Clinical Characteristics of Individuals With Serious Mental Illness and Type 2 Diabetes

Baseline data were from 157 adults enrolled in a randomized controlled trial funded by the National Institute of Mental Health that compared a novel intervention with treatment as usual among individuals with comorbid serious mental illness and diabetes. The study was conducted in a safety-net primary care setting. It was approved by the local institutional review board, and written informed consent was obtained. Enrollment for the ongoing study began in November 2011. Inclusion criteria include having schizophrenia, schizoaffective disorder, bipolar disorder, or major depression confirmed with the Mini-International Neuropsychiatric Interview (5) and having diabetes, confirmed by history or laboratory values. Exclusion criteria include being actively suicidal or homicidal, having dementia, or being unable to participate in groups.

Diabetes control over the previous three months was assessed by measuring glycosylated hemoglobin (HbA1c). Depression was measured with the Montgomery-Åsberg Depression Rating Scale (MADRS) (6). The Brief Psychiatric Rating Scale (BPRS) measured global psychopathology (7). Diabetes knowledge was assessed with the Brief Diabetes Knowledge Test, which has been used successfully with people with serious mental illness (8). Health literacy was assessed with a measure utilized to identify persons with limited health literacy in primary care (9).

Analyses were conducted with SAS, Version 9.2, and R. We used p values from the nonparametric Kruskal–Wallis one-way analysis of variance by ranks across diagnostic groups to compare subgroups of people with serious mental illness. Bivariate Spearman correlations between HbA1c, MADRS, and BPRS scores, both overall and within the three diagnosis groups were assessed. Linear regression models examined whether the association between the dependent variable (HbA1c) and the independent variables (MADRS, BPRS, and diagnosis group) changed after the analyses controlled for covariates. In separate models that also adjusted for any lower-level interaction terms and main effects, our primary measures of interest were a three-way interaction term between MADRS, BPRS, and diagnosis; a two-way interaction term between diagnosis and MADRS and diagnosis and BPRS; main effects for diagnosis, BPRS, and MADRS; a main effect for only diagnosis; a main effect for only BPRS; and a main effect for only MADRS. Covariates included factors related to diabetes control (race, gender, health literacy, diabetes knowledge, and diabetes duration).

We then performed residual analysis and made appropriate transformations to the data when necessary, if regression assumptions were violated. As a result of the residual analysis, we used a logarithm transformation for HbA1c. Finally, using a structural equation modeling approach, we separately assessed associations between the five BPRS factors (affect, positive symptoms, negative symptoms, activation, and resistance) and HbA1c, forming latent constructs using confirmatory factor analysis. All statistical tests were two-tailed, with significance defined as α=.05.

Seventy-seven participants (49%) had major depression, 40 (26%) had schizophrenia or schizoaffective disorder, and 40 (26%) had bipolar disorder. The mean±SD age was 52.9±9.8. A total of 102 (65%) were women, 62 (40%) were African American, and 15 (10%) were from other racial minority groups. The mean duration of serious mental illness was 18.1±12.2 years, and the mean duration of diabetes was 10.4±7.8 years.

The diagnostic groups differed by race, with the proportion of Caucasians significantly larger in the schizophrenia group (p<.01). The duration of diabetes was significantly longer in the depression group (p<.02). The duration of mental illness was significantly longer in the schizophrenia group (p<.05). Health literacy for completing forms was significantly poorer in the schizophrenia group (p<.02). No significant cross-group differences were noted in age, education, or gender. [A more detailed summary of subgroup characteristics is presented in an online data supplement to this report.]

The diagnostic groups differed significantly (p<.01) in HbA1c levels. The mean level was highest in the major depression group (all groups had a mean level ≥7.5%). Diabetes knowledge was significantly poorer in the schizophrenia group (p<.01). The groups did not differ in insurance status, marital status, insulin use, body mass index, or self-reported presence of significant general medical comorbidity as measured by the Charlson Comorbidity Index (10).

Participants were generally moderately to severely depressed (mean MADRS score for all groups in the sample >21). Global psychopathology severity was low (mean total BPRS score for all groups <45). The groups differed significantly in BPRS total score (p<.001) but did not significantly differ in MADRS scores. No association was found between diagnosis and HbA1c in our linear regression model after adjustment for race, gender, health literacy, diabetes knowledge, and duration of diabetes. No significant interaction terms with diagnosis group were noted in any of our linear regression models.

We analyzed the relationship between HbA1c and total scores on the BPRS and MADRS by using regression analysis. Our hypothesis that BPRS and MADRS scores would be positively correlated (r=5.57) was confirmed. However, although HbA1c and MADRS scores were positively associated (r=5.26), the bivariate Spearman correlation between HbA1c and BPRS scores was close to zero (r=5.02). After adjusting for race, sex, health literacy, diabetes knowledge, and duration of diabetes in our multivariate regression model, we found that both MADRS and BPRS scores were associated with HbA1c levels (Table 1). In the multivariate model, MADRS total scores appeared to be positively correlated with HbA1c, whereas the association of BPRS scores with HbA1c was negative. In regard to BPRS factors, the affective factor was positively associated with HbA1c, similar to the MADRS scores and HbA1c. The other BPRS factors, which are conceptually concordant with psychosis, were inconsistently and weakly associated with HbA1c. Longer duration of diabetes was associated with poorer diabetes control. We found no significant interaction between MADRS and BPRS scores.

In our sample of patients with serious mental illness in a primary care, safety-net health system, psychiatric symptoms appeared to be related to diabetes control. After the model controlled for covariates, more severe depressive symptoms were associated with poorer diabetes control. After control for covariates in our statistical model and adjustment for depression severity, the severity of global psychopathology, as measured by the BPRS, was unexpectedly and negatively related to HbA1c levels. Clinical interpretation of MADRS and BPRS scores in our sample suggests that participants were mainly depressed and that symptoms of schizophrenia or bipolar disorder, such as hallucinations or mania, were relatively subtle. It seems counterintuitive that lower BPRS scores would be related to worse diabetes control. However, it must be noted that psychotic symptoms in this sample were mild. In the low range of BPRS total scores, minor differences in scores may be of limited clinical relevance. It is also possible that individuals with more obvious psychotic symptoms receive additional clinical attention that is helpful for diabetes outcomes. It also must be noted that the inconsistent and relatively weak finding for the BPRS psychosis factor limits interpretation of the relationship between BPRS and diabetes control in this analysis.

Pervasive and severe depressive symptoms appear to have been a major impediment to diabetes self-management in our sample. This is consistent with increasing awareness of the impact of depression on diabetes, although reviews of this relationship are limited by the fact that they have not generally included people who have depression in the context of chronic psychotic or cyclical mood disorders (3,11).

Serious mental illness that is characterized by disorders of thought, perception, or mood present unique challenges to clinicians and caregivers. Our findings, although limited by cross-sectional methods, single-site enrollment, and absence of data on health service use, have clinical relevance. In our sample, patients with serious mental illness and diabetes tended to be quite depressed. Low health literacy, low levels of functioning, and limited insight might all be expected to prevent individuals from being able to benefit from diabetes programs available in routine primary care settings. Psychiatric symptoms, especially severe depression, may impede management of the many components of successful diabetes self-care. Effective self-management approaches need to take into account the specific and unique challenges that serious mental illness can create for treating comorbid diabetes. A growing literature suggests that diabetes care that is targeted to people with serious mental illness can improve outcomes (4), but more research is needed on refining approaches and on how such approaches can be translated to practical venues for diabetes care.

The research was supported by award R01MH085665 from the National Institute of Mental Health (NIMH). The project was also supported by grants UL1 RR024989 and KL2TR000440 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH). The contents of this report are solely the responsibility of the authors and do not necessarily represent the official view of NIMH, NCRR, or NIH.

Dr. Sajatovic has received research support or served as a consultant to Amgen, Bracket, Janssen, Merck, Ortho-McNeil Janssen, Otsuka, Pfizer, Prophase, Reuter Foundation, Woodruff Foundation, and Reinberger Foundation. The other authors report no financial relationships with commercial interests.