The applicant, Mr. Geoffrey Kahn, proposes to identify key, modifiable predictors of attempted suicide among children and youth who have had contact with Child Protective Services (CPS), and quantify the impact of hypothetical preventive interventions that could be implemented for these children. This research will be the foundation of the applicant?s doctoral dissertation, and will serve as one component of a comprehensive training plan to prepare the applicant for a career in independent research. Mr. Kahn will be mentored by a team of co-sponsors and consultants with expertise in suicide prevention, advanced statistical methods, child welfare research, and the CPS system. He will complete coursework in analysis of complex survey data and machine learning methods, participate in working groups on suicide prevention, machine learning, and causal inference at the sponsoring institution, Johns Hopkins Bloomberg School of Public Health, and present his findings to diverse audiences. The advanced training and focused mentoring that Mr. Kahn will receive through this Fellowship will position him to conduct future, independent research studies on the prevention of suicide, particularly suicide among youth. Suicide is the second leading cause of death among youth aged 15-24 years in the US. Suicide rates have been increasing for nearly two decades, with that increase accelerating in recent years. Children who have had contact with Child Protective Services are at increased risk for numerous mental health problems, including suicidal ideation and attempts. The opioid epidemic is thought to be contributing to an increase in the already substantial number of children entering the CPS system. Finally, suicide has been remarkably difficult to predict, with recent meta-analyses showing that most currently identified risk factors do not predict future suicide attempts or deaths significantly better than chance, and predictive accuracy has not improved over the last 50 years of research. Suicide is likely caused by interaction among numerous risk factors, but most studies have examined only single risk factors. New machine learning (ML) methods have been shown to out-perform traditional regression in predictive models, and there are classes of ML algorithms that can identify complex interactions among large numbers of variables. The applicant proposes a multi-part analysis of suicide risk factors among youth with CPS contact. The proposed project will: 1) assess the role of CPS case outcome on access to specialized mental health care and subsequent suicide attempt; 2) use machine learning methods to develop and validate a predictive model for suicide attempts; and 3) apply causal inference methods to quantify the impact of hypothetical interventions on the key risk/protective factors identified in the study. The proposed project is consistent with NIMH?s strategic priorities, as it will elucidate complex groups of risk factors which contribute to suicidal behavior and identify the most promising targets for intervention.
The proposed project will identify key predictors of suicide attempts among children and youth who have had contact with Child Protective Services (CPS). Machine learning algorithms will be used to evaluate large numbers of variables and find clusters of factors, rather than single risk factors, that predict suicide attempts in a longitudinal cohort of youth. Finally, causal inference methods will be used to quantify the potential impact of intervening on key risk factors which have been identified AND have been determined to be feasible targets for preventive intervention after consulting with experts from the CPS system.
