Predictors of Body Mass Index in Individuals with Serious Mental Illness Receiving Integrated Health Services

Background and purpose: The physical health of persons with serious mental illness (SMI) has been a major focus of national efforts that seek to reduce health and health care disparities by providing integrated primary and behavioral health care (PBHC) services to this population. Persons with SMI are disproportionately affected by obesity and associated cardiometabolic conditions, which result in increased mortality and decreased longevity. Researchers have attributed this elevated risk to certain psychotropic medications that are indicated for weight gain; however, few studies have examined the health characteristics associated with body mass index (BMI) among clients with SMI. The current study sought to address this gap.

Research Questions: What combination of relevant biopsychosocial factors best predict BMI scores among individuals receiving community-based PBHC services?

Methods: Researchers collected sociodemographic (age, race, gender, education level), psychosocial (social support, psychological and daily functioning), health (objective biomarkers, self-assessed health, psychotropic medications) and health-risk (tobacco, alcohol, and substance use) data from the records of persons (N=184) enrolled in an integrated PBHC program within a community mental health center.  Social support was measured with the 4-item Perception of Social Connectedness subscale of the Mental Health Statistics Improvement Program (MHSIP).  Psychological functioning was assessed with the K6, and daily functioning was measured with the 8-item Perception of Functioning subscale of the MHSIP.  Health indicator data were used to measure systolic and diastolic blood pressure (BP), BMI, blood glucose, high-density-lipoproteins (HDL), low-density-lipoproteins (LDL), triglycerides (TRI), and lipid total.  Health status was self-assessed with a general self-rated health item asking participants to rate their overall health. Bivariate analyses were computed to examine zero-order correlations between major variables of interest.  Multinomial logistic regression (MLR) was then conducted to examine the predictive ability of the significant health correlates (gender, tobacco use, illicit substance use, education level, TRI, and BP) on 4 categories of BMI status (normal weight, overweight, obese, and morbidly obese).

Results: The majority of the sample was female (57.6%) and African-American (77.1%) and participants were, on average, 44.9 years old (SD=11.88). The average BMI score was 32.6 (SD=9.4) with the majority of participants at risk (81.5%, BMI>24.99). The MLR model correctly classified 44.8% of all cases, with 7.7% normal weight, 45.5% overweight, 14.7% obese, and 78.0% morbidly obese individuals correctly classified.  The model was statistically significant in distinguishing BMI categories (-2 Log Likelihood=378.66; X²(18) =58.53, p<.001). Relevant predictors accounted for approximately 32% of the variance in BMI categories (Nagelkerke pseudo-R² =.324). Among predictors, TRI and BP were significantly associated with obesity (OR=1.011,p<.05, OR=1.073,p<.01; respectively). Morbid obesity was significantly predicted by TRI, BP, gender, and substance use (OR=1.012, p<.05, OR=1.063, p<.01; OR=0.272, p<.01, OR=5.433, p<.01, respectively). Overall, higher TRI, higher BP, female gender, and no substance use indicated little change in the likelihood of being obese or morbidly obese.

Conclusions: Results provide preliminary evidence that the extant research examining obesity in persons with SMI may overemphasize the role of psychotropic medications and underemphasize gender differences and substance use. Findings further underscore the importance of employing a biopsychosocial approach to health.