Living Well: An Intervention to Improve Medical Illness Self-Management Among Individuals With Serious Mental Illness

Individuals with serious mental illness (schizophrenia, schizoaffective disorder, major depression, and posttraumatic stress disorder) have reduced life expectancy compared with the general population because of their elevated rates of chronic medical conditions such as diabetes and cardiovascular and respiratory disease (1–4). Poor self-management of chronic medical conditions among this group exacerbates the course of medical illness (5). Strategies to improve illness self-management among individuals with serious mental illness and co-occurring medical conditions are needed (6).

Peer interventions for illness self-management are led by individuals who are themselves managing a chronic medical condition and can provide effective models. One such intervention that has established efficacy is the Chronic Disease Self-Management Program (CDSMP), which consists of six peer-led group sessions, each lasting for 2 to 2.5 hours, that are focused on goal setting and problem solving (7). A growing workforce of peer specialists with lived experience of mental illness could be used to promote illness self-management among adults with serious mental illness (8).

Four studies have examined the CDSMP for individuals with serious mental illness. In a nonrandomized, pre-post intervention study, Lorig et al. (9) trained certified peer providers working at community mental health centers in the CDSMP; survey data indicated improvements on numerous health indicators. Druss et al. (10) adapted the CDSMP for individuals with serious mental illness, creating the Health and Recovery Peer Program (HARP), which consists of six 2.5-hour sessions. HARP has been tested in two randomized controlled trials (RCTs; 10, 11) with positive results; in the second RCT (N=400), HARP was associated with significant improvements in physical health– and mental health–related quality of life and mental health recovery (11).

Living Well (12) is also based on the CDSMP but was substantially revised in content and structure for ease of implementation with consumers with serious mental illness in outpatient mental health settings (e.g., shorter sessions, 12 sessions instead of six, repetition of material, option for cofacilitation by a nonpeer). In a pilot RCT (N=63), compared with usual care, Living Well was associated with significant improvements in self-efficacy, patient activation, illness self-management behaviors, and health-related quality of life at postintervention. Notably, the Lorig et al. (9) study was uncontrolled, whereas the Druss et al. (10, 11) and Goldberg et al. (12) studies used a usual care control condition. Thus, although a growing literature supports the use of the CDSMP model with individuals with serious mental illness, a large RCT with an active control condition has not occurred. Moreover, the CDSMP model has not been tested among veterans with serious mental illness; given that veterans on average have poorer health than nonveterans (13), this is a significant gap.

This study aimed to conduct a large RCT comparing Living Well, a 12-session group intervention co-led by a peer (a veteran with co-occurring mental and physical health conditions) and a nonpeer facilitator, with 12 sessions of a didactic Medical Illness Education and Support group led by a nonpeer facilitator. Active ingredients of the Living Well intervention that were not present in the control condition were as follows: a peer facilitator; skills training in goal setting, action planning, and problem solving; the setting of weekly health goals; and positive reinforcement of small steps toward health goals. We hypothesized that participants in the Living Well intervention would exhibit greater improvements in physical health– and mental health–related quality of life (the primary outcome), self-efficacy, patient activation, and illness self-management behaviors and greater decreases in emergency room (ER) use than participants in the active control condition at posttreatment and three-month follow-up.

Methods

Participants and Procedures

Participants were recruited from outpatient programs at three VA medical centers in the Mid-Atlantic region—two urban, one rural—between January 2014 and April 2016. Potential participants were identified by means of chart review, clinician referral, and self-referral. Inclusion criteria were a chart diagnosis of schizophrenia spectrum disorder, bipolar disorder, major depression with psychotic features, or posttraumatic stress disorder; chart diagnosis of a chronic respiratory or cardiovascular condition, diabetes, or arthritis; engagement in mental health services at a study site; capacity to consent; and approval from the participant’s mental health provider. Exclusion criteria included serious cognitive impairment or participation in a concurrent psychosocial treatment trial at the investigators’ home research center.

All study procedures were approved by the appropriate institutional review boards. Interested and eligible participants completed written informed consent. After giving consent, eligible participants completed baseline assessments and were randomly assigned to a condition. Assessors were master’s-level research assistants who received thorough training on all assessment measures, including intensive in-person training overseen by the principal investigator (RWG), repeated observation and feedback on assessments, and ongoing supervision.

We found no difference by gender or ethnicity among the proportion of participants who were randomly assigned to a treatment condition versus those who declined to participate or who did not complete a baseline assessment; in terms of race, African Americans were more likely to be randomly assigned (29%) than whites (19%; χ²=29.5, df=2, p<.001). [A figure showing the CONSORT diagram is available as an online supplement to this article.] Assessment measures were repeated after treatment (3 months after baseline) and at follow-up (6 months after baseline). Assessors were blind to treatment condition.

Interventions

Peer and nonpeer group facilitators for both conditions were trained and supervised by the study’s principal investigator (RWG) in weekly supervision sessions. Group sessions for both conditions were video recorded for fidelity.

Living Well is a manualized 12-session group intervention designed to enhance self-efficacy and motivation through education and skills training in action planning and problem solving. Sessions are 75 minutes long. A range of health-related topics are covered (e.g., healthy eating, physical activity, medication management, symptom management, and making good use of health care). Participants create weekly action plans related to each topic, report on their progress on previous action plans, and engage in group-based problem solving to overcome barriers to completing action plans. Peer providers complete weekly check-in calls with participants to provide problem solving for action plans. All groups were cofacilitated by a nonpeer provider (generally a research assistant with a master’s degree in clinical or counseling psychology—approximately the equivalent of a master’s-level clinician) and a peer provider, a veteran who had lived experience with co-occurring mental health and general medical conditions. Group leadership was shared equally by the peer and nonpeer facilitators. Both facilitators presented didactic information, aided participants in weekly goal setting and action planning, and helped facilitate problem solving. The peer facilitator also engaged in self-disclosure regarding his or her own illness self-management when it was relevant to session material and participants’ experiences.

The Medical Illness Education and Support group was a manualized 12-session intervention focused on living with a chronic medical condition. Sessions included standardized didactic review of common challenges experienced by people living with a chronic medical illness. Sessions were of similar length and frequency as Living Well. Groups were facilitated by a single nonpeer facilitator.

Measures

Basic demographic and psychosocial information was collected from participants. Current and lifetime medical conditions were solicited on the basis of interview items from the National Health and Nutrition Examination Survey III (14). Physical health– and mental health–related quality of life was assessed using the 12-item Short-Form Health Survey (SF-12; 15), a widely used instrument with good psychometrics in this population (16). To assess attitudinal change, we used several measures with strong psychometric properties: the six-item Self-Management Self-Efficacy scale, based on a measure used in the original CDSMP evaluation (17); the Patient Activation Measure (18), which measures self-reported ability to actively participate in treatment encounters; and the six-item internal health locus of control subscale of the Multidimensional Health Locus of Control Scale (19), which measures self-perceived control over one’s health.

We used the Instrument to Measure Self-Management (IMSM; 17), which is based on a measure used in the original CDSMP evaluation, to assess change in specific and general self-management behaviors and the Morisky Medication Adherence Scale (20), a self-report scale with good psychometric properties (20), to assess medication adherence.

To assess psychiatric symptoms and recovery, we used the Behavior and Symptom Identification Scale 24—Modified (BASIS-24; 21), a self-report instrument that measures psychiatric symptoms and has well-established psychometric properties, and the Maryland Assessment of Recovery Scale, a 25-item self-report instrument that measures recovery orientation and has good internal consistency and test-retest reliability (22).

We extracted ER use data from electronic medical records for the six months before baseline and the six months of the study. The outcome of ER visits was coded as binary, indicating whether the individual had an ER visit focused on a medical issue during either of the six-month timeframes.

Data Analysis

For continuous scale outcomes, we used linear mixed-effects models to test the effectiveness of Living Well compared with Medical Illness Education and Support. A random effect for class accounted for an intraclass correlation resulting from common group membership. Within-individual correlation among measurements over time was accounted for with an unstructured error correlation matrix. Regression terms in the model included intervention, time, and intervention×time interactions. Time was entered into the model by using two dummy variables—one for posttreatment and one for the follow-up time point—to compare group mean change from baseline to posttreatment separately from baseline to follow-up. The interaction terms estimated mean change in the Living Well group minus mean change in the control group. All available outcome measure data from baseline, posttreatment, and follow-up were included (intention-to-treat). All continuous outcome scales were checked for skew, and an appropriate transformation was applied as needed.

For the ER visit (yes-no) outcome, we used a logistic mixed-effects model for binary outcome. That model was specified parallel to the linear mixed-effects model for continuous scale outcomes, except that time represented two time periods—the 6-month period prior to baseline (“pre”) and the 6-month period after baseline (“post”).

This model estimates the probability of an ER visit during a time period on a log-odds scale. In particular, the interaction term estimated the difference in pre- to postchange in probability of an ER visit between the Living Well group and the control group on a log-odds scale. A negative coefficient estimate represents a greater reduction in Living Well relative to the control condition.

Statistical significance was defined as p<.05. All analyses were performed using SAS version 9.4.

Results

Intervention Attendance and Fidelity

Living Well participants attended a mean±SD of 5.4±4.4 of 12 weekly sessions, and Medical Illness Education and Support participants attended a mean of 6.4±4.2 of 12 weekly sessions. In the Living Well condition, 99 (80%) participants attended at least one group session, and 65 (52%) attended at least five group sessions. In the Medical Illness Education and Support condition, 100 participants (85%) attended at least one group session, and 76 (64%) attended at least five group sessions. Living Well participants were offered three monthly booster sessions after the 12-week curriculum, and they attended a mean of 0.8±1.1 booster sessions, with the majority of participants (N=74; 60%) not attending any booster sessions. Table 1 summarizes demographic and other characteristics of the participants.

A total of 30 Living Well sessions and 18 Medical Illness and Support sessions were rated by independent reviewers on facilitator competence and adherence to content. Each item was scored on a 3-point scale (0, unacceptable; 1, acceptable; and 2, excellent). Across the 30 recorded Living Well sessions, the full curriculum and all possible facilitator pairings were represented at least once. Mean scores were high in both the Living Well group (adherence=1.99; competence=1.98) and the Medical Illness and Support group (adherence=1.97; competence=1.95). No rated sessions included any ratings of unacceptable.

Outcomes

Treatment outcomes by group are displayed in Table 2. We found no group differences in SF-12 General Health Functioning score. At posttreatment but not at follow-up, Living Well participants exhibited greater increases in mental health–related quality of life and Medical Illness Education and Support participants exhibited greater increases in physical health–related quality of life. Living Well participants showed greater improvement on the Self-Management Self-Efficacy scale at posttreatment and follow-up and greater increases on the internal health locus of control subscale and the Patient Activation Measure at posttreatment but not follow-up. Participants in Living Well showed greater improvement on the IMSM behavioral and cognitive symptom management subscale at posttreatment and follow-up and greater improvement on the IMSM physical activity subscale at follow-up but not posttreatment, indicating a delayed effect. We found no other significant group differences for the IMSM and no group differences for the Morisky Medication Adherence Scale.

Living Well was associated with a greater decrease in BASIS-24 overall symptoms at posttreatment but not at follow-up and with delayed improvement on the BASIS-24 relationships subscale at follow-up. No other significant group differences were found for the BASIS-24 scales. There was a nonsignificant trend for greater improvement on the Maryland Assessment of Recovery Scale among Living Well participants at posttreatment but not at follow-up. No group differences in ER use were found at any time point.

Discussion and Conclusions

In a large, well-controlled RCT, participation in Living Well, a medical illness self-management intervention for individuals with serious mental illness co-led by peer and nonpeer facilitators, resulted in significant improvements in mental health– but not physical health–related quality of life immediately after treatment, as well as in improvement in self-efficacy and behavioral and cognitive symptom management at both posttreatment and follow-up, compared with participation in an active control condition, a didactic medical illness and support group. Compared with the control condition, Living Well was also associated with improved patient activation, psychiatric symptoms, quality of social relationships, and physical activity–related self-management.

Living Well had a large effect on self-efficacy that was maintained at follow-up. Adults with serious mental illness often have poor self-efficacy for health behaviors (23), which may result from limited exposure to role models and social support for wellness (24). Using a peer facilitator and a group-based format, Living Well provided a credible source of social support for health behaviors, which may have been a key ingredient in enhancing illness management self-efficacy. Illness management self-efficacy may, in turn, have led to increased self-management behaviors at postintervention or 3-month follow-up (i.e., behavioral and cognitive symptom management, physical activity–related self-management).

At posttreatment, participants in Living Well exhibited improvements across health attitudes, behavioral and cognitive symptom management, psychiatric symptoms, and mental health–related quality of life. At follow-up, Living Well was associated with improvements in interpersonal relationships. Although we did not test a mediation model, it may be that Living Well helped participants feel empowered to engage in symptom management behaviors, which were then associated with a decrease in psychiatric symptoms and an improvement in relationship quality and mental health functioning. The majority of these outcomes were significant at posttreatment and not maintained at follow-up, indicating that a longer intervention with more intensive strategies for maintenance and generalization may be warranted.

Although use of behavioral or cognitive symptom management strategies improved at posttreatment among Living Well participants, self-management of physical activity did not increase until three-month follow-up. It may be that individuals with serious mental illness initially implement stress management strategies because this is an area of more immediate need or because these strategies may easiest for them to internalize because they have previously been exposed to them in their mental health treatment. If Living Well were to be expanded to a longer intervention, facilitators might focus in the first several weeks on stress management before shifting to a focus on other health behaviors.

The only outcome that exhibited greater improvement in the Medical Illness Education and Support group was physical health–related quality of life, which was not maintained at three-month follow-up. It may be that to have an impact on physical health outcomes, an exclusive focus on physical health topics is required. Alternatively, it may be that for Living Well to have an impact on physical health–related quality of life, a longer intervention may be required to allow for attitudinal and behavioral changes to have eventual downstream impacts. Of note, the delay in physical activity change may mean that changes related to physical health–related quality of life would occur after the three-month time point; future studies with longer follow-up periods should explore this possibility.

Although several findings indicate that delivery of Living Well over a longer time period might be necessary to promote maintenance and generalization, it is possible that the intervention could be less intensive and still be effective. Living Well required a significant time commitment (attendance at 12 weekly face-to-face sessions), which may have contributed to the approximately 75% of potential participants who were contacted and declined to participate in the intervention. Among those who did consent, level of attendance varied widely, with half of the participants attending four or fewer sessions. Despite this, Living Well was associated with better short-term outcomes than the active control condition across a number of measures. This finding indicates that exposure to all of the intervention content was not necessary for a positive outcome; thus the active ingredients of Living Well (peer contact; skills training in goal setting, action planning, and problem solving; weekly goal setting; reinforcement of small steps toward health goals) could be delivered in a more condensed and easily accessible format, promoting more widespread participation. For example, there is a growing evidence base for Web- and app-based health and wellness interventions, supplemented with peer coaching support, for individuals with serious mental illness (25–27). Future studies should examine innovative ways to package and deliver Living Well to maximize accessibility, efficacy, and generalization.

Living Well did not affect ER use. Notably, the rate of ER use among the sample was considerable, which speaks to the medical illness burden among this population as well as the barriers that make it hard for such individuals to engage proactively in health care. It is possible that a psychosocial group intervention such as Living Well is not enough to address this problem, which may be attributable more to health care system issues. An intervention that directly connects individuals with serious mental illness to preventive health care may be needed to address their high levels of ER use.

The large sample size, recruitment of a veteran population with significant illness burden, and an active control condition were strengths of this study. Limitations included the fact that many of the outcome measures were based on self-report, the follow-up period was relatively brief, and the population was a majority male population that was engaged in mental health services, which could limit generalizability. In addition, attendance at intervention sessions was sporadic, with half of the participants attending four or fewer sessions; future studies should examine the optimal dose of the intervention required to achieve a positive benefit. Nonetheless, our findings support the use of a peer-cofacilitated psychosocial intervention to improve aspects of medical illness self-management among veterans with serious mental illness, especially in the short term. Specifically, the intervention was found to promote self-efficacy, patient activation, and other health attitudes; increase some illness self-management behaviors; attenuate psychiatric symptoms; and improve mental health–related quality of life. Future studies should examine strategies to maintain and generalize gains over a longer period and examine augmenting Living Well with direct provision of preventive health care.