Examining the intersection of mental illness and suicidal risk in the shadow of a pandemic: A Machine Learning Approach

Abstract

To develop the suicidal recovery model for adults with mental illness during the pandemic and better serve them in the mental health system, it is necessary to ensure that we can identify the intersection of mental illness and suicidal risk. Therefore, we used machine learning to examine the intersection of mental illness and suicide aged 17 years old and above adults in the Mideastern state-funded mental health service (n=29,267) during the calendar years of 2019 and 2020. Classification, regression tree analyses, and chi-square automatic interaction detection (CHAID) were used to identify the intersection of mental illness and suicidal risk and determine their classification accuracy. In the COVID-19 pandemic year, self-injurious behavior, depression, adjustment to trauma, danger to others, impulse control, anger control, age, sleep, and psychosis were identified as the critical factors associated with suicidal risk. However, danger to others, impulse control, anger control, and age were associated with suicide risk only in 2020, but not in 2019. Overall, self-injurious behavior, depression, danger to others, psychosis, adjustment to trauma, anxiety, sleep, and interpersonal were intersected with suicidal risk.