A systematic review of service models and evidence relating to the clinically operated community-based residential mental health rehabilitation for adults with severe and persisting mental illness in Australia

Records were sought that described: (1) Australian Community Rehabilitation Units for people affected by severe and persisting mental illness, (2) the characteristics of the consumers engaging with these services, and (3) their service experiences and outcomes. Eligibility criteria for inclusion in the systematic review were a service focus on adults (18–65 years) with schizophrenia and related disorders. Rehabilitation units targeting consumers aged < 18 years or > 65 years, non-psychotic disorders (e.g. drug and alcohol, acquired brain injury, or physical rehabilitation) or specifically dual diagnosis consumers were excluded. These exclusion criteria related to the focus of the service model. There was no exclusion of consumer data from included records based on diagnosis or co-morbidity.

Databases were searched from 1995 onwards. This limit was applied as these services emerged in the context of the deinstitutionalisation process [15] with the first Community Care Unit opening in Victoria in 1996 [16]. No language specifiers were used. No exclusions were made based on study type, with emphasis placed instead on the assessment of the quality of included studies.

Parallel strategies were employed to identify relevant grey-literature using internet-based searches and published literature in academic databases (PubMed, CINAHL, PsycINFO and EmBASE).

Full details of the search strategy are provided in Additional file 1: Literature search strategy. The initial search string applied to the PubMED database is illustrative of the process undertaken: “schizophrenia” [MeSH Terms] OR “schizophrenia” [All Fields]) AND (“rehabilitation” [Subheading] OR “rehabilitation” [All Fields] OR “rehabilitation” [MeSH Terms]) AND residential [All Fields]) AND (“1995/01/01” [PDAT]: “3000/12/31” [PDAT]).

Final extraction of records from all databases occurred on the 08/02/2018. Records were sequentially imported into an Endnote X8 database, and duplicates were removed. Two authors (SP & GH) independently screened records for eligibility at the title and abstract level. Additional records were identified through snowballing [17], with the reference lists of records assessed at the full-text level inspected to identify relevant documents. Attempts were made to contact authors of included records to identify relevant research and documentation.

Records presenting unique quantitative and/or qualitative research data were subject to quality appraisal (full appraisals are provided in Additional file 2: Quality appraisal). Given the anticipated prominence of non-comparative methodologies in the published literature, the decision was made to rely on methodologically targeted quality appraisal instruments. Relevant checklists from the National Heart, Lung and Blood Institute (NIH) were used in the appraisal of ‘observational cohort and cross-sectional studies’, ‘systematic review and meta-analyses’, ‘case series studies’, ‘before-after (pre-post) studies with no control groups’, and ‘observational cohort and cross-sectional studies’ [18]. Studies presenting qualitative data, and the qualitative components of mixed-methods studies were assessed using the CASP Qualitative Checklist [19]. As the CASP Qualitative Checklist does not produce a global quality rating, the tripartite global rating system (Good, Fair, Poor) used in the NIH checklists was also applied to these records. Two of the authors (SP & GH), completed the quality assessments independently, and a final rating was determined following discussion to reach consensus. For studies where SP was one of the authors, DS and GH completed the quality appraisal.

Data extraction was completed independently by two authors (SP & GH). Discrepancies were resolved through discussion to reach consensus. Relevant content was extracted into matrices. Data of interest included descriptive information outlining service models (e.g. physical environment, philosophy of care, and ‘treatment and support’), service user characteristics (e.g. age, sex, diagnostic information, and chlorpromazine dose equivalence), and quantitative and qualitative research findings (e.g. outcomes, comparison with other services, consumer and staff perspectives). Descriptive data about service models were synthesised into a domains-based classification system [20], with the emergent typology derived from key emphases in the descriptive content. The typology was then used to organise the results into meaningful groups for interpretation.

Descriptive data regarding service user characteristics was pooled across studies, with means for service types and the associated models subsequently derived. Where data was available across both service types, differences in frequency of dichotomous variables was statistically compared using Chi-square or Fisher’s exact test. Fisher’s exact test was applied when the total sample size for pooled data was < 1000 cases [21] or if the frequency in any of the 2 × 2 cells was less than five [22]. A p ≤ .01 significance level was adopted to accommodate for the multiple comparisons and reduce the risk of Type 1 error. The absence of consistent documentation of standard deviations/standard errors for continuous variables (age and chlorpromazine dose equivalence) prohibited statistical comparison of the pooled means for these variables.

In relation to outcome data, no statistical synthesis was pre-specified or completed for the quantitative data. Findings from quantitative and qualitative studies were synthesised and tabulated.

From the 24 records providing service descriptions, two types of Community Rehabilitation Units were identified– the original Community-Based Residential Care (C-BRC) type operating from the mid-1990s until the early 2000s, and the more recent Transitional Residential Rehabilitation (TRR) type (see Table 1 & Additional file 3: Service types and models). These types and the associated named service models differed in the physical environment, philosophy of care, and the available treatment/support.

C-BRC service models emerged to meet the needs of institutionalised people with severe mental illness, predominantly schizophrenia, transitioning to community-based care. Service models associated with the C-BRC type were the Community Residences in New South Wales [23] and Community Care Units in Victoria [15, 24, 25]. These services focused on the provision of accommodation in a cluster housing configuration, in addition to rehabilitation and support to people affected by severe and persisting mental illness. Key features were the availability of 24-h clinical support and an individualised treatment focus on living skills development and community integration. Group-based therapies/programmes were explicitly de-emphasised under the original Community Care Unit model [26] and were not an essential feature of Community Residences. While C-BRC services were initially intended to provide a permanent residence, by the start of the twenty-first century their focus shifted towards a more transitional model of support [24, 27, 28].

The TRR type is distinct from C-BRC in its emphasis on the transitional nature of rehabilitation support as well as the inclusion of ‘recovery’ as part of the philosophy of care. The shift to a transitional focus coincided with: the initial cohort consumers in C-BRC services not requiring 24-h support on an ongoing basis [23, 24, 27]; most of the initial C-BRC-Community Care Unit cohort transitioning to more independent settings in the community [24, 27]; and an emergent emphasis on recovery-oriented care in mental health policy and planning frameworks [10, 29, 30]. Service models aligning with the TRR type are the Community Care Units in Queensland and Victoria [10], Community Rehabilitation Centres in South Australia [31], the Community Recovery Program in Victoria [32], and Hawthorne House in Western Australia [33]. Cluster housing remains the most common environmental configuration for TRR type services, and a focus on independent living or dual occupancy units predominates. Under TRR, there is greater emphasis on the availability of individual and group therapies/programs. Since 2014, deviations from the traditional clinical staffing configuration have emerged, including the substantial integration of peer support workers under the ‘Integrated Staffing Model’ and partnerships with Non-Government Organisations [32]. These alternative staffing configurations have been linked explicitly to the goal of further realising recovery-oriented care [29, 32, 34].

Data describing consumer characteristics from 15 records were organised according to the service typology in Tables 2 and 3 (additional detail available in Additional file 4: Consumer characteristics). There were limitations in the consistency and comparability of reported data. Variables reported in 50% or more of records for both C-BRC and TRR services were: age, sex, mode of referral, the presence of an involuntary treatment order, primary diagnosis, and total chlorpromazine dose equivalence of antipsychotic medication. Across both service types, most consumers were males aged in their 30s–40s with a primary diagnosis of a schizophrenia spectrum disorder.

Differences arose in the characteristics of consumers from the C-BRC and TRR service types. TRR consumers were significantly more likely to be male (X² 19.98, p < .01) and to be referred from the community rather than an inpatient service (Fisher exact test, p < .01). Schizophrenia spectrum disorders were the most common primary diagnosis across both service types, although this was significantly less frequent within the TRR service type (X² 21.24, p < .01). TRR consumers were also significantly less likely to be subject to an involuntary treatment order (X² 106.24, p < .01). C-BRC and TRR consumers did not differ significantly with respect to the frequency of being born in Australia (Fisher exact test, p = 0.68), or rates of comorbid disorders (substance use (Fisher exact test p = 0.78), developmental disorders (Fisher exact test, p = 0.07), and physical illness (Fisher exact test, p = 0.16).

Statistical comparison of the mean values for age and chlorpromazine dose equivalence between the C-BRC and TRR groups could not be performed as standard deviations/standard errors were not reliably reported. Descriptive statistics suggest divergence between the C-BRC and TRR services, in that consumers entering the more recent TRR services were younger (pooled x̅ 35, range of available means 30–39; versus 40 years, range of available means 38–43), and on lower total chlorpromazine dose equivalence of antipsychotic medication (pooled x̅ 614, range of available means 573-728 mg/day; versus 878 mg/day, range of available means 936-1127 mg/day).

With regards to quantitative research, only six records were identified. Findings from these studies are detailed in Table 4 (with additional detail in Additional file 5: Included research). Meta-analysis and statistical assessment of publication bias and sensitivity analysis were not appropriate given the insufficient number of comparable studies [35]. Follow-up outcome data was available from two C-BRC studies assessed as ‘fair’ quality. These studies found most consumers remained in the community at long-term follow-up [23, 24], with one study noting high levels of ongoing disability [24]. One ‘poor’ quality report presented follow-up outcome data for the TRR type; the favourable pre-post outcomes with respect to reduced inpatient bed-days and symptoms observed in this study should be interpreted cautiously [31].

Twelve records presented qualitative research findings that considered a range of stakeholder perspectives (consumer, staff and family). Findings are detailed in Table 5 (see also Additional file 5: Included research), and the findings relating to consumer experiences, which reflect the majority focus of the qualitative research, is additionally summarised in text. Consumer reflections about their expectations and experience of care at Community Rehabilitation Units were consistently positive, regardless of service type and name. Two studies explored long-term follow-up in C-BRC services. These studies described ongoing challenges faced by consumers, including impoverished social networks [23, 24] and accommodation instability [24]. All records presenting qualitative findings relating to C-BRC services were assessed to be of poor quality. Seven of the nine records considering TRR models focused on Community Care Units, which were of mixed quality (Good = 4, Fair = 1, Poor = 2); the remaining two records from non-Community Care Unit settings were assessed to be of poor quality.

The typology enhances understanding of the nature and function of Australian Community Rehabilitation Units for people affected by severe and persisting mental illness. Within the Australian context the typology identified potential bases of service non-comparability, whereby even identically named services were discrepant from each other. The matrix conceptually grouping the key emphases of services models supports the appropriateness of comparing extant Australian service models in future research. Additionally, differential features associated with sub-types and specific service models will facilitate the exploration of differences in service outcomes that may emerge. A relevant example of this may be the impact of novel staffing models being trialled, such as the integration of peer support work within the multidisciplinary team and non-government organisation partnerships.

The availability of domains-based classification presents an opportunity to synthesise existing research internationally, and for future research collaboration. By identifying the key emphases of Australian service models, the typology will allow identification of the extent to which these models are similar, and the appropriateness of comparing and synthesising current and future quantitative and qualitative research findings through meta-analytic methods. One potential opportunity for comparison is the overlap identified between the two Australian Community Rehabilitation Unit types, and two of the five types of supported accommodation emerging in the recent English classification system of supported accommodation - ‘The Simple Taxonomy for Supported Accommodation’ (STAX-SA) [36]. C-BRC aligns with ‘Type 1,’ which is characterised by ‘staff on site’, ‘high support’, ‘limited emphasis on move-on’ and ‘congregate setting’. TRR aligns with ‘Type 2,’ which is distinguished from ‘Type 1’ by its ‘strong emphasis on move on’. This alignment suggests the appropriateness of comparing Australian TRR services and English services classified as ‘Type 1’ in future research.

The limitations in research exploring the outcomes for TRR consumers are concerning given the increasing investment in these services [1, 10, 12] and ongoing adaptation of service models in Australia [34]. There continues to be a strong mental health policy emphasis on the provision of community-based alternatives to inpatient psychiatric services [37, 38]. The findings of this systematic review support concerns that this policy drive is uncoupled from the evidence base [9]. Multiple stakeholders and outcomes are relevant to understanding whether these services are achieving their purpose and offering value for money. Important outcome considerations relating to consumers include functional recovery (e.g. gains in employment, accommodation stability, and living skills), clinical recovery (e.g. improvement on symptomatic measures), as well as consumer defined recovery and quality of life. The outcome of service provision on carer burden is another relevant consideration. Outcomes of relevance to service planners include reductions in inpatient psychiatric bed days, emergency department presentations, and other indexes of service utilisation. None of these outcomes are adequately addressed in the literature.

The consistent finding of positive expectations and reflections on the experience of Community Rehabilitation Units by consumers across qualitative studies provides some support for the value and acceptability of these services. Additionally, available research exploring consumer’s commencement expectations of TRR services found they understood the model [39, 40], and most were actively involved in the decision to come [40]. These findings counter the international trend toward de-emphasising and de-valuing clinically focused TRR in favour of an emphasis on the provision of permanent housing and the mobilisation of relevant support around the consumer [6‐8]. However, the lack of follow-up qualitative and quantitative research examining the associated short- and long-term outcomes for Community Rehabilitation Units remains problematic. There are significant risks associated with increased investment and adaptation of service models that remain largely untested.

This review deliberately focused on Community Rehabilitation Units for people affected by severe and persisting mental illness in the Australian context. The review did not consider other types of accommodation and support that may be provided to this group, including substance use rehabilitation, non-clinical psychosocial rehabilitation services delivered by non-government organisations, and non-rehabilitation focussed supported accommodation. While these alternative residential services are outside the scope of the present review, a comprehensive understanding of the service array is critical for answering the complex questions of what works for whom, under what circumstances, and why.

The two defined service types (C-BRC and TRR) reflect changes in the approach to these services over time, rather than two distinct types that are currently operating. This limits the utility of the C-BRC type to efforts directed towards understanding historical aspects of care.

While the process of pooling service user data facilitated statistical comparison between the C-BRC and TRR types, there are several limitations associated with this approach. At least partial overlap of consumer characteristic data is anticipated due to the combination of cross-sectional and service commencement data covering overlapping time periods. Additionally, the use of cross sectional data may create bias towards more severely impaired consumers who received longer durations of residential care.

Significant gaps and methodological limitations in the literature were identified, particularly with regards to qualitative research for C-BRC services and quantitative research for TRR services. There were no randomised controlled trials, and naturalistic observational designs predominated. Given the socially complex nature of these services the absence of studies involving randomisation, blinding and controls was anticipated. Moreover, there are ethical and practical barriers to attempting randomized controlled trials of such services [41].