The Illness Management and Recovery (IMR) program is a curriculum-based approach to helping consumers with severe mental illness set and achieve personal recovery goals and acquire the knowledge and skills to better manage illnesses (
1). IMR has been tested in multiple quasi-experimental studies (
2–
7) and three randomized controlled trials (
8–
10). A growing body of research has also addressed implementation and adaptations of IMR (
11), illustrating its widespread popularity.
IMR was developed to incorporate effective strategies for managing illnesses such as schizophrenia (
12), including psychoeducation, cognitive-behavioral approaches to medication adherence, relapse prevention, social skills training, and coping skills training. The information and skills taught in IMR are conceptually organized around the stress-vulnerability model (
13,
14) and are aimed at improving illness through reducing biological vulnerability (for example, increasing medication adherence and reducing substance abuse), reducing the impact of stress (for example, providing coping skills training and facilitating meaningful activities), and increasing social support (for example, improving relationships). The IMR curriculum includes ten topic modules taught individually or in groups. By teaching consumers to effectively manage mental illness and work toward personal goals, IMR seeks to help them become more active in treatment and make progress in recovery, including both subjective (for example, hopefulness) and objective (for example, improved role functioning) aspects of recovery; and to reduce use of inpatient and crisis services.
Research on IMR as a package has demonstrated the program’s effectiveness, particularly regarding increased illness management and symptom reduction (
11). The randomized trials have been conducted in very different settings: 13 community agencies in Israel (
8), three supportive housing settings in New York City (
9), and six Swedish psychosocial rehabilitation centers (
10). Each of these studies used a “treatment as usual” condition that varied substantially; however, common elements included case management, medication management, and access to other rehabilitation services. Although not consistent across outcome domains, these trials all produced results showing advantages for IMR participants. However, no studies have compared IMR to equally intensive interventions as a means of controlling for nonspecific treatment factors, such as clinician attention. More rigorous evaluation of the benefits of the IMR curriculum and teaching methods above and beyond nonspecific therapeutic factors is needed.
In addition, none of the prior published studies of IMR has addressed the needs of veterans. The U.S. Department of Veterans Affairs (VA) is investing resources to improve mental health treatment for veterans, including evidence-based community mental health programming (
15) and an increased emphasis on recovery-oriented care (
16). Despite these initiatives, veterans’ needs remain great. For example, in 2004, the average length of an inpatient psychiatric hospitalization in VA facilities was 13 days; in addition, rates of readmission were high, with 31% readmitted within six months (
17). One study directly comparing veteran with nonveteran consumers with schizophrenia found similar levels of subjective recovery (for example, hope, satisfaction with life, and empowerment) but lower perceived knowledge about illness management among veterans (
18). Taken together, these findings suggest the need for interventions that can help veterans take greater control of their illness and rely less on costly mental health services.
The purpose of this study was to rigorously test IMR in a sample of veterans and nonveterans with schizophrenia spectrum disorders to determine whether IMR improves outcomes above the effects observed for an active control group. We hypothesized that IMR would lead to greater improvements in symptoms, functioning, illness management, medication adherence, and subjective experiences of recovery, as well as reduced utilization of crisis and hospital services.
Methods
A randomized controlled trial compared group-based IMR to an equally intensive problem-solving (PS) control group. Both interventions were offered weekly for nine months. All participants continued to receive other usual treatment, which could include medication management, case management, individual or group therapy, and other psychosocial treatment. The study took place at a Psychosocial Rehabilitation and Recovery Center of a VA medical center and at a community mental health center (CMHC) in the same city. Recruitment occurred between September 2008 and October 2010, with assessments conducted at baseline, nine months, and 18 months. Procedures were approved by the Indiana University–Purdue University Indianapolis Institutional Review Board and the Research and Development committee of the Roudebush VA Medical Center
Participants
Inclusion criteria were as follows: currently receiving (or newly admitted to) mental health services at the VA or CMHC, age 18 or older, diagnosis of schizophrenia or schizoaffective disorder, and willingness and ability to give informed consent. Exclusion criteria were as follows: medical condition limiting participation in an 18-month study (for example, end-stage renal disease) or evidence of severe cognitive impairment as assessed by a screening instrument (
19). All 52 participants from the VA were veterans, and four of the 66 CMHC participants were veterans.
Participants were recruited through clinician referrals, self-referral, and a registry of individuals who had participated in previous research. Members of the research team attended clinical team meetings to describe the study and distribute brochures, which were then given to consumers and posted in treatment areas. A total of 118 participants were recruited and randomly assigned to either IMR (N=60) or the PS control group (N=58). [A flow chart showing study recruitment is included in an online
data supplement to this article.]
Program models
IMR was offered in small groups (fewer than eight), cofacilitated by either an experienced master’s-level clinician or a doctoral-level psychologist and by a doctoral student in clinical psychology. The initial facilitator was trained directly by a developer of the IMR model and participated in biweekly phone supervision for several months; we added two lead facilitators over time, and the first one moved during the study period. Through the rest of the study, facilitators met regularly for peer supervision. Facilitators used the IMR curriculum (
20), incorporating psychoeducation, cognitive-behavioral approaches, relapse prevention, social skills training, and coping skills training. Facilitators worked with group members to set personal recovery goals and address progress toward those goals throughout the intervention. Home assignments helped participants apply newly learned skills or make progress on goals. Groups were open to rolling admission across the study period, and participants were considered to have graduated from the group when they had the opportunity to attend nine months of sessions. All participants had the opportunity to attend the full number of sessions.
The PS intervention was used to control for attention and time in groups. Participants were encouraged to discuss current concerns and receive group support for solving problems, but we did not use structured problem-solving tasks. These groups were led by the same facilitators who led the IMR groups. They helped establish group expectations (attendance and confidentiality), encouraged participation, and provided process-oriented observations. There was no formal curriculum, goal setting, or homework assignments.
To ensure that the experimental condition was following the IMR model and that the control condition was not, we audiotaped all sessions and rated fidelity using the IMR Treatment Integrity Scale (
21), a 16-item, behaviorally anchored scale. We randomly selected 60 IMR and 20 control sessions to rate. Raters were four graduate students with experience providing IMR and trained to make fidelity ratings; they were not told which condition they were rating. On a 5-point scale, with higher scores indicating greater fidelity, mean total scale scores were significantly higher for IMR sessions (3.4±.8) than for PS sessions (2.2±.3) (t=11.68, df=78, p<.001). Means for each of the items were higher for IMR than for PS, except for “enlisting mutual support,” which was similar across conditions.
Measures
Participants in both conditions were interviewed at baseline, nine months, and 18 months by trained raters blinded to study condition. Participants were paid $20 for each interview. We obtained data on emergency department and hospital utilization through medical records. At baseline, diagnoses were assessed with the psychosis modules of the Structured Clinical Interview for DSM-IV (
22) administered by either a clinical psychologist or trained doctoral student. Because IMR was designed to help consumers become more active in their own treatment and to make progress in recovery, we included a range of measures to tap activation and both subjective and objective indices of recovery.
Psychiatric symptoms were assessed by the Positive and Negative Syndrome Scale (PANSS) (
23), a widely used, 30-item, interview-based rating scale. The PANSS has demonstrated satisfactory internal consistency, test-retest reliability, and validity (
23). Raters were trained to an interrater reliability of .80 prior to interviewing participants. We examined the total score and five factors —positive symptoms, negative symptoms, emotional discomfort, hostility or poor impulse control, and cognitive symptoms (
24).
Quality of life was assessed with an abbreviated version of the Quality of Life Scale (QLS) (
25), commonly used with patients with schizophrenia. The QLS includes questions and objective indicators for interviewers to rate social and occupational functioning during the prior four weeks. The abbreviated version includes seven of the original 21 items, providing a reliable, brief measure (
26).
Illness self-management was assessed with the consumer-rated Illness Management and Recovery Scale (
27). Items are rated on a 5-point behaviorally anchored scale, and the mean across all 15 items forms an overall score, with higher scores indicating better self-management. The IMR Scale has shown internal consistency, stability, sensitivity to change over time, and correlations with indices of functioning, symptoms, and recovery (
28–
30).
Patient activation was assessed with the short-form, mental health version of the Patient Activation Measure (
31). Possible scores range from 0, least activation, to 100, highest activation. The 13-item mental health version has been shown to have strong reliability in Rasch analyses and test-retest reliability and to correlate with related constructs (
32); it has been used in other samples with schizophrenia (
33–
35).
Medication adherence was assessed with the Morisky Scale, a four-item scale with adequate reliability and validity across populations (
36), including persons with severe mental illness (
37). Participants answered yes or no to items related to medication adherence; a score sums the number of yes responses, and a lower score indicates better adherence.
Perceived recovery was assessed using the total mean score of the Recovery Assessment Scale (RAS) (
38). Respondents endorse 41 items (for example, “I have a desire to succeed”) on a scale from 1, strongly disagree, to 4, strongly agree (without a neutral option). The RAS total score has shown good test-retest reliability and internal consistency and correlates with measures of self-esteem, empowerment, and quality of life (
38).
Hope was assessed using the six-item adult version of the State Hope Scale (
39), which has high internal consistency and convergent and discriminant validity (
39). The scale has also been used with individuals with schizophrenia (
40,
41).
Service utilization was extracted from medical records, including number of visits to the emergency department and number of admissions to and length of stay in inpatient units. Data were extracted for nine months prior to baseline, baseline to nine months, and nine to 18 months.
Data analysis
We compared the IMR and the PS participants on background and outcome variables using chi square tests for categorical variables and t tests for continuous variables. Intent-to-treat analyses compared changes in IMR and PS groups on the outcome measures over time. To explore whether outcomes differed between consumers who had some engagement in their randomly assigned intervention, we also conducted analyses to compare participants who attended at least one group. These analyses produced results similar to those of the intent-to-treat analyses and are not presented here.
We followed an analysis approach similar to those used in recent randomized controlled trials, with mixed-effects regression analyses of mean-response profiles (
42,
43). We included the baseline measure for a given outcome variable as a covariate, as well as site, and we adjusted the fit of mean-response profile models (
44) using SAS PROC MIXED for continuous outcomes (interview data) and SAS PROC GENMOD (Poisson regression) for count outcomes (service data). This approach can accommodate missing data as well as correlated residuals by selecting appropriate covariance structures with maximum likelihood estimation (
45). The group main effects in this model (that is, the difference in group mean-response profiles between the IMR and PS groups at posttreatment and follow-up) were the primary test of the study hypotheses. To evaluate changes over time across the outcome measures from baseline to nine and 18 months, we performed similar analyses but included all measurement occasions as dependent variables using SAS PROC MIXED and SAS PROC GENMOD. Finally, we examined whether rates of intervention exposure (percentage attending the intervention) were related to changes in outcomes over time. We employed the significance level at a p value of .05 or less.
Results
Participants assigned to IMR did not differ significantly from those assigned to the PS control group on baseline measures (
Table 1). Most participants were male (80%), African American (61%), and living independently (70%). The mean age was 47.7±8.9. Most were unemployed (86%) and reported an annual income of less than $10,000 (63%).
Rates of treatment participation were low for both groups but did not significantly differ. For IMR, 14 individuals (23%) attended no sessions, 29 (48%) attended fewer than half the scheduled sessions, and 17 (28%) attended half or more. For the PS control group, 20 (34%) attended no sessions, 28 (48%) attended fewer than half the scheduled sessions, and ten (17%) attended half or more. Older age, lower hostility, fewer psychotic symptoms, and more education were associated with higher attendance in both conditions (
46).
Outcomes across time are shown in
Table 2. Analyses of mean-response profiles revealed no group differences between IMR and PS on any of the outcomes. Time effects showed improvements in several variables, including all symptom domains (Cohen’s d >.5), functioning on the abbreviated QLS (Cohen’s d=.4), self-reported illness management (Cohen’s d=.4), and emergency department visits (Cohen’s d=.3). There were main effects for site, with consumers from the VA having lower scores on some of the outcome variables (that is, self-reported illness management [p<.05] and total RAS scores [p<.05]), but there were no site-by-group interactions. Additional analyses accounting for percentage of sessions attended indicated that treatment exposure did not explain the variance in outcomes.
Discussion
Participants in both groups improved significantly in a number of domains related to illness management, including symptoms, psychosocial functioning, self-rated illness management, and emergency department use. However, in contrast to three other randomized trials showing some advantages for IMR over usual care (
8–
10), there were no significant differences between the IMR and PS groups on any outcomes. Here we offer several possible reasons for these findings.
This was the first randomized trial of IMR that controlled for nonspecific treatment factors with an active and equally intensive intervention, rather than usual services. Although the control intervention did not take a structured approach to teaching problem solving, it did provide a forum to discuss problems and concerns with experienced clinicians and to receive support from group members, which have been identified as effective components of group interventions (
47,
48). Our fidelity ratings showed no differences between the IMR and PS groups in enlisting mutual support. Thus it is possible that the benefits of IMR over usual services shown in previous studies were primarily attributable to nonspecific therapeutic factors in IMR groups, rather than the specific curriculum, teaching methods, and recovery orientation of IMR. A recent Cochrane review of cognitive-behavioral therapies for schizophrenia reported a similar lack of findings when these approaches were compared with active control groups (
49). However, there are important alternative explanations to consider related to implementation of and participation in the group interventions.
Participation rates in both the IMR and PS groups were lower than desired, which could have limited the effect of the intervention. Only 28% of consumers assigned to IMR and 17% of those assigned to the PS group participated in more than half the scheduled groups, and 23% and 34%, respectively, attended no sessions. Although the percentage of sessions attended was not related to outcomes in our analyses for either group, it is conceivable that a minimum threshold of exposure to IMR is necessary to detect treatment benefits, and even higher levels of exposure may be necessary to distinguish specific from nonspecific effects of IMR.
Rates of exposure to IMR in this study are lower than those reported in the previous three controlled trials of IMR. One study in a residential setting reported that 54% of consumers (almost twice our rate) attended at least half of the sessions (
9). Participants in Färdig and colleagues’ (
10) sample attended an average of 30 of the 40 sessions offered, and all attended at least half; however, participants were recruited on the basis of consistent attendance in prior mental health services, and implementation of IMR focused heavily on consumer engagement (Färdig R, personal communication, Dec 19, 2012). The controlled trial that took place in Israel reported that 7% were dropped from analyses for not participating, but completion rates for IMR were not reported (
8).
In our study, participants were asked to attend weekly for nine months. Although group leaders attempted to facilitate attendance with reminders and phone calls, assertive outreach was not feasible. This may represent a common scenario in community mental health settings and suggests that additional strategies may be needed to engage participants in a weekly intervention for the better part of a year. A possible limitation in our study design was rolling admission into groups. This was done so that participants would not wait long between recruitment and intervention; however, rolling admission may have led to less group cohesion or mutual accountability, which could have had an impact on participation. In addition, in both the VA and CMHC settings there was a lack of integration of IMR with other services and with documentation in medical records. The lack of integration may have limited the ability of other treatment providers to facilitate consumers’ work on recovery goals, encourage IMR attendance, and reinforce skills learned in the program.
The interaction of IMR with other services is poorly understood and may have affected responsiveness in the current trial. Veterans participating in this trial generally had received extensive services prior to participating in IMR, including psychoeducation, coping skills, and relapse prevention groups. It is unclear in reports of the previous IMR trials how many participants accessed other rehabilitation services concurrently or previously. Although we controlled for site in the analyses, more work is needed in the VA context, particularly because illness management is a required component of recovery-oriented care in VA psychosocial and recovery centers (
50). Finally, the sizes of our samples were somewhat smaller than our a priori power calculation (N=62 for each group assuming 20% attrition), and attrition was higher, resulting in slightly less overall power to detect moderate effects.
Conclusions
The findings highlight important areas for future study of IMR effectiveness. Although outcomes for IMR participants were not better than those for PS group participants, participants in both groups experienced improvements in several domains. Issues of treatment exposure, potency, context, and population served may have all affected outcomes and serve as important targets for future study.
Acknowledgments and disclosures
The research reported here was supported by the U.S. Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service (IAC 05-254, HFP 04-148, and TPP 05-179). The authors thank providers and veterans from the Richard L. Roudebush VA Medical Center and providers and consumers from the Midtown Mental Health Center in Indianapolis. They also appreciate the assistance of their research team who helped with the project. The views expressed in this article are those of the authors and do not necessarily represent the views of the U.S. Department of Veterans Affairs.
The authors report no competing interests.