Self-Harm Events and Suicide Deaths Among Autistic Individuals in Ontario, Canada

Abstract

Importance: Reasons for elevated suicide risks among autistic people are unclear, with insufficient population-based research on sex-specific patterns to inform tailored prevention and intervention.

Objectives: To examine sex-stratified rates of self-harm events and suicide death among autistic individuals compared with nonautistic individuals, as well as the associated sociodemographic and clinical risk factors.

Design, setting, and participants: This population-based matched-cohort study using linked health administrative databases in Ontario, Canada included all individuals with physician-recorded autism diagnoses from April 1, 1988, to March 31, 2018, each matched on age and sex to 4 nonautistic individuals from the general population. Self-harm events resulting in emergency health care from April 1, 2005, to December 31, 2020, were examined for one cohort, and death by suicide and other causes from April 1, 1993, to December 31, 2018, were examined for another cohort. Statistical analyses were conducted between October 2021 and June 2023.

Exposure: Physician-recorded autism diagnoses from 1988 to 2018 from health administrative databases.

Main outcomes and measures: Autistic and nonautistic individuals who were sex stratified a priori were compared using Andersen-Gill recurrent event models on self-harm events, and cause-specific competing risk models on death by suicide or other causes. Neighborhood-level income and rurality indices, and individual-level broad diagnostic categories of intellectual disabilities, mood and anxiety disorders, schizophrenia spectrum disorders, substance use disorders, and personality disorders were covariates.

Results: For self-harm events (cohort, 379 630 individuals; median age at maximum follow-up, 20 years [IQR, 15-28 years]; median age of first autism diagnosis claim for autistic individuals, 9 years [IQR, 4-15 years]; 19 800 autistic females, 56 126 autistic males 79 200 nonautistic females, and 224 504 nonautistic males), among both sexes, autism diagnoses had independent associations with self-harm events (females: relative rate, 1.83; 95% CI, 1.61-2.08; males: relative rate, 1.47; 95% CI, 1.28-1.69) after accounting for income, rurality, intellectual disabilities, and psychiatric diagnoses. For suicide death (cohort, 334 690 individuals; median age at maximum follow-up, 19 years [IQR, 14-27 years]; median age of first autism diagnosis claim for autistic individuals, 10 years [IQR, 5-16 years]; 17 982 autistic females, 48 956 autistic males, 71 928 nonautistic females, 195 824 nonautistic males), there was a significantly higher crude hazard ratio among autistic females (1.98; 95% CI, 1.11-3.56) and a nonsignificantly higher crude hazard ratio among autistic males (1.34; 95% CI, 0.99-1.82); the increased risks were associated with psychiatric diagnoses.

Conclusions and relevance: This cohort study suggests that autistic individuals experienced increased risks of self-harm events and suicide death. Psychiatric diagnoses were significantly associated with the increased risks among both sexes, especially for suicide death, and in partially sex-unique ways. Autism-tailored and autism-informed clinical and social support to reduce suicide risks should consider multifactorial mechanisms, with a particular focus on the prevention and timely treatment of psychiatric illnesses.

Figure.. Cumulative Mean Function Curves for Self-Harm Events by Group

Estimated mean cumulative numbers of self-harm events by attained age are illustrated separately for autistic females, autistic males, nonautistic females, and nonautistic males. Shaded areas indicate 95% CIs of the estimated means. Numbers of individuals at risk in each group at intervals along the x-axis were calculated to indicate those who were not censored at the attained age.