Predictors of suicidality: An exploratory study using machine learning paradigm

Pages: 1055-1062
Pragya Malik, Sukanya Ray, and Rajendra Sharma (Department of Clinical Psychology, AIBAS, Amity University Madhya Pradesh)

A growing number of suicides calls for predictability of such attempts in order to prevent suicides. A significant number of suicidal attempts are found to be non-linked to any psychiatric syndrome. Amongst the known risk factors, stressful life events, bereavement, childhood abuse have been found to be significant. Spiritual well-being may act as a protective factor. Suicide rates are found to differ in males and females. The present study was aimed at exploring these factors in males and females as predictors of suicidality using advanced Data Science methods. 14 individuals with at least one suicidal attempt referred from a Psychiatry OPD in Gwalior, Madhya Pradesh. Beck Scale for Suicidal Ideations, Presumptive Stressful Life Events Scale, The Grief Evaluation Measure, Childhood Trauma Questionnaire, Spiritual Well-being Scale, Beck Depression Inventory were used for measuring the predictor variables (both quantitative and qualitative). Data Science methods and Machine Learning paradigm using programming (Python 3.6, Version: Anaconda) languages and Decision Trees, Extra Trees, Logistic Regression were used to analyze the data. There seems to be significant discrimination in the way the predictor variables interact to result in suicidality in males and females. Beck Depression Inventory scores, Childhood Trauma Questionnaire scores were found to be relevant predictors explaining the discrimination seen between males and females. These preliminary findings suggest the need to model determinants and vulnerability to suicidal behavior, in prevention of suicides.