Characteristics of Cardiovascular Disease Prediction Models Considering Mental Disorders: A Systematic Review.

backgroundPrognostic models for cardiovascular disease (CVD) risk have commonly included predictors such as cholesterol levels. Mental disorders are robust predictors of CVD and associated mortality, leading to approximately 2-fold increases in risk. This systematic review aimed to narratively summarize the key characteristics, strengths, and limitations of all CVD prediction models that consider mental disorders.

methodsA literature search with medical subject headings/key terms related to CVD, mental disorders and prognostic modeling was conducted in Medline and EMBASE. Included studies were: cohort studies of CVD prediction model development, validation, or recalibration that included mental disorders as prognostic factors/covariate(s), or the population of interest. All studies were screened by 2 independent reviewers, followed by data extraction. The Prediction Model Risk Of Bias Assessment Tool was used to critically appraise bias. A narrative synthesis was used to summarize mental disorder and sociodemographic factor/intersectionality inclusion.

resultsThere were 31 unique models identified (n=35 records including external validations). Considering these, 77% included mental disorders as a covariate, whereby depression and/or anxiety were the most considered (71%, n=22 studies). Most models were published within the last 5 years, included measures of socioeconomic status; however, many models lacked intersectionality considerations. Only one study was identified with a low risk of bias, while the majority had analytic concerns.

conclusionsDepression and/or anxiety were the most commonly considered mental disorders in modeling, despite larger associations with CVD for other disorders. Further CVD prediction modeling should consider a broader array of mental disorders, and limit methodological biases.