinguished the voice produced in the region of the trachea which is `the sign of the passions of the spirit' (he observed that that voice was also found in other animals). In contrast, `man is the only animal which nature has endowed with the gift of speech'. Aristotle even stated that `speech is natural for man who is by nature a political or civil animal'. Galen, much later (130–200), as a philosopher-physician, contributed additional anatomical and physiological knowledge by associating the voice with the function of the `vocal organs' and speech with organs such as `the tongue […], nose, lips and teeth.'

The Greeks were not content with a partial analysis but went on to develop a veritable synthesis, a general theory of reasoning, from two major techniques, dialectics, the art of controversial discussion (of which the masters were Socrates and Plato), and, above all, rhetoric, the art of public speaking (so important in a democracy).

Objective argumentation uses logical arguments intended to win the auditors' adoption of the new proposals presented by the orator on the basis of common sense or doxa . The second argumentation, subjective argumentation, in addition to the choice of words and arguments, uses the tone of voice and the rate of speech but also gestures, expressions and posture. This approach combines ethos , the moral image that the orator projects through his discourse, and pathos , the passions that the orator elicits in his auditors and through which he elicits their acceptance.

Adaptation to the audience involves the judicial (at the origin of rhetoric), deliberative (on addressing citizens) and epidictic (praising or blaming) genres. The latter genre was to be allocated to literature.

In the various phases of elaboration of the discourse, heuresis ( inventio ) and taxis ( dispositio ) concern the search for and ordering of the arguments, the organization of the discourse, lexis ( elocutio ), the style, the rhythmic figures and also the period. The latter, a true partition updated by the intonation, presents the audience with an argument that it can follow orally from one end to the other promoting continuity of speech that is itself sustained by a sort of vocal legato . This fully reflects the ideal verbal fluidity that was dear to Greek orators. It led them to stigmatize, using a detailed vocabulary, all the cases of constitutional or accidental dysfluency or disfluency. Lastly, the hypocrisis ( actio ) refers to delivering a speech that has been prewritten and memorized before being carefully enunciated. The voice is then advantageously deployed as a component of a motor set in which expressions and gestures also play their role.

In a different social and political context, the Romans were above all to contribute an improvement in techniques aiming to optimize exploitation of the speaker's vocal and gesture potential. The pronuntiatio mainly concerned the voice and the actio the gestures. The orators were practitioners of an effective speech whose esthetic aspect was stressed: eloquence ( bene dicere ). Thus, Quintilian (30–100) in `On the oratory institution' developed a program of teachings. Several passages of the text are devoted to working on the voice, breath, facial expressions and body gestures, without overlooking the hygiene and training of the memory ( memoria ) on which good restitution of the discourse depends.

The teaching of rhetoric was to ensure both the transmission of the knowledge acquired with respect to the function of speech and the voice (in particular the projected voice) but also the uninterrupted transfer of experience accumulated first by the masters of the declamatory art, the `phonasques', then by educators and trainers (for both the speaking and singing voices). Thus, the experience acquired crossed the centuries and remains available to voice professionals: singers, actors, lawyers, preachers and teachers.

All the philosophical reflection of Greek and Latin antiquity, harnessed in the works on rhetoric and grammar (with their phonetic classifications), was to be resumed and developed continuously until the 18th century, and even later, but in a more critical manner.

The Christian Middle Ages attempted to reconcile the heritage of the Romans with the Christian message. The first sentence of the Gospel according to Saint John `At the beginning was the Word and the Word was with God and the Word was God' gave rise to considerable reflection on language, first of all by the Fathers of the Church (Saint Augustine compiled a whole theory of signs), then by the theologians of the Middle Ages. In the 13th century, Saint Thomas Aquinas set up a distinction between the intellectual processes: `the word of the heart' and `the word which includes the image of the voice' (the word thought with syllables) differentiated from the motor processes of speech: `the word of the voice' (the word pronounced with its syllables).

At the start of the Middle Ages, poetry also assumed functions fulfilled by the spoken voice in primarily oral cultures. The other oral form, the oratory art, was incarnated by preachers who were to obey new constraints: the preacher places his voice at the service of the word of God in order to convince his congregation (and threaten his flock when the dangers of sin are to be evoked).

The most direct access to the works of the Greek philosophers was via Arab letters. But the Arab grammarians were also to develop a study of their own language, classical Arabic, which, as was the case with Sanskrit, is a sacred language, that of the Koran. It is thus important to pronounce the language correctly and a fine analysis of the articulation sites was intended to control the pronunciation. The importance attributed to chanting the Koran led to a review of the vocal categories, some of which were considered incompatible with chanting the sacred Text. In particular, it was important to be attentive not to lose the melodic line when the sound was softened.

In the Renaissance, the printed word was rapidly to assume dominance but reading out loud and the art of correctly pronouncing continued to be valued in cultural practices in which orality was necessary: oration, poetry, theater. A degree of suspicion emerged: while a well delivered text affected the audience, the vocal effects were not to supersede the content.

Increasing attention was paid to the `vulgar languages' even though Latin remained the model for grammatical description. The first phonetician grammarians were stimulated by the projected reform of spelling. In addition, curiosity with respect to the anatomy of the human body developed: Leonardo da Vinci drew a series of diagrams of the larynx. During the Renaissance, research on language tended to distance itself from the speculations of the preceding centuries and increasing importance was attributed to the analysis of hard facts.

In the 17th century, while eloquence continued to be displayed in church and at the theater, the first salons, in which speech was exchanged in the intimate context of a relationship involving a small number of selected people, opened their doors. The art of conversation thus developed in a social environment that was privileged in terms of birth or wealth. It was appropriate to respect the proprieties. The correct use of language was regulated: eloquence was an art. Pronunciation of the language continued to be a subject of the grammars (although one of the most prestigious, that of Port Royal, innovated little in that field). Phonetic analyses `for themselves' and classifications of both vowels and consonants were modified and refined. Thus, in 1694, Dangeau published a classification of the consonants which was a precursor of French phonology. Even more so, the 17th century was that of the scientific study of the voice which Marin Mersenne suggested calling `phoniscopy'.

As Patrick Dandrey stressed, `the analysis of the voice in the 17th century was entrusted to the natural philosopher and physician, on the one hand, and to the grammarian and priest, on the other, to the specialist in the human body and to the specialist in the spirit or soul, combining their complementary qualities in the person of the orator who cultivated his voice in order to place it in the service of God or the world.'

The 18th century was marked by a great curiosity with respect to language and, in particular, the origin of language. Philosophical reflection was of primary importance and the most essential works on `the origin of understanding' were those of Locke, Hobbes, Hume, Leibniz and, in France, Condillac, who, in 1746, clearly formulated the theory of the arbitrary nature of the linguistic sign. The clear distinction between the speaking and singing voices, although supported by strong arguments, was called into question by Jean-Jacques Rousseau, author of a veritable anthropological concept of the voice: music and, more accurately, singing were considered to be the origin of the spoken language.

Salons continued to play the role of `resonance chambers' for artists and the learned but the latter could also meet in the Academies, which had almost all been created (the last, the Academy of Medicine, was created at the start of the following century).

Linguistic research progressed with the publication of grammars for an increasing number of languages. The latter development also resulted in the first major attempts to relate languages (Gyarmathi was the first to call attention to the close relationship between Hungarian and Finnish). Persistent discussions on reforming spelling continued to stimulate phonetic analyses which became increasingly independent of the grammars.

The anatomical and, above all, physiological studies of the larynx advanced. The production of the human voice was described by analogy with musical instruments by Dodart, in 1700, and Ferrein (to whom we owe the terms `vocal cords') in 1741.

The first experimental phonetic studies were launched and Abbe Mical and Von Kempelen built the first `speaking machines', the first great mechanical simulations of the phenomena of the production of human speech combining the vocal source and supraglottic resonance chambers.

Speech and the voice subject to specialization

In the 19th century, speech and the voice became the subjects of specific scientific studies. Two main analytical models interacted at first, then opposed one another. The first, the historical model, ensured the independence of linguistics as a historical science. The second, the physiological model, initially used by numerous different disciplines, including comparative grammar, gave rise to the major medical specialities which were to take charge of the hygiene and then pathology of disorders of the speech and voice (without overlooking language disorders).

The 19th century was the century of in-depth reflection on history conducted in the context of German philosophy (Hegel and the different interpretations of his theses). Again in Germany, two major research currents derived from the `discovery' of Sanskrit succeeded each other. Comparative grammar resulted in a number of languages being grouped in the Indo-European family. Subsequently, the neo-grammarians identified, in that evolution, the `phonetic laws without exception'. Profoundly influenced by Charles Darwin's theory of evolution (`The origin of species' was published in 1859), a comparative grammar specialist, August Schleicher, came to consider all languages as organisms and linguistics as a natural science. However, opposition to that point of view mounted increasingly throughout the century, culminating in the idea that `the only method for scientific study of language is the historical method' (the peremptory affirmation of Herman Paul).

The 19th century was also a century of physiology. In the fields of speech and the voice, a new term, phonation, with physiological orientation, led to reappraisal of the voice as the first indispensable stage in the production of speech. Speech was no longer reduced to articulation but put into relation with thought and, thus, with the function of the brain. Broca defined the major brain centers for language (in 1861) and Wernicke defined the centers for oral and written language understanding (in 1874) in relation with various types of aphasia. Thus, a new research avenue came into being, that of neurolinguistics. The research investigated the relationship with the clinical characteristics of acquired speech and language disorders and their repercussions at linguistic level on the basis of an increasing number of cases. In a complementary perspective, the study of hysteria led the neurologist, Sigmund Freud, pursuant to cooperation with Doctor Breuer, to develop the `cure by speech'. Taking into account Freud's complementary thinking on the role played by sexuality and the importance of transfer in treatment, that period (the last years of the century) may be considered those of the birth of psychoanalysis.

In France, the grammarians remained much attached to the `general grammar' (on the Port-Royal model) and were influenced by the ideologists' school, which remained very potent at the start of the century. This delayed the commitment of French grammarians to comparative research (particularly as conducted in Germany). French grammarians preferred more practical tasks such as the compilation of school grammars. Comparable concerns led to the pursuit of research with a view to developing phonetic transcription. At first, this was the work of the `phonographs of vocabulary', dictionary writers who wished to indicate how words whose spelling was likely to induce incorrect pronunciation should be pronounced. But the research also moved further into other fields: the teaching of living languages (beginning with English) with Paul Passy, then Daniel Jones, and dialectology. Dialectology consisted in collecting the spoken forms of the regional languages of France, which were threatened with extinction, particularly with the acceleration of the political centralization movement following the French Revolution in 1789. Two phonetic transcription alphabets were created with two different points of view. That of the International Phonetic Association (created in 1886) was pedagogic in aim and used a limited number of characters. It was more oriented towards the characterization of invariant units. The other alphabet, that of the `Romanists' was intended to reflect the variability of the oral forms of `patois'. The new `instrument' was to renew the attention paid to `aural phonetics'. In a manner complementary to historical linguistic research, `aural phonetics' addressed the phonetic description of the languages then spoken in Europe but with an orientation that remained orthoepic in that it was designed to promote a standard. Phonetic transcription was then endowed with a role (which had long been, in part, that of orthographic writing) of conserving the data on spoken languages. Edison's invention of the phonograph in 1878 called into question transcription. Dialectology, by collecting speech on the very sites of its pronunciation (the French countryside), appeared to be a linguistic `field' study based on observations of subjects speaking and was thus differentiated from speculative linguistics which frequently had an ideological cast. The question as to whether an accumulation of facts can replace a theory remained, however.

Throughout the century, technical innovations enabled enhanced physical and physiological analysis of the phenomena of speech and voice. This was the case, for instance, with the laryngoscope, a mirror enabling in vivo examination of the vocal cords. The instrument , used in 1854 by Manuel Garcia, was mainly promoted by Johan Nepomak Czermak, thanks to the complementary contribution of artificial light. In the field of speech production, the kymograph, invented by Karl Ludwig in 1847, strengthened the development of the graphic method in France under the impetus of Jules Étienne Marey.

Elsewhere, acoustic analysis moved further in two directions which continuously competed with each other. One direction, more mathematical, was derived from the work of Joseph Fourier on the analytical theory of heat (1822) while the other, more physical, was derived from Von Helmholtz studies on the timbre of musical sounds, then speech, using resonators.

The scientific and technical progress was conducive to the creation of experimental phonetic laboratories, one of the first of which was that of Abbé Rousselot. The laboratory was first housed at the Catholic Institute in Paris, then at the College of France in 1897. The laboratory rapidly became a meeting place for phoneticians, teachers, physicians and rehabilitators. Suzanne Borel-Maisonny, a student of Abbé Rousselot, specialized in rehabilitation with respect to disorders of language, speech and voice. Although the term `orthophony' was created in 1828, it was not until the 20th century that `orthophonic rehabilitation' (speech therapy) in an office under medical control was officially recognized. The term phoniatricians appeared in Germany in the context of experimental phonetic research directed by H. Gutzman. The latter was named professor of phoniatrics in Berlin in 1905.

Abbé Rousselot, who had defended a thesis on dialectology and who had a solid training in historical linguistics, consistently presented experimental phonetics as a means of acquiring more refined knowledge of the processes of phonetic evolution in languages even though linguistics were becoming independent as a historical science. Rousselot succeeded in convincing the dialectologists of the value of experimental analyses of their speech but other linguists and even some `oral' phoneticians such as Paul Passy were very reticent and even critical of the use of instrumentation that was considered superfluous (for them, the ear remained the best instrument for analyzing speech and the voice). Even though Abbé Rousselot continued to plead in favor of linguistic phonetics, the various phonetic practices appeared too different to be wholly incorporated in linguistics (particularly as the latter was then conceived).

The first schism was the work of the two most prestigious representatives of the `Kazan school' (Poland), Baudouin de Courtenay and Kruzewski. They proposed splitting phonetics into two independent parts: one, `anthropophonic', considering sounds from the objective viewpoints of physics and physiology, and the other, phonetics in the strict sense of the word or in a `morphologico-etymological' sense, addressing sounds in relationship with the meaning of words. In the latter case, the term sound was rejected and Kruzewski proposed replacing it with the term phoneme (borrowed from Ferdinand de Saussure, who had in turn borrowed it from Dufriche-Desgenette, 1873).

In France, with a different perspective, Michel Bréal (with Baudry) was at the origin of the term phonetics, `science of the sounds of language', created in 1897, with an eye to the term semantics, `science of meanings'. As G. Bergougnioux was to remark in his work entitled `Aux origines de la linguistique française ' (1994): `language has a dual status - material by its sound component […] and mental by its meaning – […]. Linguistics has some difficulty assuming the heterogeneity.'

Whatever the case may be, the experimental phonetics laboratories that developed in France and worldwide over the next century pursued the example of their illustrious forerunner and continued to bring together the various specialists in the analyses of speech and particularly the voice: phoneticians, dialectologists, teachers, physicians, therapists and engineers. All came together with a shared commitment to interdisciplinary research that was all the more laudable in that it would be difficult to maintain over the following century with the increasing independence of the disciplines, their increasing specialization and the institutional consequences of those trends.

New `instruments of reflection' and accelerated technological developments contribute to deepening and broadening the understanding of speech and the voice

Linguistics became increasingly independent with the decisive contribution of Ferdinand de Saussure whose `Course' was published in 1915. Linguistics, for which de Saussure clearly distinguished two approaches, one historical, diachronic, and the other current, synchronic, addressed `language', a social phenomenon defined as a code. `Speech', a set of phenomena rather related to more individual use of language (in that sense, `speech' includes a large proportion of vocal phenomena) moved into the background. The insistence on the formal nature of the linguistic object was taken up by Noam Chomsky with `generative grammar' privileging `skill', an indispensable prerequisite in order to account for `performance', i.e. that which is directly observable in the acts of speech. Many of the phonologists who came afterwards, from structural phonology to the theory of optimality, considered the separation between phonetics, a (natural) science of the substance, and phonology, a science of the form, to have been established. Nicolas Troubetzkoy, to whom we owe that fracture (1939), had set up a research area, phonostylistics (developed later and mainly by Yvan Fonagy), to take account of the expressive and appellative functions of language and no longer only the representative function. Ferdinand de Saussure, whose theses have recently been reevaluated (in particular the dichotomy between language and speech, pursuant to the discovery of new manuscripts), had left a possible opening by making linguistics a field of semiology, a `science of signs at the heart of social life'. The research related to the sciences of culture (Rastier) was to be oriented in that more symbolic direction.

Attempts to develop a typology of voices (speaking and singing voices) via ethnolinguistics and ethnomusicology expanded our understanding of the fundamental mechanisms of the voice and moved study beyond the narrow framework of the cultural restraints of Western civilizations only.

The 20th century and start of the 21st century gave rise to unprecedented expansion of research into language, the various languages spoken worldwide and speech and the voice replaced in a more general context of spoken communication. All the studies are characterized by their increasing profoundness thanks to the use of technological resources that progressed throughout the last century and in the beginning of this one: mechanical analyses were transformed by the development of electronics, itself superseded by information technology. Computer workstations equipped with software that is continuously updated to more or less successfully adapt to researchers' and practitioners' ceaselessly changing requirements have profoundly modified the context of scientific studies of speech and the voice and circulation of the results with respect to both basic research and applications. It would be tedious to list all the technical resources for acoustic and physiological analyses that have succeeded each other. It should be stressed, however, that we owe to those resources a refined understanding of the processes of speech and the voice. Computers enable a single real-time representation of synchronous analyses from different sources. This has culminated in multimedia through which gestures, expressions and exchanges of looks can be related to acoustic analysis of the signals of speech together with perception (giving rise to fine phonetic transcriptions). The analyses are supported by rigorous designs in order to control the extreme variability of linguistic phenomena. However, they may be criticized in that they address a `laboratory speech (and voice)' whose artificial character has frequently been criticized. However, although the value conferred on certain phonetic parameters is debatable, it is possible to simulate speech and the voice through analog synthesis of the vocal tract by a source that is itself simulated. In addition, analytical recognition of speech has been a remarkable tool for elucidation of the crucial processes in acoustic and phonetic decoding (from the acoustic signal to the meaning). Research is continuing on a more mathematical basis currently, in order to ensure reliability, with regard to the various applications of speech recognition.

European `continental' philosophy remains greatly influenced by phenomenology and has addressed intentionality, a concept frequently applied to language conducts. A broadening of the scope of the study of language activity occurred pursuant to the philosophical reflection on `ordinary language' initiated by Wittgenstein and pursued by several representatives of analytical philosophy, particularly in the English-speaking world (with Austin for the pragmatic approach). Conversational analyses focus on everyday language and study, for example, the succession of speakers in usual communication settings. The analyses also address the relative share of individual skills and the social context in the organization of language exchanges. The subject of conversation analysis is then the discourse in the interaction with its system of turns of speaking (the listener is more than a simple foil in the active communication). Thus, a fine description of the organizational forms specific to conversations based on detailed analyses of authentic interactions leads to a search (which may often appear mythical) for a `spontaneous' speech whose vocal components, the prosodic and, above all, melodic and rhythmic phenomena, but also the gestures, play a major role in the meaning of the statements. The studies have shown the disfluency which can characterize conversational speech with its pauses, hesitations and false starts, etc., which it would be good to distinguish (but is that always possible?) from the dysfluency that is more related to language disease.

The separation between speech and voice as two very distinct stages in the production of the phonic units of language is no longer so assured. This is, moreover, the conclusion reached by speech therapists when they target the placement of the voice which is largely affected by supraglottic phenomena. In the same perspective, a complement to the International Phonetic Alphabet was compiled in 1994 to transcribe pathological language. Not only speech is concerned by the extension (extIPA) but also vocal quality thanks to the use of specific characters. A system of labeled markers enable isolation, in the phonetic transcription, of the particular vocal sequences of the enunciation.

A new stage was reached, particularly in the last third of the 20th century, with the development of cognitive sciences which analyze the function of the brain in the production of speech using brain imaging (generally speaking, this is a new stage in the in vivo use of medical imaging that has ceaselessly developed over the century). Prior neurolinguistic research had already shown the fascinating complexity of brain function in the pathological changes in language characterizing the various types of aphasia.

Integrating the data derived from cumulative experience (frequently of a holistic nature) over more than 2000 years, from the Greek `phonasques' to the speech therapists, logopedists and phoniatricians, and scientific laboratory analyses, also fragmentary, remains a complex undertaking. The increasing profoundness and widening scope of the analyses of speech and the voice are not devoid of new epistemological questions: How can the complexity be managed? By using an interdisciplinary approach? Experimental design enables control of the variability of linguistic phenomena but is necessarily frequently accompanied by a simulation or more generally a modeling that is more synthetic (hence the crucial role played by formalization).

And what about speech and the voice in the classroom? The teacher's position with respect to the students has profoundly changed, as can be readily observed today, at the beginning of the 21st century. It would appear increasingly difficult in teaching, at all levels, to deliver lecture-like courses. The need to introduce a significant amount of interaction in the spoken communication in class is now widely admitted. The need necessitates regulation of speakers' turns and organization of discussions that must itself be judiciously regulated. The heterogeneity of the student groups may give rise, for an increasing number of students, to a feeling of linguistic insecurity and a malaise in teachers who are frequently situate their `public'. In consequence, not only must sociolinguistic factors, which are frequently stressed, be taken into account, but also, and above all, the sociocultural factors in the new relationships between students and teachers.

The sketch of the historical epistemology of speech and the voice has frequently disclosed a strong ideological investment: what is in question is what is specific to man with, in the background, the opposition between nature and culture. Another major separation reinforces the first dichotomy: the speech of reason responds to the voice of passion. But has man always mastered his speech and, to an even greater extent, his voice? Pathological psychology research results, particularly those of the `master of suspicion', Sigmund Freud, provide grounds for doubt. Moreover, the voice characterizes, in itself, human experience. Have we not then overestimated the meaning of speech to the detriment of that which, drawing on the philosophers of the Middle Ages, we might call the significance? All meaning transits through a particular speaker whose voice takes on the personal resonances of his experience.

From glottal sound to speech

The vocal fold or plica vocalis is a composite fleshy fold. Two vocal folds are positioned inside the larynx, at the juncture between the trachea and the pharynx. They are oriented in an anterior-posterior direction. Toward the front, they rest on the thyroid cartilage (Adam's apple) and toward the back on two small cartilages known as the arytenoid cartilages. The speaker controls the inter-fold gap, as well as lengths, shapes and tensions of the vocal folds by contracting them and changing the relative positions of the laryngeal cartilages. The 3-dimensional space between the vocal folds is known as the glottis.

The sound produced by the pulsed airflow through the oscillating glottis is called voice. The voice is produced via the transformation of aerodynamic into acoustic energy. This transformation depends on several aerodynamic and biomechanical factors (subglottic pressure, mass, stiffness and viscosity of the vibrator, tension and contraction of the intrinsic and extrinsic laryngeal muscles). The laryngeal sound propagates through the resonating supraglottic cavities of the vocal tract, whose morphological and parietal characteristics (energy losses, resonances) determine speech timbre and contribute to voice quality.

Each vocal fold consists of a mucous membrane, ligament and vocal muscle. The mucous membrane is separated from the vocal ligament by a slippage space known as Reinke's space. The mucosal wave is caused by the subglottal air during expiration, separating the closed vocal folds. The undulation occurs along three axes: from bottom to top, anteroposterior and lateral. The chemical constitution of Reinke's space, in particular the presence of hyaluronic acid, plays an important role in favoring the flexibility and slippage of the mucosal membrane over the vocal ligament and contributes to the efficacy of the laryngeal vibrator.

Vocal fold vibrations are self-sustained. They result from a circular chain of causality, which relates elastic and viscous forces generated by the movement of the vocal folds to the aerodynamic forces in the glottis. The latter derive from the flow of air from the lungs. By regulating the manner in which the vocal cords vibrate, the speaker determines the timbre, pitch and loudness of his vocal emission, controls sentence intonation, accentuates syllables and switches between voicing and non-voicing of the speech sounds.

Voice disorders are the felt, perceived or audible consequences of an anomaly or acquired or congenital organic lesion of the vocal folds. Disorders may also be due to defective control of the respiration or the laryngeal articulation or a psychological problem. Acquired organic lesions may reflect vocal abuse or overuse.

Several physiological and environmental factors may contribute to impairing the vibrator. A subglottal pressure that is too strong, in the low register, at high intensity, will increase the amplitude of the mucosal wave, which may become excessive and give rise to microtraumata that cause tissue changes and the appearance of nodules. Aging affects individuals unequally and may have an effect on the flexibility of the vocal folds. Voice training appears to slow the aging of the voice. Dry air is reported to increase the viscosity and rigidity of the mucous membrane with, as a consequence, changes of the voice acoustic and efficacy (more effort is required for the same result). Adequate fluids intake contributes to correct laryngeal function. Finally, the environment has an influence on the function of the vocal folds: physical factors (inhaled substances), biological agents (viruses, bacteria), irritant chemicals, and pharmaceuticals (hormones, drugs) may all play a role.

Acoustic analysis of the voice contributes to the vocal assessment

The physical etiologies of voice disorders are various: abnormal vibratory modes of the vocal folds, excessive turbulence of the air, amplification of vocal tremor or `jitter', parasitic vibrations of laryngeal structures that normally do not vibrate, uncontrolled transients between different vibratory modes. Irrespective of the exact etiology, rapid disturbances in the durations and amplitudes of the speech cycles are frequently referred to as vocal jitter and shimmer (or shimmy). Other vocal symptoms include dysprosody and deficient coordination of articulation and voicing.

Frequently, the effects of voice disorders on speech are described using acoustic, mechanical or aerodynamic cues, which are quantitative data obtained from the speech signal or other signals (e.g. electrolaryngography, contact microphone or measurement of the air flow rate) that are recorded in a non-invasive manner. These cues summarize properties of the speech signal that are clinically relevant. They are selected on the base of the laryngeal disorders that are investigated, the vocal symptoms described, the searched-for correlations with other levels of description, the tasks requested from the patients, the linguistic performances involved as well as the instrumentation and type of the recorded signals.

Voice assessment includes an acoustic analysis and perceptual evaluation of a patient's vocal emission. Often, acoustic indices are obtained using dedicated software. Software that is available in clinical practice is based on research that has been performed 10 or 20 years ago, and which does not include recent analytical methods meeting clinicians' current requirements.

Expert assessment concerns the evaluation of vocal handicaps that are considered to be the consequence of a voice disorder related to the practice of a profession or exposure to a vocal risk. The assessment of the handicap of a professional speaker may be problematic because intensive and prolonged professional use of the voice is difficult to simulate under clinical or laboratory conditions.

Currently, few analysis systems are available that are suitable for use in the field. Those that are have been purpose-built. As a consequence, little experience has been accumulated and use is not widespread. There is no consensus on the cues that must be recorded.

Few studies are available on the use of acoustic cues of voice quality during vocal rehabilitation. Such an application of acoustic cues would consist in informing the patient on his voice and the time course of its quality. Use in rehabilitation is legitimated by the observation that only acoustic cues carry relevant information on the patient's voice, which is the same for the caregiver and the patient (biofeedback). The speech signal propagating in the air is the only signal that is shared by listener and speaker, who does indeed not perceive his voice in the same way as the listener. Moreover, the speaker, obviously, cannot directly observe his/her glottal articulation or vocal folds.

Perceptual evaluation and acoustic analysis of voice disorders are components of voice assessment. In practice, clinicians expect a feeble or moderate correlation between perceptual evaluation, acoustic cues and patient self-assessment. In general, clinicians are resigned to observing that voice assessment requires a multilevel representation of the voice with a moderate correlation between the perceptual features, acoustic cues and subjective indices.

This is the situation that currently prevails in clinical practice. However, it is not easy to determine to what extent it reflects listener cognition or lack of understanding of the perceptually relevant acoustic cues. Perceptual and acoustic voice analysis combined with acoustic voice synthesis might contribute to a systematic investigation of the acoustic disturbances and clinically relevant percepts.

Perceptual evaluation by the caregiver and self-assessment by the patient also contribute to the vocal assessment

The voice is multidimensional. Multi-parametric analysis is therefore required for diagnosis and in order to orient disorder management. Among the various determinations available, perceptual evaluation plays a special role. Perceptual evaluation consists in determining, on the one hand, the timbre of the voice and its esthetic qualities, while, on the other hand, determining the articulatory and prosodic behavior, and the intelligibility of the speech. Numerous perceptual scales are available and vary as a function of the evaluation criteria targeted (timbre, intelligibility, linguistic criteria). The GRBAS scale consists of 5 components (grade, roughness, breathiness, asthenia, strain) and 4 severity grades for each component (0, 1, 2, 3). For instance, a very dysphonic, moderately rough, slightly breathy and somewhat strained voice would be scored: G 3 R 2 B 1 A 0 S 2 . It mainly evaluates timbre and phonatory behavior. It is the most widely used scale and does not depend on the user's language or professional skill. In addition to perceptual evaluation, the subject's own experience of his/her voice disorder, and its repercussions and consequences on the subject's everyday life is to be incorporated in the assessment of voice disorders. Several instruments measuring quality of life are available. The most widely used scale is the `Voice Handicap Index' (VHI) which comprises 30 items divided into 3 subscales: physical, emotional and functional. An abridged version consisting of 10 items is used in the United States.

A voice disorder may have a markedly greater impact than its solely acoustic expression because of its repercussions on the subject's professional and social life. It can as well have consequences on the mental, physical, emotional and communicational life.

The medical criteria used to evaluate a voice disorder do not reflect the degree of communicational handicap. It is therefore important to allow the patient to evaluate that handicap, particularly since the assessment may not always be correlated with the objective reality of the dysphonia.

The WHO international classification of handicap considers handicap to be the consequence of a disorder that limits an individual's activities both through the objective disorder and through the personal and environmental factors that may modify the patient's perception of his disorder. The definition is thus multidimensional and admits that two people with similar disorders may experience different limitations of their activities due to personal and environmental differences. "The classification takes into account the social aspects of handicap and proposes a mechanism for determining the impact of the social and physical environment on an individual's function. The environment is to be adapted to the person, not the person to the environment."

The US FDA has estimated that voice disorders cost between USD 30 to 150 billion per year in terms of losses of productivity, treatment and education. In France, no public health study has yet evaluated the financial cost of voice disorders for voice professionals.

Different forms of the disease are associated with the professional use of the voice, particularly for teachers

A number of laryngeal lesions, particularly of the vocal cords, are related to tissue reaction to biomechanical overload (supraphysiological phonatory conditions, phonotrauma). The lesions consist in erythema and edema of the vocal cords (concept: `laryngitis'), vocal polyp or polypoid forms (acute phonotrauma), vocal cord nodules and, to a degree, contact ulceration and dorsal granuloma on the vocal cords (chronic phonotrauma). However, some of those manifestations are not specific in the sense that phonotrauma is not necessarily the only etiologic agent. For example, laryngeal erythema may in part derive from environmental factors (dust, dry air, etc.). Vocal cord polyp (in particular the angiectatic form) preferentially occurs on a mucosa that is already congested (for example, due to cigarette smoke). Contact ulceration and dorsal granuloma may in part result from gastroesophageal reflux. Vocal cord nodules may be considered the most specific lesion of chronic phonotrauma at an early stage when the symptoms are still mild. Nodules are identified at a significantly higher frequency in female primary school teachers than in a paired homogeneous group (nurses). For some other laryngeal diseases, intensive use of the voice (which may be accompanied by straining) is considered a possible etiologic cofactor: this is the case for Reinke's edema (polypoid corditis) and hyperplastic chronic laryngitis in which smoking and alcohol are the main etiologic agents. Minor congenital anomalies (sulcus, mucosal web in the anterior commissura) may play a promoting role as may diathetic factors (allergy of the upper airways and treatment by inhaled steroids) or chronic diseases such as gastroesophageal and gastropharyngeal reflux.

The etiology of vocal cord nodules appears to be associated with a particular vibratory mode of the vocal cords which limits the contact (collision, then detachment) between the free margins of the vocal cords at a particular site: the union of the anterior third and posterior two thirds. The vibratory mode requires three biomechanical conditions: incomplete dorsal adduction, a curved resting position (rather than rectilinear) around which an oscillatory movement occurs, and a sufficient oscillatory amplitude (without which contact does not take place).

Voice disorders in general and vocal cord nodules in particular are very largely preponderant in women. The main factor is the mean vibration frequency of the speaking voice and hence the frequency of potential microtrauma (115 Hz in men and 210 Hz in women). In addition, incomplete dorsal glottal closure is almost to be considered physiological in women (2/3 of normal subjects). The slight curving of the free margin develops with vocal fatigue and the sufficient oscillatory amplitude required to induce microtrauma naturally accompanies voice projection and the need to raise one's voice. The tissue reactions are initially reversible but subsequently become definitive.

Environmental factors are also important: the acoustics of the classroom, outside background noise, the noise of the class, excessively dry air, etc., together with the stress that accompanies speaking. Stress exerts various physiological effects, generally negative, but with a few exceptions. Those effects include dryness of the mucosa, postural rigidity, undifferentiated increase in the tone of the laryngeal intrinsic muscles, changes in respiration, etc.

Screening methods and aptitude tests remain little developed. A few attempts have been made based on case history-taking and perceptual evaluation of voice quality, laryngoscopic clinical examination or certain physical voice tests such as maximum phonation time (MPT) determination or study of the dynamics of the fundamental frequency of speech as a function of imposed sound intensity levels. At present, the sensitivity and specificity of those tests are not sufficient for their widespread use to be advised.

Voice disorders occur more frequently in female teachers

Most of the studies that should enable estimation of the prevalence and incidence have exploitation and interpretation limitations, in particular due to the absence of a consensus definition of the functional disorder. There is no clear medical nosologic classification. In the International Classification of Diseases (ICD), voice disorders are listed under several headings (ENT, neurology, psychiatry).

From a methodological point of view, most of the studies of voice disorders are based on:

• self-assessment questionnaires (frequently differing between studies) but which enable statistically adequate recruitment;
• specialized investigations (video laryngoscopy, recorded-voice evaluation software) limiting the size of the samples and for which result interpretation depends on the operators. These difficulties explain the great diversity of the published figures.

In the overall population of the United States, the prevalence of voice disorders was estimated to be between 3 and 9% in the most recent studies.

With regard to teachers, epidemiological studies of voice disorders have been conducted in European countries (Finland, United Kingdom), the United States and Australia. Several studies show that voice problems occur more frequently in teachers than in the overall population and that the difference is significant. Thus, in the United States, the prevalence is about 12% in teachers and 6% in non-teachers. A recent Finnish study showed that the prevalence of vocal symptoms in teachers had significantly increased over 12 years (1988–2000).

Studies have also shown that in teachers who remain in the profession, the prevalence of voice problems increases with age, peaking in the age group, 50–59 years. The prevalence is very markedly higher for women.

Voice disorders as a function of age (after Roy et al., 2004).

In France, voice disorders are not clearly identified as an occupational disease despite the large number of teachers who account for 2.7% of the country's working population. Very few studies have been conducted in that occupational sector. However, voice disorders are beginning to be taken into account in the teacher-training establishments (IUFM), which are attempting to set up awareness promotion and prevention programs.

In a study of the French National Education General Mutualist Association (MGEN) conducted in 2005, (10,288 survey respondents of which 3,904 working teachers), among the teachers, 26% of the men and 50% of the women reported always or frequently having voice disorders. The trends observed show that the prevalences of voice disorders are markedly higher in female nursery-school teachers.

In the teaching profession, an increase in voice-related complaints has certainly occurred over a few decades. The increase is probably related to the greater number of women working in educational professions, particularly in nursery and primary schools. Female gender is a risk factor for teachers. Certain types of teaching are associated with a higher risk (music, nursery school, primary school, sport). The studies to date have not evidenced a higher risk in language teachers.

In the United States, voice disorders give rise to 2 days of sick leave per year and per teacher for an estimated total annual cost of USD 638 million and a substitute teacher cost of USD 220 per day. In Finland, 5% of teachers have their professional capability called into question due to voice disorders.

Environmental conditions are risk factors to be taken into account

Ergonomic factors are to be taken into account among the risk factors for voice disorders. In the first place comes acoustic pollution of classrooms, amphitheaters and gymnasiums which constitutes a major factor for straining the voice. Recommendations exist in France but only cover new buildings and they are perhaps inadequate compared to the WHO recommendations. Based on the acoustic recommendations, the French recommendations do not sufficiently take into account the sensory difficulties in auditory perception that are frequent in children and the need for additional acoustic power from the speaker. This may be necessary with children whose mother language is different from the language used for teaching (signal/noise ratio of about 15 dB). External acoustic pollution (urban environment, airport, etc.) is a recognized factor with regard to learning difficulties in younger children.

Acoustic standards in classrooms

View in own window

RT60: time taken for the sound pressure to decay by 60 dB after the source has stopped emitting; Leq: level equivalent

Other factors are to be taken into account:

• the duration of the teacher's phonation; complaints are more frequent in the afternoon;
• the relative humidity of the facility air, a factor frequently mentioned in studies and correlated with good hydration in order to ensure good quality mucus which is necessary for the correct vibration of the vocal cords;
• the quality of the air in order to ensure, in occupational facilities with non-specific pollution, the required number of air changes, taking into account the number of students in the class;
• air pollutants. Dysphonia due to chemical pollutants (ammonia, welding fumes, solvents, diesel fumes, mold, ozone, formaldehyde) has been identified in a few studies conducted in occupational environments. In a single recent study, physics and chemistry teachers were shown to be at vocal risk. Teachers in vocational high schools are also possibly at risk, particularly when facility ventilation is inadequate.

Some teachers do not complain of voice disorders. This raises the question of between-individual differences in the emergence of disorders. The variability may be due to constitutional factors but also perhaps to the development of more pertinent vocal adaptation strategies whose recognition and implementation would be a simple means of prevention. The need to adapt to different categories of students and to different facilities is a requirement of the teaching profession and may explain the occurrence of voice disorders in subjects without vocal disorder history, sometimes at the start of the career, or even sporadically throughout the career. European directive 89/391/EEC dated June 12, 1989, is a framework directive that defines the fundamental principles for the protection of workers and provides a template for priority implementation of collective solutions: necessary information for teachers on the vocal risk (giving appropriate instructions), a priori assessment of risk factors (acoustic risk, chemical risk), combating risks at the source by ensuring the absence of degradation, the improvement of the acoustic characteristics of classes by judicious location of establishments (external acoustic pollution), improving ventilation, replacing hazardous products (felt-tip pens containing solvents, dry effacement of boards). Individual preventive measures may be adopted by teachers suffering from voice disorders by adaptation of the work to the subject taking into account technical progress (use of amplification systems, creation of a vocal rest room for dysphonic teachers). With regard to reducing noise levels in classrooms, it is possible to attenuate intrusive noise by enhancing sound proofing and to limit the relative disturbance induced by technical equipment by modifying its installation. With regard to reverberation control, it is possible to position semi-absorbent acoustic panels, taking care not to excessively decrease the reverberation time. That type of arrangement also enables absorption of the noise generated by the students. A last type of treatment, which is very important, consists in strengthening the initial reflections (those which take less than 50 ms to reach the ears) whose role is to naturally strengthen the voice and the intelligibility of the message and thus decrease the teacher's vocal effort.

It should be noted that in other occupational settings different from teaching a vocal risk is also present. In modern societies, for one third of the workers, the voice is the leading tool. This applies to call-center operators, reception personnel and workers who communicate in noise-polluted settings.

The diagnosis of speaking voice impairment requires functional and etiological assessment

The diagnosis of a speaking voice impairment or dysphonia is conducted in the course of a phonation assessment. The assessment includes a functional assessment and an etiological assessment.

The functional assessment is clinical and instrumental. The interview is undoubtedly the longest part since it is important to elucidate the complaints (vocal fatigue, straining, respiratory difficulties), the results of vocal straining or overwork, professional constraints and working conditions, and identify the associated symptoms and all `exacerbating' factors. This enables assessment of the subject's psychological status and the potential impact of vocal disorders, assessment of `vocal behavior' (respiratory technique, posture, muscle relaxation), patient's self-assessment of the quality of his/her voice and perceptive analysis. The GRBAS scale is increasingly used.

All the acoustic parameters of the voice may be impaired: frequency (or tone), intensity, timbre, rate, articulation. Objective acoustic analysis enables measurement of all those parameters. The difficulty resides in the absence of a consensus with regard to the pertinence of the indices measured, their correlation with the subjective evaluation and the anatomical findings. This is due to the instability of the voice as a signal from a physical point of view, the manner in which the signal is acquired (acoustic or microphone, electrolaryngography), the phonetic material analyzed (vowel or continuous speech), vocal emission conditions (mean intensity, high intensity, usual frequency, speaking voice, singing voice, etc.). All the analytical systems are computerized with dedicated programs ranging from the most simple to the most sophisticated. The programs concomitantly analyze one or several parameters. There is thus a question of choice and resources.

Acoustic measures determine the usual tone or fundamental frequency of the voice, its irregularities or jitter, and its mean intensity and intensity irregularities or shimmer. An examination frequently conducted by phoniatricians and speech therapists is the phonetogram, or dynamic range of the voice, which correlates two acoustic parameters: frequency and intensity. The test protocol has been standardized by the Union of European Phoniatricians.

The analysis of timbre is more complex and the interpretation of spectral analysis data is difficult except for trained specialists.

Aerodynamic measurements determine air pressures and flow rates. Some are easy to conduct such as the maximum phonation time (MPT). Others require more sophisticated apparatus with flow rate transducers.

Etiological assessment is conducted by a specialist physician, an ENT specialist or phoniatrician. The physical examination studies the quality of the resonators (mouth, teeth, tongue, soft palate, temporomandibular articulation). Laryngeal examination is an essential and indispensable component since it addresses the morphological aspect and dynamics of the vocal cords. The examination is conducted with flexible (nasoendoscopy) or rigid endoscopes enabling good quality imaging and enabling adjunction of a stroboscope. In general, digital video recording is conducted during the investigation. The morphology of the vocal cords (color, length, lesions) and their mobility are studied, while the stroboscope enables determination of the quality of their contact and the amplitude and symmetry of mucosal undulation. A standardized examination protocol recommended by the European Laryngological Society is available.

Following the vocal assessment which is to at least include a functional assessment (conducted by a phoniatrician or speech therapist), acoustic determinations (fundamental frequency, intensity), MPT determination and laryngeal endoscopy, dysphonia may be diagnosed: the dysphonia may be purely dysfunctional with anatomically normal vocal cords, or vocal cord lesions may be present in a context that is frequently dysfunctional. Bilateral nodules, unilateral polyp or edema of the vocal cords are the lesions that are most frequently encountered, particularly in teachers.

The prognosis is assessed and treatment prescribed: simple medicinal treatment for laryngitis or gastroesophageal reflux, speech therapy or phoniatric rehabilitation to restore the `correct vocal practice' in the context of dysfunction, surgical treatment, if necessary, which is given concomitantly with or followed by rehabilitation since the dysfunction frequently `creates' the lesion, over time.

Further investigations may be necessary for more accurate diagnosis or in order to prescribe particular management: hearing assessment, respiratory function tests, imaging (laryngeal CT-scan or MRI) for a mobility disorder or trauma, laryngeal electromyography in the event of a disorder of mobility or if spasmodic dysphonia is suspected (laryngeal dystonia). The latter enables both diagnosis and treatment by botulin toxin injection.

Vocal rest is the best anti-inflammatory for the acute phases of dysphonia

In the acute phases of dysphonia, most frequently in the context of a chill with laryngitis but also in the event of pharyngeal or bronchopulmonary inflammation or an acute episode of vocal straining, resting the voice is the best anti-inflammatory. This means that the teacher requires a few days of sick leave. Optimal anti-infective treatment indication is outside of the scope of this expert review since treatment depends on the clinical situation. In certain cases of severe dysphonia not resolving on rest or in the event of particular occupational requirements, anti-inflammatory treatment with systemic corticosteroids (prednisolone: 1 mg/kg/day) may be prescribed. The treatment is of short duration and dosage tapering is not required. Patients with peptic ulcer require concomitant gastric demulcents.

The efficacy of aerosols, particularly in the event of medicinal combination therapy, has not been demonstrated. With regard to drug administration by aerosol, in line with the recommendation of the consensus conference on pneumology, it is preferable to use only normal saline aerosols (with a view to moisturization) possibly with a steroid (anti-inflammatory). Corticosteroid sprays are intended for the treatment of bronchopulmonary diseases, particularly asthma, and are not appropriate for the vocal cords. Their use has even been suspected of inducing dysphonia related to the propellant gases in some cases.

Vocal cord surgery is one option for voice disorder treatment

The surgical procedure consists in treatment of lesions that have become established over time. Most of the time, the specific lesions following voice straining in teachers are responsive to rehabilitation and surgical treatment is only used in the event of failure of the latter.

Overworking and straining the voice are the key factors in the emergence of dysfunctional lesions. Lesion management must thus focus on the dysfunction and vocal cord surgery is always to be accompanied by vocal hygiene measures such as the use of an amplification device, medicinal treatment of concomitant diseases and, above all, rehabilitation measures. In all cases, speech therapy is to be prescribed at least post-operatively. It is preferable for the patient to meet with the speech therapist before the surgical procedure.

For the chronic phases of dysphonia, there is no specific treatment having demonstrated its efficacy with regard to improving the voice. A number of patients have nonetheless been treated for diseases considered promoting factors for dysphonia. This is particularly the case for gastric antacid treatment in patients presenting with gastroesophageal reflux.

Surgery consists in insertion of a laryngoscope (metal tube) into the oral cavity with the patient under general anesthesia. Orotracheal intubation is usually conducted. The tube enables imaging of the vocal cords directly or via a surgical microscope. The procedure is short but in most cases the patient is hospitalized at least for a day or one or two nights depending on the healthcare establishment's regulatory regime. As with any surgical procedure, the patient is to be warned of the risks associated with the procedure.

Surgery is conducted using specific instruments enabling the basic surgical procedures: retraction, detachment, gripping, section. Some surgeons use a CO2 laser whose beam, coaxial with the microscope view, enables the mucosa to be sectioned and microvessels to be coagulated. With the more recent laser systems, the unwanted thermal effects (burning of the mucosa around the section site) are minimal and currently there is no documented reason for recommending section by microsurgery rather than laser section or vice versa.

The treatment of nodules and polyps consists in gripping the polyp and retracting it from the vocal ligament plane in depth so as to implement excision with minimal healthy tissue damage. Some surgeons conduct infiltration of Reinke's space in order to implement a `hydro-dissection'. Excision may be implemented by microsurgery or by laser. After excision, it is not necessary to suture the excision margins. The wound bleeds very little. The excision specimen is systematically forwarded to the histology laboratory for analysis. On returning to consciousness, the patient generally does not suffer at all. Although there is no consensus, absolute rest of the voice for a few days is usually recommended.

The other surgical procedures are variants on the foregoing. Edema is sometimes very similar to a polyp from the point of view of appearance and consistency. Excision is conducted in accordance with the same principles. In other cases, the edema is more `fluid' and may be aspirated after practicing a longitudinal section of the mucosa at the superior surface of the edema. Granuloma of the vocal apophysis is usually treated by laser excision. At the end of the procedure, some surgeons conduct a complementary procedure consisting in mitomycin application or cortisone infiltration in situ . Other surgical procedures are not specific to teachers and will not be considered here.

For teachers, surgical treatment is most frequently combined with sick leave. During the post-operative period, authors recommend suspension of vocal activity for at least 3 to 6 weeks, depending on the surgical procedure conducted. Resumption of professional vocal activity fulltime will only take place after assessment of the quality of the voice and its potential since there is a strong risk of relapse in the event of work resumption before complete cicatrization of the vocal cords. In certain cases, if possible professionally, resumption of teaching part-time is suggested.

Rehabilitation is to be adapted to the subject's personal situation and the context of the dysphonia

The laryngeal organ which ensures vocal vibration also has other roles: a sphincter to protect the upper airways, contribution to deglutition and sometimes a degree of regulation of the respiratory rate.

The voice is intended not only to be heard but, through speech, to have an action on the interlocutor (voice projection and projectional act). This points to the role of the voice in speech, i.e. in the verbal exchange relationship. The role of vocal projection in the verbal exchange is thus to be defined. Speech contributes to psychological dynamics. The great concepts of modern psychiatry enable illustration of the psychological functions of the voice, which define it as the founder of a vocal identity.

The voice is produced through the synchronous functioning of several systems that have functions that are usually independent: the oral cavity and tongue, the larynx, the respiratory system and the abdominal and diaphragmatic muscle system. These systems when acting in conjunction are subject to constraints in the implementation of their movements or combinations: if the respiration is excessive and very energetic, it is associated with a cervical muscle contraction which tightens around the larynx and reduces its mobility; excessively low or high laryngeal pressures reduce vibratory efficacy; the latter may be offset by linguopharyngeal tone modifying the vocal timbre. The constraints thus have a limited margin for adaptation beyond which symptoms will develop, clinical then organic.

When the constraint margin of the system is exceeded, dysphonia occurs. The voice is thus not univocal and management needs to be organic, functional and psychological to ensure complete coverage of the crucial role of the human voice.

The management of vocal symptoms is generally conducted by ENT specialists, phoniatricians or speech therapists. They consider the patient overall and are able to distinguish the organic, somatic, functional and psychological factors. This calls for receptiveness, empathy and professionalism. Management by rehabilitation calls for the personal maturity of the caregiver in order to comply, first of all, with the cultural identity and vocal requirements of the social group to which the patient belongs, while taking into account the psychological dimension specific to the voice and to the principles of communication. Rehabilitation is thus to be tailored to the patient and conducted at the patient's rhythm in compliance with the diagnostic indications and the medicinal or surgical therapeutic orientations.

Singing teachers and coaches manage vocal problems that may be encountered by teachers. Management is not a question of therapy but of optimizing vocal potential. The teaching is intended to enhance the vocal gesture for esthetic aims, at the rhythm of the musical group or choir or in accordance with the requirements of the score.

Specific voice research has markedly developed since the start of the 20th century through the actions of numerous learned societies and research laboratories. Training is given to teachers through teacher-training colleges (IUFM) but still remains too rare, although teachers are convinced of the benefit of voice training, particularly since 60% of teachers have already experience vocal symptoms (fatigue, aphonia, loss of strong voice, change in timbre).

Rehabilitation is based on the organic and functional diagnosis insofar as rehabilitation will have specific features depending on the disease considered. The main organic lesions to which teachers are exposed are lesions due to straining: polyp, nodule, edema. Dysphonia of a functional etiology includes vocal fatigue, changes in timbre, loss of strong voice and aphonia. Psychogenic aphonia requires a specific approach targeting the context of occurrence and specific management.

Rehabilitation is to be adapted to the teacher's personal situation and to the context of dysphonia (junior school, nursery school or university, noisy facilities, disruptive students, etc.). Immediate assistance is to be offered, such as voice amplification, which is particularly effective. The caregiver must also take into account the stress factors which are particular for teachers. The caregiver must be attentive to resolving them in as far as possible. The objectives of vocal rehabilitation take into account the nosologic context: pre- or postoperative, isolated dysfunctional dysphonia or dysphonia associated with a laryngeal disease, psychogenic dysphonia.

In order to orient the `treatment program', caregivers use particular instruments. The caregiver first defines the treatment aim with the patient by assessing voice quality (GRBAS scale, study of posture, patient's ability to modify his vocal production, patient self-assessment of voice). The acoustic measurements are mainly used to strengthen the patient's self-assessment and enable the patient to assess progress. The acoustic measurements also reassure and encourage the patient. They also demonstrate the state of the voice and constitute forensic documents. Stroboscopic examination generates crucial information for the caregiver but also acts as a visual aid enabling the patient to understand his/her dysphonia. Initial care addresses the information and advice given to the patient with respect to eradicating irritant factors, voice misuse and the responsibility of environmental working conditions.

Care then addresses the conditions of vocal projection and the situations and dynamics thereof. The various styles of voice, related to intention, are addressed: reflective and confidential voices, vocal projection voice (containing both spontaneous speech and prepared speech), the voices of opposition, insistence and distress. Voice placement work thus takes into account the overall bodily dynamic: laryngo-pharyngo-buccal, the sites of vibration, pressurization and articulation. Voice placement also takes into account breathing and its use in phonation. There are two main poles: the superior tension triangle (laryngeal, pharyngeal and buccal region) and the inferior tension triangle (low back and abdominal region) which are particularly involved in voice straining and require relaxation, depending on the case, to regulate and re-equilibrate the tensions.

Rehabilitation training is of two types:

• during rehabilitation, the patient may learn or further elucidate the manner in which he/she articulates or makes a sound or syllable resonate. This study isolated from the context of communication and only centered on implementation technique may be considered mechanical;
• when it is necessary to reintegrate the mechanical implementation of vocal production within a verbal chain, rehabilitation addresses the truly functional aspect, i.e. re-associating speech with its semantic content and linguistic values with the intention of the speaker acting on his/her interlocutor. The vocal and socio-phonatory code modify the mode of implementation, which moves from the mechanical to the functional.

Traditionally, in France, speech therapists offer on average 30 half-hour sessions once per week, frequently with two sessions per week to begin with. According to the international literature, the number of sessions is often more limited. No comparison of the efficacy of different rehabilitation session durations for a given disorder is available.

The criteria for resolution of dysphonia are as follows:

• a voice present at all times and under all circumstances of life, enabling what is usual with the voice: talking quietly and loudly, shouting, calling, singing;
• no tiring of the voice or, if there is tiring, tiredness that rapidly and spontaneously resolves;
• auditor understanding at syntactic, semantic and pragmatic level: the voice reflects the feeling experienced thanks to the prosody that it confers on the speech;
• a voice considered an integral part of the user's personality and identity.

The assessment of the therapeutic efficacy of rehabilitation techniques is based on numerous studies comparing and contrasting the results of various investigations such as the GRBAS, phonetogram, laryngeal stroboscope, acoustic and aerodynamic data, handicap self-assessment and listening tests. The assessments have confirmed the value of voice rehabilitation in the event of symptomatic impairment of the voice due to straining mechanisms.

Voice rehabilitation uses deliberate attention to transform the pathological vocal gesture. The aim is to confer, through that learning, a vocal automatism, the correct vocal gesture.

Training is an essential component of a preventive program

Under the European legislation, various organizations are responsible for identifying voice problems. The organizations take into account the health risks and consider whether voice problems require recognition as an occupational disease. The European Committee for Standardization has defined the ergonomic criteria which may be applied to the field of the voice and speech. The organizations have proposed the term `vocoergonomics' for a multidisciplinary field which combines both the scientific and practical aspects. The European Agency for Health and Safety at Work (EAHSW) recommends that the employer should take measures to ensure the safety and health of the employee. This is documented in European directive 89/391/EEC which has been incorporated into French labor law. However, there are no preventive measures. Nonetheless, various countries are currently addressing research on voice disorders with a view to compiling preventive programs.

Preventive programs consist of two approaches: the indirect approach and the direct approach. The indirect approach stresses vocal hygiene and the understanding of the anatomy and physiology of the larynx and vocal cords. Vocal hygiene recommendations consist in a list of instructions on how to avoid vocal disorders. The literature stresses a list of `to do's and don'ts' for voice care. The don'ts are frequently stressed more. The explanation of the anatomy and physiology of the larynx and vocal cords must be comprehensible for a non-specialist public. In consequence, audiovisual programs are used to show the functioning of the vocal system in a readily understandable manner.

The direct approach stresses voice exercises but also insists on vocal hygiene. Voice exercises address various parameters: the position of the body, breathing (learning the various techniques), resonance, articulation, the voice (quality, timbre, intensity) and voice projection. In order to meet high requirements, the functions of the voice need to be increased and strengthened. Three techniques were developed in Antiquity: (1) the position of articulation, (2) the positioning of the voice in the mask and (3) the search for obstacles to strengthen the different parameters of voice and speech. Role playing exploiting the resources of the imagination are often used. It would seem that combining the direct and indirect approaches is the most optimal solution for obtaining the best possible results.

Various voice training programs have been designed to prevent voice disorders. These programs are beginning to be used in the training of teachers and other voice professionals such as radio broadcasters.

Recent preventive programs are based on the combination of the direct and indirect approach and incorporate a subdivision with the abbreviation TTTT: Test, Theory, Training and Transfer. The `tests' are indispensable in order to identify the characteristics of the voice. The `theory' may be considered the indirect approach. `Training' consists in voice exercises. `Transfer' aims to transfer what has been learned with regard to vocal hygiene and vocal techniques into everyday life. Group training has a positive effect on the function and the quality of the voice. After 18 months of training, a significant difference (p < 0.001) in voice quality was evidenced. The group that underwent training obtained a dysphonia severity index (DSI) of 4.3 while the control group had a DSI of 3.2. DSI is an acoustic index which reflects the quality of the voice: a score of −5 denotes a poor quality voice (G 3 = Grade 3; bad voice quality) and a score of +5 a good quality voice (G 0 = Grade 0: normal voice). DSI provides information on the time course of voice quality.

Time course of voice quality for a trained group and control group (taken from Timmermans et al, 2003).

A TTTT program, included in teacher training, contributes to structural prevention. At the start of the school year, the student teachers underwent a perceptual evaluation (`test' ) and, in the event of a voice disorder, the physician, speech therapist or phoniatrician was contacted. The `theory' was imparted after the test and before in-school teacher training. Voice `training' was conducted in groups in two phases. The first phase was given before teacher training and the other afterwards. The motivation for applying the voice techniques is stronger after teacher training when the student teachers have experienced the difficulties associated with adopting a `teacher's voice'. `Transfer' is necessary to consolidate the new vocal technique during or after the course.

For working teachers, the Training, Guide, Video (TGV) program includes training materials, a guide for the school administration, a guide with the schedule for voice assessments and a video explaining how the larynx functions, the diseases of the larynx and the recommendations for vocal hygiene. The program is implemented in the four TTTT stages.

Basic facilities are components of prevention: good classroom acoustics and the use of sound amplifiers. The latter are all the more necessary the greater the acoustic pollution: student behavior (talking increasingly loudly) and the style of teaching have changed. Moreover, according to the surveys, the number of students per class has increased.

It is thus important to prevent potential problems before they occur by daily action targeting vocal hygiene, awareness promotion and training.

• Recommendations

Numerous studies have shown that teachers frequently experience voice disorders, which are strongly related to the exercise of their profession, during their careers. This finding should lead to teachers being better informed and trained with respect to the potential for prevention and management of that type of disorder.

In the context of training, a teaching module on vocal physiology, voice diseases and trauma, and the effects of stress and anxiety could be an initial approach to the instrument that the voice constitutes. Similarly, training in correct use of the vocal apparatus in a teaching context (addressing a class, capturing attention without increasing vocal intensity, disarming aggressiveness, etc.) and an awareness of the rules for vocal hygiene (fluids intake, warming up, pauses in prolonged phonation) and the risk factors (smoking, caffeine, poor room acoustics, etc.) are fully in line with a preventive approach. Future teachers should also learn to understand their vocal capability and its limitations better. The ideal would be recourse to a speech therapist in order to obtain early personalized assistance. Although preventive programs exist, their implementation for teachers requires adaptation to the profession in France.

At work in their educational establishment, teachers should also have access to information on the acoustics of their classroom and should be consulted on potential arrangements that could enhance the quality of their working environment.

Although frequent in teachers, voice disorders nonetheless remain poorly understood and poorly managed by healthcare professionals. They are frequently responsible for sick leave. The treatment modalities still remain very heterogeneous and call for reflection on the harmonization of medical practices and the efficacy of rehabilitation speech therapy under the auspices of the French High Authority for Health (HAS).

The analytical and synthetic review of the scientific literature conducted by the expert group has enabled identification of a number of gaps in our epidemiological and clinical understanding of voice disorders. The review has also drawn attention to working hypotheses for improvement of the instruments used to evaluate the voice in occupational settings and for improvement of the acoustic environment.

Furthering research

Economic studies.

Very few studies of the cost generated by absenteeism and voice disorder treatment have been conducted. In the United States, the absenteeism related to voice disorders due to noise pollution of classes was estimated to cost USD 2.5 billion per year and 2 days of sick leave per teacher per year. The cost of a substitute teacher was USD 220 per day. Those costs are far from negligible.

The expert group recommends that studies should be conducted in order to estimate the efficacy and cost of medical and voice therapist management of voice disorders and conduct a cost-benefit analysis of preventive programs.

Epidemiological research on voice disorders in teachers

In France, only partial data on voice disorders in teachers are available.

In order to elucidate the situation in France, the expert group recommends initiating a multicenter study of teachers using the Voice Handicap Index or another scale with a view to collecting information on the number of cases of dysphonia with or without consultation (complaint), the number of sick leaves, their frequency, duration and reasons, and in order to monitor the time course of the prevalence of the problem with a view to enhancing prevention. The expert group also suggests conducting a prospective study over 5 years, then 10 years. The study would address teachers having received information on the voice during teacher training vs. those who did not in order to study the frequency of voice disorder occurrence in the exercise of the teaching profession. It is possible that correct information during initial teacher training would be sufficient to protect the teachers.

Research on voice amplification systems

Various studies of the use of portable individual amplifiers or the amplification system incorporated in the facility infrastructure have been published in the literature. The systems are reported to be highly effective, in particular with regard to reducing the vocal intensity of teaching and enhancing the signal/noise ratio. Moreover, a judiciously deployed loudspeaker system prevents the adverse effects of an intensity gradient (the back-row problem). The results of pilot studies conducted in the United States, unfortunately only over the short term, are reported to be extremely interesting with benefits for both the teachers and students. The concept of teacher comfort is clearly experienced by the interested parties.

The expert group recommends further research on voice amplifier prototypes. The main problems that have still to be solved seem to be of a technical nature (improvement in quality, need to prevent the teacher having to adjust the amplifier while teaching, siting optimization, maintenance of the wireless microphone, type and positioning of the loudspeakers, optimal adjustment of the signal/nose ratio, etc.). Once a prototype has been produced, it could be installed in a number of pilot classrooms.

Developing high performance techniques enabling functional assessment of the voice

Functional assessment and documentation of a patient's voice using a small number of numerical cues are widespread. Functional assessment is based on speech signals or other signals, which are acquired in a non-invasive manner. That is, their recording does not interfere with the speaker's performance. Methods of analysis that have been developed recently are more flexible and reliable than those available in the form of dedicated computer programs based on heuristic methods developed 20 years ago. Conventional clinical methods, for instance, are often exclusively designed for the analysis of sustained speech sounds.

It is conjectured that the quasi-permanent monitoring of the voices of speakers at risk in their occupational environment could contribute to both prevention and expert assessment of voice disorders. Currently, this is only a plausible hypothesis since the existing analyzers are not portable and the speech data they process are limited.

In consequence, the expert group recommends promoting research projects with the following objectives:

• development of prototype clinical workstations based on recent analytical methods which answer criticisms formulated with regard to existing methods. The aim is to let clinical practice benefit from recent advances in the laboratory;
• development of demonstrators of portable analyzers enabling monitoring of a speaker's speech in the context of his/her daily work and acquiring relevant data on vocal production. The aim is to test the conjecture that vocal dosimetry is an instrument for prevention and expert assessment of voice disorders in professional speakers.

Developing research on the individual characteristics and situational adaptability of the voice in class

The technical resources for voice analysis, in its various acoustic components, and the IT resources for data processing now enable specific research to be envisaged. The research would be designed to answer the question: Does voice adaptability differ over time as a function of the individual characteristics of the teachers? In order to answer that question, the time course of the prosodic phenomena of speech could be monitored over a fairly long-duration teaching task: acoustic parameters of the accent, intonation, rhythm (succession of durations and syllable stressing), elocution rate, tempo and speed, and pause distribution.

In order to ensure satisfactory recording of intonation, it would be appropriate to use a contact microphone (type: laryngeal microphone) which could be issued to volunteer teachers. A room microphone would also be necessary in order to promote the intelligibility of the words spoken.

An aspect of the question which is worthy of further investigation consists in the changes in the syntactic function of intonation in settings of prolonged use of the voice in front of a student public.

The expert group recommends studies to generate reference data on the use of the projected voice over a certain duration in the implementation of a teaching task.

Developing research on the interface between classroom acoustics and voice disorders

A gradual increase in the level of environmental noise to which a speaker is exposed induces a natural raising of the voice. The vocal effort of teaching is thus linked to the noise level in the classroom. The teacher wishes to use a volume that enables clear understanding of his/her message. Noisy and reverberating environments not only limit the understanding of the message but also limit the auditors' and speakers' endurances. In consequence, the probability of scholastic success is reduced in acoustically-unfavorable teaching environments and all the occupants of the classroom, teachers and students, suffer from the situation.

The expert group recommends studying the relationship between the acoustic conditions of classrooms and the emergence of voice disorders. All acoustic improvements targeting the teacher will have favorable repercussions on the students' receptiveness.

Created: 2006.

• Cite this Page INSERM Collective Expertise Centre. The voice and its disorders in teachers [Internet]. Paris (FR): Institut national de la santé et de la recherche médicale; 2006.
• PDF version of this title (310K)

In this Page

Other titles in this collection.

• INSERM Collective Expert Reports

Related information

• NLM Catalog Related NLM Catalog Entries

Similar articles in PubMed

• Review Dyslexia Dysorthography Dyscalculia: Review of the scientific data [ 2007] Review Dyslexia Dysorthography Dyscalculia: Review of the scientific data INSERM Collective Expertise Centre. 2007
• Review Cancers: Long-term prognoses [ 2006] Review Cancers: Long-term prognoses INSERM Collective Expertise Centre. 2006
• Review Glycol ethers: New toxicological data [ 2006] Review Glycol ethers: New toxicological data INSERM Collective Expertise Centre. 2006
• Review Alzheimer’s Disease: Scientific, medical and societal implications [ 2007] Review Alzheimer’s Disease: Scientific, medical and societal implications INSERM Collective Expertise Centre. 2007
• Review Obesity: Assessment and the evaluation of obesity prevention and management programs [ 2006] Review Obesity: Assessment and the evaluation of obesity prevention and management programs INSERM Collective Expertise Centre. 2006

Recent Activity

• The voice and its disorders in teachers The voice and its disorders in teachers

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

Connect with NLM

National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894

Web Policies FOIA HHS Vulnerability Disclosure

Help Accessibility Careers

The Federal Register

The daily journal of the united states government, request access.

Due to aggressive automated scraping of FederalRegister.gov and eCFR.gov, programmatic access to these sites is limited to access to our extensive developer APIs.

If you are human user receiving this message, we can add your IP address to a set of IPs that can access FederalRegister.gov & eCFR.gov; complete the CAPTCHA (bot test) below and click "Request Access". This process will be necessary for each IP address you wish to access the site from, requests are valid for approximately one quarter (three months) after which the process may need to be repeated.

An official website of the United States government.

If you want to request a wider IP range, first request access for your current IP, and then use the "Site Feedback" button found in the lower left-hand side to make the request.