Toward Linguistic Recognition of Generalized Anxiety Disorder

Abstract

BackgroundGeneralized anxiety disorder (GAD) refers to extreme, uncontrollable, and persistent worry and anxiety. The disorder is known to affect the social functioning and well-being of millions of people, but despite its prevalence and burden to society, it has proven difficult to identify unique behavioral markers. Interestingly, the worrying behavior observed in GAD is argued to stem from a verbal linguistic process. Therefore, the aim of the present study was to investigate if GAD can be predicted from the language people use to put their anxious worries into words. Given the importance of avoidance sensitivity (a higher likelihood to respond anxiously to novel or unexpected triggers) in GAD, this study also explored if prediction accuracy increases when individual differences in behavioral avoidance and approach sensitivity are taken into account.MethodAn expressive writing exercise was used to explore whether GAD can be predicted from linguistic characteristics of written narratives. Specifically, 144 undergraduate student participants were asked to recall an anxious experience during their university life, and describe this experience in written form. Clinically validated behavioral measures for GAD and self-reported sensitivity in behavioral avoidance/inhibition (BIS) and behavioral approach (BAS), were collected. A set of classification experiments was performed to evaluate GAD predictability based on linguistic features, BIS/BAS scores, and a concatenation of the two.ResultsThe classification results show that GAD can, indeed, be successfully predicted from anxiety-focused written narratives. Prediction accuracy increased when differences in BIS and BAS were included, which suggests that, under those conditions, negatively valenced emotion words and words relating to social processes could be sufficient for recognition of GAD.ConclusionsUndergraduate students with a high GAD score can be identified based on their written recollection of an anxious experience during university life. This insight is an important first step toward development of text-based digital health applications and technologies aimed at remote screening for GAD. Future work should investigate the extent to which these results uniquely apply to university campus populations or generalize to other demographics.

Screening for Generalized Anxiety Disorder From Acoustic and Linguistic Features of Impromptu Speech: Prediction Model Evaluation Study

“…Comparing our results with a prior study that aimed to predict GAD, there were studies that achieved the above-random prediction for GAD [ 21 , 23 ], and there were studies that did not [ 28 ]. Our models performed significantly better than the random model, and we speculated that this might be attributed to the larger and demographically broad sample size that enabled our model to learn a large amount of information in predicting GAD.…”

“…We also aimed to obtain broader demographics than those in the prior study. Most prior studies focused on a certain type of demographics, such as the study by McGinnis et al [ 13 ], which focused on children, and the studies by Week et al [ 17 ], Salekin et al [ 20 ], and Rook et al [ 23 ], which focused on undergraduate students. Both these types significantly limited the age range of the participants.…”

“…Rook et al [ 23 ] hypothesized that the worrying behavior in GAD comes from the verbal linguistic process. They attempted to predict GAD using only linguistic patterns.…”

“…In addition, the number of participants might not be the only factor affecting generalizability. Apart from the studies by Di Matteo et al [ 28 ] and Baird et al [ 21 ], the prior studies were mostly limited to very specific demographics: McGinnis et al [ 13 ] focused on children; Weeks et al [ 17 ], Salekin et al [ 20 ], and Rook et al [ 23 ] focused on undergraduate students at a university or college.…”

Predicting Generalized Anxiety Disorder From Impromptu Speech Transcripts Using Context-Aware Transformer-Based Neural Networks: Model Evaluation Study

“…Rook et al [ 19 ] attempted to predict GAD from linguistic patterns because they believed that the worrying behavior in GAD comes from the verbal linguistic process. A total of 142 undergraduate participants (n=56, 39.4% men and n=86, 60.6% women) were recruited for their study and were asked to recall and write about an anxious experience during their university life.…”