Individuals with serious mental illnesses, such as schizophrenia, bipolar disorder, and recurrent depression, die earlier than individuals in the general population, losing nine to 32 years of life on average (
1–
4). Much of the premature mortality among people with serious mental illness is due to medical comorbidities, such as diabetes mellitus. Type 2 diabetes is common among persons with serious mental illness, and the likelihood of poor outcomes and complications is worsened by inactivity and poor diet as well as by treatment with second-generation antipsychotics (
5–
9).
Active self-management is crucial in minimizing the morbidity and mortality associated with chronic mental disorders and chronic general medical conditions (
10–
13). There are few practical treatment models for active self-management of serious mental illness with comorbid general medical illness (
10–
16), and a recent literature review found that the strength of evidence was low for most interventions (
13). Targeted training in illness management (TTIM) is a novel self-management approach that targets serious mental illness and comorbid diabetes concurrently. TTIM focuses on enhancing care engagement, including the use of peer educators to facilitate communication and to model behavioral changes. Preliminary work suggests that TTIM is highly acceptable to patients and may improve outcomes of both mental illness and diabetes when added to usual care (
17).
This randomized controlled trial (RCT) assessed effects of TTIM versus treatment as usual among 200 individuals with serious mental illness and diabetes. The primary aim was to test whether an intervention that targets serious mental illness and comorbid diabetes among patients in a primary care system will improve serious mental illness symptoms, functional status, general health status, and diabetes-specific outcomes.
Methods
This project was a prospective, 60-week RCT testing the effects of TTIM versus treatment as usual among 200 individuals with serious mental illness and diabetes. Individuals were randomly assigned to TTIM or treatment as usual by using a computer-generated list, allocation concealment, a 1:1 allocation ratio, and block randomization of consecutive patients, with block sizes ranging from four to eight patients. Primary outcomes were psychiatric symptom severity, functioning, general health, and diabetes control. Research assessments were conducted at baseline (prior to randomization) and at 13, 30, and 60 weeks. The study was conducted between December 2011 and June 2015.
Study Participants
Inclusion criteria included having schizophrenia, schizoaffective disorder, bipolar disorder, or major depressive disorder confirmed by the Mini-International Neuropsychiatric Interview (MINI) (
18); having type 2 diabetes; age ≥18; ability to communicate in English; and ability to provide written, informed consent. Exclusion criteria included active risk to self or others, inability to participate in study procedures, pregnancy, or dementia. The study was approved by the local institutional review board (IRB). Details on recruitment and retention methods have been published elsewhere (
19). Individuals were identified by clinician and self-referrals. In addition, individuals who had a serious mental illness on their problem list or who were being treated with medication for a serious mental illness (lithium, mood stabilizers, or antipsychotics) were identified from the health system’s electronic health record. Using an IRB-approved process, these individuals were consecutively contacted and invited to participate in the study.
TTIM Intervention
Designed for individuals with serious mental illness and diabetes, the TTIM intervention is a group-based psychosocial treatment that blends psychoeducation, problem identification, goal setting, behavioral modeling, and care linkage. TTIM is derived from the Life Goals Program (LGP), developed by Bauer and McBride (
20), and the Diabetes Awareness and Rehabilitation Training (DART), developed by McKibbin and colleagues (
12). Whereas LGP focuses mainly on mental health outcomes and DART focuses on diabetes outcomes, TTIM combines these foci and enhances social support by using peer educators.
TTIM is delivered in a two-step process described in detail elsewhere (
21,
22). Step 1 consists of 12 weekly, group-format, in-person sessions with six to ten participants per group that are codelivered by a nurse educator and a peer educator with serious mental illness and diabetes (
21,
22). [A table listing the topics for each session is available as an
online supplement to this article.] In step 2, which begins after the group sessions have ended and lasts for 48 weeks, participants have brief telephone maintenance sessions with peer educators and nurse educators. Telephone sessions, which last from ten to 15 minutes, occur every other week for the first three months and monthly thereafter.
Treatment as Usual
Patients with diabetes in this safety-net health system visit their primary care practitioner, on average, four to six times per year. [Materials describing the site characteristics and standard care are available in the online supplement.]
Intervention Evaluation
TTIM session attendance was assessed at each session. Acceptability of the intervention was assessed upon completion of step 1 with a brief, self-rated questionnaire. Fidelity to TTIM processes, content, and format was evaluated by a standardized checklist completed by noninterventionist study staff who randomly attended 20% of TTIM sessions.
Measures
Demographic and clinical variables, including health literacy screening (
23), were assessed at baseline. The self-reported Charlson Index evaluated medical comorbidity (
24). Primary outcomes evaluated four key domains: mental illness symptom severity, functioning, general medical health, and diabetes control. Other outcomes, which were not evaluated at 13 weeks, included two additional physical markers (body mass index [BMI] and blood pressure) and secondary outcomes directly related to diabetes control.
Serious mental illness symptoms.
The Clinical Global Impression (CGI) is a broad measure of global psychopathology that evaluates illness severity (
25). Possible scores range from 0 to 7, with higher scores indicating greater psychopathology. The Montgomery-Asberg Depression Rating Scale (MADRS) is a ten-item depression severity scale that is widely utilized in studies with patients with serious mental illness (
26). Possible scores range from 0 to 60, with higher scores indicating worse depression. The Brief Psychiatric Rating Scale (BPRS) measures psychotic and nonpsychotic symptoms in serious mental illness (
27). Possible scores range from 7 to 126, with higher scores indicating greater symptom severity.
Functional status.
The Global Assessment of Functioning (GAF) is a single-item scale that measures global functioning (
28). Possible scores range from 1 to 100, with higher scores indicating better functioning. The Sheehan Disability Scale (SDS) measures role impairment in three domains (work/school, family life/home, and social life) (
29). Possible scores range from 0 to 30, with higher scores indicating greater disability.
General health status.
The 36-Item Short-Form Health Survey (SF-36) is a self-report of general health (
30) divided into a physical component summary (PCS) and a mental component summary (MCS). Norm-based scores are placed on the same metric with a mean of 50 and a standard deviation of 10. Scores above 50 reflect higher functional status compared with the average population, and scores below 50 reflect lower than average functioning.
Diabetes control.
Diabetes control was evaluated with serum glycosylated hemoglobin (HbA1c) drawn at study baseline and at 30 and 60 weeks; HbA1c is an indicator of relative diabetes control over the past three months, ideally with scores of 7.0% or lower.
Secondary outcomes directly related to diabetes control.
TTIM is designed to affect behaviors that are key for diabetes control by increasing diabetes knowledge and self-care activities. Diabetes knowledge was assessed with the Brief Diabetes Knowledge Test (
12,
31). Higher scores indicate greater diabetes knowledge. Diabetes self-care was measured with the Diabetes Self-Care Activities Questionnaire (DSCA), a brief self-report of diabetes self-management that includes questions about general diet, diabetes-specific diet, exercise, glucose testing, foot care, and smoking (
32).
Data Analysis
Analyses were conducted in SAS, version 9.3; SPSS, version 23; and R software for a 64-bit Windows operating system. The level of significance was set at α=.05, except where noted otherwise. The groups were compared according to original assignment, regardless of their level of participation (intent to treat). The group × time interaction effect was compared by using linear mixed-effects analyses for each outcome. These series of mixed-effects models also included main effects for group and time along with a random intercept. A mixed-effects approach is well suited for handling missing data with a maximum likelihood algorithm because it assumes that the missingness is dependent on the data at hand (“missing at random” assumption) (
33). Bonferroni corrections were conducted to adjust for multiple comparisons within each of the four main outcome domains (mental illness symptom severity, functioning, general health, and diabetes control). Given the broad inclusion criteria used to identify patients with diabetes, post hoc analyses were conducted to compare changes in HbA1c among subgroups of patients demarcated on the basis of baseline HbA1c. Two groups were formed on the basis of guidance from the American Diabetes Association (ADA), which recommends adjusting threshold levels of HbA1c for individuals with comorbidities, such as mental illness. Wilcoxon rank-sum tests were conducted to compare change in HbA1c between the two arms over 30- and 60-week periods.
Results
Sample Description
A total of 358 individuals were screened for eligibility; 200 individuals were randomly assigned to one of the two study arms (TTIM, N=100; treatment as usual, N=100). Of individuals allocated to TTIM, 16 individuals (16%) never participated in a single TTIM session. The mean±SD number of group sessions attended was 7.2±4.6, and 65 (65%) TTIM participants completed six or more sessions. Most (N=158, 79%) of the sample received general medical care within the system, and 92 (46%) received mental health care within the system. [A figure illustrating the flow of patients throughout the study is available in the online supplement.]
Descriptive baseline statistics are shown in
Table 1. There were no clinically important differences between TTIM and treatment as usual as assessed by standardized absolute mean differences (
34). Individuals had substantial baseline depressive symptoms and low levels of psychosis as reflected in mean baseline scores on the MADRS and the BPRS, respectively. Functional status was low, with a mean GAF score of just over 50.
Use of psychotropic medication was extensive, with 82 (41%) on a first- or second-generation antipsychotic drug, 58 (29%) on a mood stabilizing drug, and 134 (67%) on an antidepressant. Among individuals with schizophrenia, 26 (53%) were on a second-generation antipsychotic drug, and 25 (45%) individuals with bipolar disorder and 22 (23%) individuals with depression were on a second-generation antipsychotic drug.
Safety and Tolerability
During the study, there were 119 adverse events among 74 participants. Adverse events occurred among six peer educators, 30 participants receiving treatment as usual, and 38 TTIM participants. There were three deaths (TTIM, N=2; treatment as usual, N=1). No adverse events were study related as determined by a data safety monitoring board.
Primary Outcomes
Table 2 illustrates mean changes in mental illness symptom severity, functioning, general health, and diabetes control from baseline to 13, 30, and 60 weeks. For psychiatric symptoms, there was a significantly greater improvement over the 60-week follow-up in both the CGI and the MADRS among TTIM versus treatment-as-usual participants. There was no significant difference between the two groups on the BPRS. Significant differences between the two groups in the psychiatric symptom domain for the CGI and the MADRS remained after Bonferroni adjustment. In terms of functioning, both groups showed improvement in GAF scores at 60 weeks, but the improvement was significantly greater in the TTIM group versus the treatment-as-usual group. There was a trend for greater improvement in SDS scores in the TTIM versus the treatment-as-usual group. Significant difference in functional status remained for GAF after Bonferroni adjustment. There were no significant group differences in SF-36 scores or HbA1c.
Secondary Outcomes Directly Related to diabetes Control
Change in diabetes knowledge was significantly better for TTIM versus treatment as usual (p<.001). Mean change from baseline to 13 weeks in diabetes knowledge was 8.47±20.1 for TTIM participants versus .11±16.3 for the control group out of a possible score of 23 (p<.02). At 60 weeks, participants in TTIM continued to demonstrate a significant improvement in diabetes knowledge (5.97±17.0), whereas the treatment-as-usual group showed a decline (−3.90±16.7) (p<.001). There were no significant differences on the DSCA.
Post Hoc Evaluations
The sample had substantial baseline heterogeneity in diabetes control, and the study design did not require that individuals meet a specific threshold for HbA1c in order to be included. A target HbA1c of <7.0% is recommended for nonpregnant adults with diabetes (
35,
36), but Ismail-Beigi (
37) recently suggested that HbA1c targets should reflect whether patients have comorbid illnesses and the complexity of these conditions. Specifically, psychological and social factors should be considered in approaching medical management of diabetes (
38,
39). This suggests that an appropriate (at least initial) HbA1c target for individuals with diabetes and comorbid serious mental illness might allow for more latitude compared with expected diabetes control among individuals without serious mental illness (
40).
Consistent with this recommendation, we conducted a post hoc subgroup analysis comparing the effects of TTIM among patients with levels of HbA1c that were moderately above the guideline target for all patients (>7.5%) versus those with better controlled HbA1c (≤7.5%). Of the analyzable sample of 196 individuals with available HbA1c data, 104 (53%) had baseline HbA1c levels of ≤7.5%. Two-sample nonparametric tests were used to investigate differences in outcomes between patients in the lower HbA1c group (≤7.5%) who were exposed to TTIM versus treatment as usual. Individuals in TTIM experienced minimal change in HbA1c over the 30- and 60-week time periods (median increases of .00 and .10, respectively), compared with treatment-as-usual recipients, who experienced worsening of diabetes control (median HbA1c increases of .25% and .50%, respectively). These changes were significantly different between study arms at both 30 weeks (p=.048) and 60 weeks (p=.024). When the analyses were limited to individuals with HbA1c levels that were further still above the ADA threshold, there continued to be a similar trend for better long-term maintenance of diabetes control among TTIM versus treatment-as-usual participants. For individuals with baseline HbA1c levels of ≤8% (N=122, 62% of analyzable data), there was a trend for TTIM participants to have better diabetes control at 30 (p=.070) and 60 (p=.055) weeks compared with treatment-as-usual participants.
Patient Satisfaction and Feedback
Of 84 participants who responded to a satisfaction survey on TTIM acceptability after completion of step 1, 98% (N=82) strongly agreed or agreed that TTIM is useful. Most (95%, N=80) strongly agreed or agreed that TTIM covers most of the important issues, whereas 92% (N=78) strongly agreed or agreed that TTIM addresses issues that are important to them.
Discussion
In this 60-week RCT, individuals with serious mental illness and diabetes who participated in the TTIM self-management approach had greater improvement in depression, global psychopathology, and functioning compared with individuals who received treatment as usual. Depression is associated with poor glycemic control and diabetes complications (
35,
36), so being able to improve mood and function in this high-risk group is clinically relevant. Glycemic control in the overall sample improved generally, and the mean results for the TTIM and treatment-as-usual groups did not differ significantly. However, for individuals with baseline diabetes control that was at or modestly above ADA recommendations (53% of the sample), post hoc analysis found better longer-term glycemic control in the TTIM subgroup. This is a clinically important subgroup of patients with serious mental illness and diabetes, who might particularly benefit from use of this relatively low-burden approach to enhance usual care. TTIM participants, as a whole, demonstrated significant improvement in diabetes knowledge at 13-week follow-up, compared with modest gains among recipients of treatment as usual. However, the TTIM group maintained these improvements during follow-up, whereas knowledge declined among recipients of treatment as usual. This is consistent with the theoretical basis of TTIM, which focuses on knowledge acquisition and application.
Despite recent attention to the benefits of integrating mental and general medical services among people with serious mental illness, progress has been hampered by a variety of barriers (
41–
45). In Ohio, where this study was conducted, individuals with serious mental illness account for 22% of the Medicaid population, and close to half (44%) of Medicaid expenditures are for patients with serious mental illness and co-occurring chronic medical conditions (
44). To address the multiple barriers to health management among persons with serious mental illness, one suggestion is to apply both “top down” approaches that address systems-level problems (reimbursement models) as well as “bottom up” approaches that actively engage patients. Self-management is a “bottom-up” approach that taps into a traditionally underused resource—the power of an individual to promote his or her health.
One strength of the TTIM trial was that it recruited relatively large numbers of persons from racial-ethnic minority groups, including African Americans and Hispanics, who made up 54% and 9%, respectively, of the sample. These subgroups are known to be particularly at risk for both diabetes and diabetes-related complications (
46–
49).
The TTIM study was unique in that it included individuals with schizophrenia and bipolar disorder and used peer educators to deliver the intervention. In contrast, a study by Katon and colleagues (
48) that targeted depression and medical comorbidity excluded individuals with schizophrenia and bipolar disorder. A substantial proportion of the TTIM study sample was also on second-generation antipsychotic drugs, which are associated with metabolic abnormalities and diabetes. Standard diabetes education may simply not be enough to alter health outcomes, as evidenced by the complications and early mortality seen among people with serious mental illness and comorbid diabetes. Our findings support the notion that partnering with peer educators to help deliver care not only can empower the patient (
21) but also may be a practical way to use the talents of individuals likely to be personally invested in helping others.
In our post hoc analyses among individuals who had reasonably good baseline diabetes control (HbA1c ≤7.5%), TTIM patients had less deterioration in HbA1c over 60 weeks compared with recipients of treatment as usual. It is not clear why individuals with poor baseline control of diabetes did not achieve better outcomes in TTIM versus treatment as usual. Local changes in the safety-net health system (Medicaid expansion) implemented during the time the study was being conducted may have promoted overall more aggressive diabetes management (
49,
50). Diabetes knowledge, which is believed to be critical to diabetes management, improved significantly more in the TTIM group. This indicator of an important intermediate goal may portend well for ultimate improvement in diabetes control. Some effective programs for people with serious mental illness include a focus on exercise and fitness (
51). Perhaps enhancing TTIM with exercise could be a future approach, as suggested by some TTIM participants.
There were a number of limitations to our study, including the single-site location, lack of data about mental health treatments and completed phone contact information, and the fact that psychotropic drugs were not a specific focus of intervention. Other limitations included possible effects of adverse events on study outcomes, even if they were not study related, and the fact that a research sample may not entirely represent a real-world population. Strengths included the randomized design, safety-net setting, and large proportion of persons from racial-ethnic minority groups. The study was rigorous in its primary analysis, including all individuals who were randomly assigned to a study arm regardless of session attendance. This article, which is the first to report outcomes of the TTIM study, focuses on overall change in symptoms, functioning, general health, and diabetes control. Future analyses will address sample heterogeneity, such as differential baseline diabetes risk variables, that may help inform the next generation of self-management interventions in serious mental illness.