Modeling Suicidality with Multimodal Impulsivity Characterization in Participants with Mental Health Disorder (Preprint)

Abstract

Suicide is one of the leading causes of death across different age groups. Impulsivity, the tendency to act on urges with low temporal latency and little forethought to the consequences of one's actions, is implicated in several clinical conditions including suicidality. Impulsivity can make a difference in the persistence of suicidal ideation and the progression of suicidal ideations to action. Currently, there are no validated markers for intent to act in suicidality and no objective measures that allow the prediction of suicidality possibly leading to suicide attempts. Impulsivity quantification could be the missing link in risk assessments of ideation-to-action suicidality frameworks.

We aimed to assess the impulsivity differences in participant groups with different suicidality and develop a suicidality prediction model using multimodal impulsivity measures.

We obtained questionnaires, behavioral tests, heart rate variability (HRV), and resting state fMRI (rsfMRI) measurements from a cohort of 34 well-characterized participants with mood disorders. The participants were categorized into three groups based on the suicidality stratification from their Mini International Neuropsychiatric Interview (MINI): none, low, and moderate-severe. Group differences in impulsivity and a suicidality prediction classifier using different impulsivity measures were investigated.

Questionnaire-based impulsivity measures, particularly the motor and attentional impulsivity, were significantly different between the suicidality groups with higher subscales of impulsivity associated with higher suicidality. HRV-based impulsivity measures were also significantly different between the groups. A multimodal system to characterize impulsivity outperformed the unimodal system in the three-class suicidality group prediction task, resulting in a classification accuracy of 96.77% when using all objective impulsivity measures.

Our work is the first reported attempt to elucidate the relative sensitivity of different impulsivity measures in relation to suicidality and employ multimodal impulsivity markers to predict suicidality. Impulsivity quantification could differentiate participants with suicidality and predict suicidality with high accuracy using a multimodal characterization. Suicidality prediction could potentially help tailor mental health interventions for an at-risk individual while the impulsivity characterization for suicidality prediction could be additionally helpful to deter progression from suicidal thoughts to acts of self-harm.