The utility of artificial intelligence in suicide risk prediction and the management of suicidal behaviors

Abstract

Objective: Suicide is a growing public health concern with a global prevalence of approximately 800,000 deaths per year. The current process of evaluating suicide risk is highly subjective, which can limit the efficacy and accuracy of prediction efforts. Consequently, suicide detection strategies are shifting toward artificial intelligence platforms that can identify patterns within ‘big data’ to generate risk algorithms that can determine the effects of risk (and protective) factors on suicide outcomes, predict suicide outbreaks and identify at-risk individuals or populations. In this review, we summarize the role of artificial intelligence in optimizing suicide risk prediction and behavior management. Methods: This paper provides a general review of the literature. A literature search was conducted in OVID Medline, EMBASE and PsycINFO databases with coverage from January 1990 to June 2019. Results were restricted to peer-reviewed, English-language articles. Conference and dissertation proceedings, case reports, protocol papers and opinion pieces were excluded. Reference lists were also examined for additional articles of relevance. Results: At the individual level, prediction analytics help to identify individuals in crisis to intervene with emotional support, crisis and psychoeducational resources, and alerts for emergency assistance. At the population level, algorithms can identify at-risk groups or suicide hotspots, which help inform resource mobilization, policy reform and advocacy efforts. Artificial intelligence has also been used to support the clinical management of suicide across diagnostics and evaluation, medication management and behavioral therapy delivery. There could be several advantages of incorporating artificial intelligence into suicide care, which includes a time- and resource-effective alternative to clinician-based strategies, adaptability to various settings and demographics, and suitability for use in remote locations with limited access to mental healthcare supports. Conclusion: Based on the observed benefits to date, artificial intelligence has a demonstrated utility within suicide prediction and clinical management efforts and will continue to advance mental healthcare forward.

Suicide on Facebook-the tales of unnoticed departure in Bangladesh

“…Police helped people who were at risk of suicide after receiving an alert call from Facebook (Singer, 2018). The use of artificial intelligence and machine learning has received increased attention in recent days for suicide prediction (Fonseka et al, 2019). In addition to social media data, the digital assistant product makers such as Apple (Siri), Google Assistant and Amazon (Echo/Alexa) are augmenting the suicide prevention services by analysing to an indication of suicidal thoughts and behaviours (De et al ., 2018).…”

Digital Mental Health Challenges and the Horizon Ahead for Solutions

“…The ability of AI to enhance risk prediction in mental health assessment may extend from mining unstructured data (eg, through text or continuous sensor monitoring), such as a study on sepsis, which revealed improved algorithm accuracy (deidentified data made available for review) [55]. The implementation of algorithm prediction systems in predicting the risk of suicide is currently limited to experimental and feasibility studies (with utility) without efficacy evaluation [16,[56][57][58].…”

An Autoencoder and Machine Learning Model to Predict Suicidal Ideation with Brain Structural Imaging

“…Figure 1. The structure of the autoencoder, which consists of six 3D convolution layers, three max-pooling layers, and three upsampling layers, resulted in compressed images with sizes of (11,13,11,16) for feature extraction.…”

“…Traditional statistics can only differentiate different groups with suicidal ideation from those without, rather than detect which individual is at risk. New analytical methods, such as machine learning, have been used in an attempt to develop algorithms to classify individual risk [8][9][10][11]. One study used machine learning algorithms based on functional magnetic resonance imaging (fMRI) neural signatures of death-and life-related concepts to detect individuals with suicidal ideation with 91% accuracy [12].…”