A Person-Centered Approach to Identifying the Characteristics of Young Children Who Self-Harm and Die by Suicide

Over the past decade, rates of suicide among children ages 5-11 years have nearly doubled, but our ability to identify and intervene with children most at risk for suicide remains extremely limited. Much of our understanding of risk factors for child suicide has been learned from adolescents or samples including both adolescents and children. Children who contemplate or attempt suicide are at risk for suicidal thoughts and behavior across the lifespan. Children differ from adults and adolescents, thus, this dearth of research focused on child suicide is highly problematic for suicide prevention. Certain combinations of psychiatric and social factors may help to explain suicidal behavior in youth and adults. However, no study has examined how combinations of risks, or clinical profiles, differ between children with and without self-harm, and whether certain clinical profiles are at greater risk for death by suicide. The proposed study will advance our ability to distinguish children who self-harm from other high-risk children, and to identify those at greatest risk for suicide.

There are three primary aims: 1) Identify clinical profiles preceding self-harm that distinguish children with and without a history of self-harm in a high-risk sample, 2) identify clinical profiles predictive of future acts of self-harm and suicide, 3A) examine sex-specific risk factors, and 3B) examine racial and ethnic disparities within sex. As an exploratory aim, we will compare clinical profiles in children with and without self-harm to the profiles of approximately 249 child suicide decedents. Based on prior research we hypothesize children who self-harm will have distinct profiles from children without self-harm, and sex-specific patterns of psychopathology will further differ within children who self-harm depending on future self-harm, age, race, and adversities. 

To achieve these aims, we will use a retrospective population-based case-control design. Our sample will include 4,280 Medicaid-enrolled children whose first medically documented self-harm occurred between ages 5-11 years and 42,800 matched children without self-harm as controls. We will include measures of psychopathology, intellectual and developmental disorders, mental health services used, maltreatment, and residence in adoptive or foster care. We also include community-level factors from the National Center for Educational Statistics including living in a Title 1 eligible school’s zip code and urbanicity or rurality of that zip code. The proposed study uses latent-class analysis (LCA) with a distal outcome using bias-corrected methods to address aims 1-3. Posterior probability estimates will be used to identify suicides. We will use LCA to address our exploratory aim.

This study has strong potential to identify clinical profiles of children who self-harm and those at greatest risk for suicide in childhood which could have significant implications for prevention and interventions. Important next steps following this research will be to create developmentally sensitive interventions and screenings tailored to the clinical profiles of children at greatest risk for suicide. Additional next steps might also include identifying greater resources for children exhibiting clinical profiles at risk for repeated acts of future self-harm to prevent suicide after treatment for self-harm, a known high-risk period for suicide.