Psychiatric factors predict type 2 diabetes mellitus in US Veterans.

Co-occurrence of type 2 diabetes mellitus (T2D) and serious mental illnesses (SMI) is prevalent yet underappreciated, and significantly contributes to increased morbidity and reduced lifespan. There is, therefore, a need to identify T2D risk factors to inform preventative approaches to the care of SMI-diagnosed patients. Our objective was to use predictive modeling methods to capture risk factors for T2D in a sample of 618,203 Veterans using data obtained from hospital electronic health records (EHR). This case-control study assessed VISN4 Veterans with and without T2D diagnoses and SMI diagnoses (schizophrenia, SZ; schizoaffective, SZA; bipolar disorder, BD; major depression, MDD; 2009-2019). Demographic variables and medications were obtained from the EHR. Following rigorous data quality control, 543,979 Veterans qualified for analysis (Age = 65.9[17.6]years; body mass index(BMI) = 28.6[6.0]kg/m; N = 157,457[29%]; and N = 506,257[93.1%]). Veterans with co-occurring SMI + T2D included N = 2,087(36.5%), N = 1,345(36.3%), N = 10,540(29.2%), and N = 20,510(30%) compared to 112,973(28.6%) non-SMI controls (NSC) with T2D. Factors that predicted T2D (R = 34%) included age, sex, BMI, race/ethnicity, psychiatric diagnoses, and commonly prescribed psychiatric medications. Significant interactions were found between age (centered) and BMI on the odds of T2D (P < 0.001), as well as interaction between sex and BMI (P < 0.001), after adjusting for confounders. Veterans with SMI (SZ, MDD, SZA, and BD) had a higher likelihood of experiencing T2D, compared to the NSCs (OR = 1.30, 95% CI = 1.21-1.40; OR = 1.07, 95% CI = 1.05-1.10; OR = 1.26, 95% CI = 1.16-1.38; OR = 1.05, 95% CI = 1.01-1.08). Finally, Veterans exposed to both selective serotonin reuptake inhibitor (SSRI) antidepressants and mood stabilizers had a 2.11 times increase in the odds of having T2D (95% CI = 2.06-2.16; P < 0.001) compared to Veterans not taking either medication. Four major psychiatric disorders (SZ, SZA, MDD, and BD) and several classes of medications used to treat them increased T2D risk. Our findings suggest that the measures assayed offer a potentially useful signal, that along with clinical, anthropometric, and biochemical measures can be used to ascertain metabolic risk. If confirmed with an independent sample, these findings could also inform medication choices made by prescribers.