TOPLINE:

Clinical factors such as previous self-harm and psychiatric comorbidity were more strongly associated with the risk for self-harm than demographic factors in individuals with substance use disorders (SUDs), a new risk model called Oxford Self-hArM after substance use disorders (OxSAMS) demonstrated.

METHODOLOGY:

• Researchers identified a population-based cohort of 449,720 individuals with SUDs in Sweden between 2006 and 2020, of whom 361,120 were allocated to the development sample and 88,600 were allocated to external validation on the basis of geographical regions.
• Participants were followed up from the date of their initial secondary healthcare contact for SUD until the first occurrence of self-harm, death, emigration, or the end of follow-up on December 31, 2020, whichever occurred first; self-harm was assessed over five predetermined periods: within 7 days and 1, 3, 6, and 12 months.
• Statistical models analysed associations of sociodemographic, clinical, treatment, and self-harm history factors with self-harm outcomes, while accounting for the time to event.
• Risk factors tested included age, sex, unemployment, income, comorbidity of mental disorders, antidepressant and antipsychotic treatment in the 6 months before the index SUD episode, previous self-harm behaviours, self-harm method involving cutting, and arrest within 10 years of the index SUD diagnosis.
• The performance of the OxSAMS tool was evaluated using Harrell's C-index and the area under the curve (AUC), and calibration measures included expected-to-observed (E/O) risk ratios and Brier scores, with external validation conducted in an independent geographical subset of the cohort.

TAKEAWAY:

• Previous self-harm behaviours showed the strongest association with subsequent self-harm (adjusted hazard ratio [aHR], 3.17; 95% CI, 3.08-3.26), followed by comorbidity of mental disorders (aHR, 2.63; 95% CI, 2.50-2.72).
• Recent antipsychotic treatment was associated with an increased risk for self-harm (aHR, 1.34; 95% CI, 1.24-1.44), as was antidepressant treatment (aHR, 1.29; 95% CI, 1.23-1.36), even after adjusting for psychiatric comorbidity, likely reflecting greater clinical severity and complexity.
• Arrest within 10 years of the index SUD diagnosis was associated with an elevated risk for self-harm (aHR, 1.30; 95% CI, 1.27-1.34); however, demographic factors such as younger age (aHR, 1.19), female sex (aHR, 1.22), and low income (aHR, 1.05) showed more modest associations, with narrow 95% CIs for all.
• The validated risk model demonstrated good discrimination, with an overall C-index of 0.76 and AUCs ranging from 0.73 to 0.79 across different time periods; good calibration was indicated by E/O risk ratios of 1.00-1.04 and Brier scores of 0.01-0.04.

IN PRACTICE:

"We developed a brief, accurate risk model for assessing self-harm among individuals with substance use disorder using routinely collected data. The OxSAMS tool can support decision-making in substance use and mental health services and potentially contribute to the prevention of self-harm in this high-risk population," the authors wrote.

SOURCE:

This study was led by Rongqin Yu, University of Oxford, Oxford, England. It was published online on May 07, 2026, in Addiction.

LIMITATIONS:

The study was limited by the lack of data on factors such as childhood maltreatment and the quality of family relationships in Swedish registries. The model relied on routinely recorded predictors rather than time-varying information, and further validation is needed for demographics outside Nordic countries, particularly in countries that differ from Nordic countries in terms of service provision and population demographics, including ethnic diversity, socioeconomic distribution, and age structure.

DISCLOSURES:

This study was funded by the John Fell Fund, University of Oxford; National Institute for Health and Care Research Oxford Health Biomedical Research Centre; Wellcome Trust; Swedish Research Council; and Marcus and Amalia Wallenberg Foundation. The authors declared having no relevant conflicts of interest. 

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.