A Recovery-Oriented Intervention for People With Psychosis: A Pilot Randomized Controlled Trial

We report the primary outcomes of a pilot randomized controlled trial (RCT) of critical time intervention–task shifting (CTI-TS) for people with psychosis that was conducted in Santiago, Chile, and Rio de Janeiro. The pilot RCT aimed to examine results for the primary outcomes in order to evaluate the capacity for and feasibility of implementing a future multisite regional RCT in Latin America, as reported elsewhere (1–3).

Latin American countries have made remarkable strides toward improving community mental health care over the past two decades (4), and several have substantially expanded and improved mental health services over the past decade. However, progress has been uneven across countries (5, 6). CTI-TS was designed for people with psychosis at the initiation of care in community mental health centers (CMHCs), a generic term for similar facilities in Latin American countries (1). The CMHCs in Chile and Brazil offer some of the best community care in Latin America, yet they share some significant limitations of CMHCs across the region. First, minimal resources are available for residential services or services in other community settings outside the clinic (7). Second, there is limited coordination of care with other types of services, such as primary or emergency care (8). Third, people with lived experience with mental illness are rarely engaged (and almost never employed) in planning or providing mental health care (1). Fourth, although some CMHCs provide rehabilitation and psychosocial services, recovery-oriented approaches are rarely offered (1). The CTI-TS was designed to address these four limitations.

In this context, we sought to introduce a recovery-oriented approach to help patients improve their quality of life while they confront ongoing mental health problems or related disabilities. Our approach included, for example, expanding the ability of people to rebuild a positive sense of self and social identity, augmenting people’s strategies for coping with mental health issues, and helping people find ways to navigate health services (2). Our primary outcome measures reflect our overarching goal of improving the lives of those with psychosis. In particular, we chose quality of life as a primary outcome because it has been among the most evident benefits of many trials in community psychiatry and because many have argued that it should be prioritized as an outcome in clinical trials of treatments for chronic conditions (9–11).

CTI-TS has roots in the evidence-based CTI model used in high-income countries (12–14); RCTs have demonstrated that the benefits of this 9-month psychosocial intervention, implemented during a critical period of transition (e.g., from hospital to community living), persist up to 18 months following treatment. Adaptations of CTI are now widely used for a variety of transitions in numerous high-income countries (see www.criticaltime.org). CTI-TS follows the model by offering 9 months of tailored support to people with psychosis after initiation into mental health care at CMHCs, a time of transition during which service users are especially vulnerable and seek help. As noted above, our purpose was not to increase retention of services, but rather to enhance participants’ quality of life. To demonstrate the effectiveness of this time-limited intervention, however, benefits must remain evident at 18-month follow-up.

Methods

Methods have been described extensively in a previous publication (1) and are summarized here.

Settings

The trial was conducted in eight CMHCs (Santiago, N=5; Rio de Janeiro, N=3) with catchment areas that included primarily residents of low socioeconomic status and a large proportion of the population in each city (about one-fifth of the population in Santiago and one-third in Rio de Janeiro) (1). Both cities are characterized by major social and economic inequalities (1). In the poorest areas of Rio de Janeiro included in this study, there were pockets of extreme poverty, pervasive violence, and frequent civil conflicts (15).

Recruitment and Consent

All trial participants were 21–65 years old, were diagnosed as having a nonaffective or affective psychotic disorder (ICD-10 codes F20–29, F30.2, F31.2, F31.5, F32.3, F33.3) that was not drug induced, and were enrolled no longer than 3 months after their first visit to the CMHC. Among 228 potentially eligible patients, 110 (Santiago, N=60; Rio de Janeiro, N=50) gave consent and were randomly assigned and enrolled in the study (see Figure S1 in the online supplement to this article). Capacity for consent was evaluated by a CMHC psychiatrist or psychologist.

Randomization

Randomization was done after the baseline interview with the Web-based randomization assignment program provided by the Data Coordinating Center at Columbia University. Participants were randomly assigned separately within the two cities to reduce imbalances between groups. A four-digit study ID was assigned to each trial participant in order to construct a data set (without identifiable patient information or treatment allocation) that could be shared among investigators.

Intervention

CTI-TS.

CTI-TS was developed after a series of CTI studies led by the team in Brazil (15–17) and refined after pilot studies were completed by teams in Brazil and Chile. Together (18–21), these efforts comprised part of a thorough, socioculturally sensitive process of adaptation led primarily by local investigators, which is described elsewhere (1). CTI-TS was intended for use in Latin American countries. A training manual (22) specifies core elements that are applicable across the region, as well as elements that can be adapted for different countries and service systems. A novel component of CTI-TS, compared with CTI, is that care is delivered by lay community mental health workers and peer support workers (PSWs) who are based in CMHCs and supervised by mental health professionals. Community mental health workers and PSWs work together in pairs, or duplas.

CTI-TS focused on areas identified as crucial for strengthening individuals’ continuum of care and forming enduring links with their informal and formal community supports. The intervention also provided practical and emotional support during a critical period of transition when individuals were especially vulnerable. CTI-TS entails three phases: initiation, try-out, and transfer of care. Initiation involves the elaboration of a recovery-oriented plan focused on a few areas (e.g., family, CMHCs, primary care) defined and selected by the participant. Try-out considers lessening involvement of both community mental health workers and PSWs and observing how the plan made in the prior phase works in practice. Finally, transfer of care is devoted primarily to ensuring the durability of what was achieved in the previous phases. The community mental health workers and PSWs were trained on CTI-TS in two phases, first with 54 hours of didactic training and then with 9 months of on-the-job training, guided by a CTI-TS training manual (22). How CTI-TS was designed, developed, and implemented is described in more detail elsewhere (1).

Usual care.

Usual care corresponded to the care that was typically delivered at CMHCs and was similar in each city (1). CMHCs usually served a population enrolled in the public health care system, provided mental health care to adults with psychoses and other severe mental disorders, and had a multidisciplinary team of mental health professionals. Although the teams offered both psychopharmacological and psychosocial services at the clinic, they generally received limited training on evidence-based psychosocial interventions and no training on recovery orientation (23). Services provided outside the clinic were often minimal.

Assessments

Outcomes.

The primary outcomes were health-related quality of life and unmet needs at 18 months. These outcomes were selected to be consistent with our purpose of improving the quality of life of people with psychosis and were assessed at baseline and 9-month follow-up. Health-related quality of life was measured with the World Health Organization Quality of Life Assessment–Brief Version (WHOQOL-BREF) (24). This instrument has 24 items that measure four domains: physical health (e.g., items on daily activities and functional capacity), psychological health (e.g., items on mental status and self-esteem), social relationships (e.g., items on social support and sex life), and environmental health (e.g., items on living in the physical environment and transportation). We added a domain called “recreation” that comprised three questions from the original, longer WHOQOL pertaining to recreational and leisure activities (“How much are you able to relax and enjoy yourself?” “How much do you have the opportunity to do leisure activities?” and “How much do you enjoy free time?”). The WHOQOL-BREF also has two general items on quality of life: “How would you rate your quality of life?” and “How satisfied are you with your health?” (see Table S1a in the online supplement).

Unmet needs were measured with the Camberwell Assessment of Need (CAN) (25). The CAN evaluates 22 areas of need (see Table S1b in the online supplement), each of which can be rated as “no need,” “met need,” or “unmet need.” As often done in studies using the CAN (26–28), our analysis collapsed these into “unmet need” versus “no need or met need.” Both instruments were previously adapted and validated in Brazil and Chile (1). Outcome assessments were conducted by interviewers who were blind to group assignment. Although this article is limited to the primary outcomes, extensive data were also collected on secondary outcomes, as described elsewhere (1).

Baseline characteristics.

Clinician chart diagnoses at baseline identified patients with nonaffective and affective psychoses. A sociodemographic questionnaire collected standard information at baseline, including the sociodemographic covariates that were entered as potential confounders in our analyses: age (21–65 years), gender (female, male), education (no school or incomplete, some high school, high school graduate, more than high school), living situation (living alone, living with spouse/partner, living with family, other living situation), and employment status (employed, occasional work, retired/pensioner, unemployed). The Life Chart Schedule (29) was used to rate the course of psychosis (continuous, episodic, other) over the prior 9 months.

Fidelity.

The CTI-TS fidelity scale evaluated the degree to which the CTI-TS team adhered to key elements of the model (22). The scale comprises 20 items rated from 1, minimal, to 5, excellent fidelity (see Table S1c in the online supplement). Raters in Santiago (F.M.) and in Rio de Janeiro (K.R.S.) were trained and supervised by a coauthor with extensive experience in rating CTI fidelity (S.C.).

Sample Size

Because this was a pilot RCT, sample size was based on the precision (or margin of error) of our outcomes. The computation used the standard deviation reported in the literature and an intraclass correlation coefficient of 0.01, as described in detail elsewhere (1). For example, applied to the total score of the WHOQOL-BREF, the expected width of the 95% confidence interval (CI) was 6.3 (see the online supplement).

Data Analysis Plan

The analysis of primary outcomes at 18 months followed the data analysis plan published elsewhere (1). We first conducted a series of descriptive analyses, including frequency distributions and measures of central tendency and dispersion with 95% CIs. Results for the CAN had a left-skewed distribution; therefore, we transformed the data to the fourth root to facilitate parametric analysis. We then conducted an intention-to-treat analysis, with 20 imputations of missing data and a fully conditional model based on the Markov chain Monte Carlo method. We used generalized estimating equations (GEE), with observations clustered within cities. The unadjusted analysis comprised the primary test for differences between the groups; we also present analyses adjusted for potential confounders that were selected before data analysis.

For secondary longitudinal analyses, we examined trends over time in primary outcomes (from baseline to 9-month follow-up and 18-month follow-up) with three-level models and with observations at different times clustered within observations, which were in turn clustered within sites.

Ethical Considerations

The study protocol was approved by the institutional review board (IRB) at Columbia University, the local IRB at the University of Chile, the local IRB at the Federal University of Rio de Janeiro, and the Brazilian National Ethical Committee. Additionally, a data and safety monitoring board at the U.S. National Institute of Mental Health’s Center for Global Mental Health Research closely monitored the study.

Results

Baseline Characteristics

Clinical and sociodemographic baseline characteristics included in the fully adjusted analyses are presented in Table 1. The most notable difference was that fewer usual care participants (N=7) than CTI-TS participants (N=12) had more than a high school education (37% vs. 63% of those with such education, respectively), but this difference was not statistically significant (see Table S2 in the online supplement).

The 18-Month Follow-up

Among the 110 trial participants, there were four deaths (two in each group) not related to the intervention. At 18-month follow-up, 49 of 55 (89%) participants in the CTI-TS group and 47 of 55 (85%) participants in the usual care group were assessed (Figure S1 in the online supplement).

Primary Outcomes: Quality of Life and Unmet Needs

With GEE analysis, unadjusted comparisons of CTI-TS with usual care at 18-month follow-up showed no statistically significant differences (Table 2). Results from adjusted GEE models also indicated no statistically significant differences (Table 2). For the longitudinal analyses, none of the interactions (group × city, group × time, and city × time) was statistically significant for the primary outcomes (data not shown; available on request).

Sensitivity Analysis

The imputation of missing data assumes that data are missing at random. Because this assumption is difficult to meet, we also conducted a complete case analysis. Again, there were no significant differences between the groups when adjusting for baseline characteristics (see Table S3 in the online supplement).

Further Exploratory Analysis

Given the absence of differences between CTI-TS and usual care, we examined whether there were changes within each group in order to shed light on these results. Both groups improved substantially on the primary outcomes over time (see Table S4 in the online supplement). Also, we examined whether results were different for the two cities and found no differences.

Fidelity

The overall mean fidelity rating for the CTI-TS teams was 88% in Santiago and 84% in Rio de Janeiro. This observation signified that the CTI-TS teams understood and sought to apply the concepts and practices elaborated in the manual. Nonetheless, fidelity ratings were low (<50%) in both cities for items representing early initiation: “early engagement” (i.e., the worker had at least one meeting with each of the following: a participant in the home, an informal support, and a service provider), “early intensive contact” (i.e., the worker had at least five contacts with a participant, two contacts with an informal support, and two contacts with a mental health provider), and “early community-based contact” (i.e., the worker had at least two meetings outside the CMHC with a participant and with an informal support) (Table S1c in the online supplement).

Discussion

The data from this pilot RCT indicate no notable differences between the CTI-TS and usual care groups for the two primary outcomes (quality of life and unmet needs) at 18-month follow-up. Because this was a pilot RCT, the study was underpowered, but we did not observe notable trends that failed to reach statistical significance. In secondary analyses, there was no time × group interaction, which would have indicated a temporary effect during the 9-month intervention that dissipated by 18-month follow-up.

We offer three potential explanations for the lack of difference in primary outcomes between the CTI-TS and usual care groups. First, participants and their relatives were not accustomed to PSWs providing mental health services, and some participants had negative reactions to this kind of support. A qualitative study suggested that most participants were deeply inspired and grateful for having the chance to talk to “someone like me” or were neutral, but there were some who perceived PSW services as “second class” and inferior to professional services (17). Our efforts to familiarize all stakeholders with CTI-TS might not have been sufficient, especially considering the novel use of PSWs in Latin America. Second, we could not build rapport with participants and their relatives prior to the intervention because the intervention focused on people making a first contact with the CMHCs. A previous trial of CTI suggested that establishing rapport prior to randomization enhanced the intervention’s effect (12). Introducing this component would have required randomization after several visits and accepting that our sample would be limited to those who already had some engagement with the CMHCs. We decided to initiate the intervention with people making a first contact in order to include the substantial number of patients who do not show up for further visits, but in retrospect this decision may have undermined the intervention’s effectiveness. Third, our primary outcome measures may not have been sensitive enough to capture the intervention effects. The CAN probably had a floor effect—the median of unmet needs at 18 months was zero in both groups. The WHOQOL-BREF has been widely used, but assessments of quality of life are influenced by extraneous factors such as variation in participants’ reference group. However, when we initiated the trial, we conducted a thorough search of existing instruments; despite their limitations, these two were the most suitable measures available.

Previous literature reviews have indicated inconsistent evidence regarding the effectiveness of peer support for people with severe mental illness (2, 30). Our intervention was unusual—although not unique—in formally employing PSWs. The best method for introducing peer support services is not yet clear, and further work is likely needed to identify the most meaningful and helpful role of PSWs.

Both the CTI-TS and usual care groups showed marked improvement in the primary outcomes from baseline to 18-month follow-up. This improvement might reflect the high quality of public CMHC services in Chile and Brazil, further enhanced by the implementation of new primary care and mental health services and policies (especially in Brazil) during the intervention period of this trial (31, 32). Given these improvements in mental health care, some participants may not have needed the intervention. Improvement in primary outcomes might also partly reflect that many participants had recently been released from hospital or referred from emergency services when the baseline assessment was done but were residing in the community and had received varying amounts of outpatient services when the 18-month follow-up took place.

 To our knowledge, this was the first RCT of a recovery-oriented intervention for people with psychosis that was conducted in public CMHCs in Latin America. It was also the first to employ PSWs in the delivery of care, including specifying their roles and recognizing their lived experience as a form of special expertise. This pilot RCT succeeded in demonstrating the feasibility of and capacity for regional RCTs and in showing that introducing new ideas and values at CMHCs is possible (e.g., peer support, optimism, and recovery orientation) (1). A major limitation of the analysis of primary outcomes was that the study was underpowered by design for establishing efficacy. Other limitations include the preclusion of rapport because of first-contact enrollment, the mixed reaction to PSWs, and the choice of instruments with low sensitivity. Yet, these limitations also indicated ways in which the intervention might be improved.

Conclusions

We are confident that in the long run, community mental health services will continue to be developed in Latin America, despite the limitations imposed by the COVID-19 pandemic and the current political challenges. Time-limited interventions may be needed because of the lack of resources for universal long-term community supports, and people with lived experience may be important for the success of such efforts. We hope that the results of this study will spark discussion about how to better design and deliver such interventions and how to best conduct regional trials to test their efficacy.