Accepted for/Published in: JMIR Mental Health
Date Submitted: May 24, 2022
Date Accepted: Jul 22, 2022
Using Voice Biomarkers to Classify Suicide Risk in Adult Telehealth Callers: a Retrospective Observational Study
ABSTRACT
Background:
Artificial Intelligence has the potential to innovate current practices used to detect imminent risk of suicide, and to address shortcomings in traditional assessment methods.
Objective:
We sought to automatically classify short segments (40ms) of speech according to low versus imminent risk of suicide in a large number (n = 281) of telephone calls made to two telehealth counselling services in Australia.
Methods:
This retrospective repeated measures cohort design included 281 helpline calls sourced from On The Line, Australia (n=266) and 000 Emergency services, Canberra (n=15). Imminent risk of suicide was coded for when callers affirmed intent, plan and the availability of means; was assessed as such by the responding counsellor, and re-assessed by a team of clinical researchers using the Columbia Suicide Severity Rating scale (C-SSRS = 5/6). Low risk of suicide was coded for in an absence of intent, plan and means and via C-SSRS = 1/2. Pre-processing involved normalisation and pre-emphasis of voice signals, while voice biometrics were extracted using the statistical language r. Candidate predictors were identified using Lasso Regression. The relationship of each candidate biomarker was assessed as a predictor of suicide risk using a Generalized Additive Mixed effects regression Model (GAMM) using splines to account for non-linearity. Finally, a Component-Wise Gradient Boosting model was trained on 60% of the data, then validated and tested on the remaining data (20%/20%).
Results:
Seventy-seven imminent risk calls were compared with 204 low risk calls. 171,968 40ms frames featuring 36 voice biomarkers were extracted from these calls. Candidate biomarkers were reduced to 11 primary markers, with distinct models needed for men and women. Classification accuracies greater than 99.9% were obtained for imminent suicide risk for each of the data partitions using these primary markers.
Conclusions:
This study demonstrates an objective, efficient and economical assessment of imminent suicide risk in an ecologically valid setting with potential applications to real-time assessment and response. Clinical Trial: ACTRN12622000486729
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.