Alcohol use disorders may be a notable psychiatric predictor of subsequent major adverse cardiac and cerebrovascular events among both men and women following trauma exposure.
Investigators conducted a matched case-control study nested within a population-based cohort of adults with documented trauma exposure in Denmark from 1995 to 2018. Using nationwide registry data, they identified 43,994 patients who subsequently experienced a major adverse cardiac and cerebrovascular event (MACCE) and 175,872 controls matched by MACCE date, sex, and year and age of trauma. Those with a history of a MACCE at the time of trauma were excluded.
The primary outcome was incident MACCE, a composite of nonfatal myocardial infarction, nonfatal stroke, coronary revascularization, or cardiovascular mortality. Psychiatric predictors were diagnoses recorded within 3 years posttrauma and prior to MACCE. The investigators developed sex-stratified extreme gradient boosting models and used Shapley Additive Explanations (SHAP) values to rank the relative importance of psychiatric predictors. The models included 91,111 men and 84,781 women in the training sets and 22,777 men and 21,197 women in the testing sets.
Alcohol use disorders ranked first among the psychiatric predictors among both sexes. The mean SHAP value among men was about 10 times that of depressive episodes, the next-ranked predictor. Alcohol use disorders were recorded in 9% of male MACCE cases compared with 5% of male controls. Among women, alcohol use disorders were recorded in 4% of cases vs. 2% of controls, while depressive episodes occurred in 6% of cases vs. 4% of controls.
In the hold-out testing samples, the association persisted after adjustment for age, income, employment status, and marital status at the time of trauma. Men with alcohol use disorders had 1.64 times the odds of MACCE, and women had 1.85 times the odds compared with those without alcohol use disorders.
In both sexes, depressive episodes, recurrent depressive disorders, posttraumatic stress disorder (PTSD) and other stress-related disorders, non-substance–induced delirium, schizophrenia, anxiety disorders, personality disorders, and bipolar disorders ranked high. Brain-related or physiological psychiatric conditions and several substance use disorders also appeared among the leading predictors.
The relative rankings differed by sex. Alcohol use disorders stood out more among men, whereas depressive episodes had greater relative importance among women. Cannabis- and opioid-related substance use disorders appeared among the leading predictors in men. Persistent mood disorders, persistent delusional disorders, and sedative-related substance use disorders appeared among the leading predictors in women. The investigators cautioned that these differences represented relative rankings within each sex-specific model and should not be interpreted as direct comparisons of effect sizes between men and women.
The findings did not causality. The investigators did not include lifestyle behaviors or clinical cardiovascular risk factors in the primary models because the analysis was designed to characterize the relative importance of psychiatric diagnoses. Registry data may not have captured less severe trauma exposures, psychiatric diagnoses made only in primary care, or subthreshold psychiatric symptoms, and about 25% of candidate psychiatric disorders were excluded because of low prevalence. The composite outcome prevented assessment of whether associations differed between coronary and cerebrovascular events. Potential registry misclassification and uncertain generalizability beyond Denmark were additional limitations. The models had modest discrimination and were intended to generate hypotheses rather than serve as clinical cardiovascular risk-prediction tools.
“Our machine learning approach identified alcohol use disorders, which have traditionally been understudied in the trauma–[cardiovascular disease] literature, as the most important psychiatric predictor of MACCE for both trauma-exposed men and women,” wrote lead study author Jennifer A. Sumner, PhD, of the Department of Psychology at the University of California, Los Angeles, and colleagues.
Full disclosures of the study authors can be found in the study.