JMIR Publications

ABSTRACT

Background:

In an age where telehealth services are increasingly being used for forward triage, the need for accurate suicide risk detection is increasing. Voice signals analysed using Artificial Intelligence is now proving capable of detecting suicide risk at accuracies superior to traditional interview based approaches, suggesting an efficient and economical approach to ensuring ongoing patient safety.

Objective:

This systematic review aimed to identify voice signal characteristics that discriminate between patients experiencing elevated risk of suicide and comparison cohorts and to identify the specific technical specifications of the systems used in classification.

Methods:

A search of Medline via Ovid, Scopus, Computers and Applied Science Complete, CADTH, Web of Science, Proquest, Dissertations and Theses A&I, Australian Policy Online and Mednar was conducted between 1995 - 2021. A total of 1074 articles were assessed for relevance with 21 included in the final review.

Results:

Candidate voice signal characteristics discriminating between suicidal and comparison cohorts included speech timing patterns (median accuracy = 95%), power spectral density subbands (median accuracy = 90.3%) and mel-frequency cepstral coefficients (median accuracy = 80%). A random effects meta-analysis was used to compared 22 characteristics nested within three studies, which demonstrated significant standardised mean differences for frequencies within the first and second formants (standardised mean difference ranged between -1.07 and -2.56) and jitter values (standardised mean difference = 1.47).

Conclusions:

Although a number of key methodological issues prevailed amongst the studies reviewed, there is significant promise in the use of voice signal characteristics to detect elevations in suicide risk. Clinical Trial: International Prospective Register of Systematic Reviews (PROSPERO) on 28th April 2020 (registration number CRD420200167413)