Peer Support for People With Severe Mental Illness
Peer Support for People With Severe Mental Illness
We evaluated the effects of peer-provided interventions on objective outcomes including hospitalisation and employment, and on self-reported outcomes including symptoms of mental health problems, quality of life, recovery, hope, empowerment, and satisfaction with services. The review protocol was pre-specified. This review of previously reported studies required no ethical approval or additional consent from participants.
Types of Trials. Randomised controlled trials (RCTs) including cluster RCTs and factorial RCTs were included. Published and unpublished trials were eligible.
Types of Participants. Studies were included if participants were adults with severe mental illness. We included participants with schizophrenia spectrum or bipolar disorder, or studies with mixed populations of people using secondary mental health services. We excluded studies including only participants with unipolar depression or personality disorders. Peers were using or had used secondary mental health services.
Types of Interventions. Included interventions were community-based peer support designed to facilitate recovery from severe mental illness. We included studies of peer support in addition to other interventions if the effect of peer support could be isolated. We excluded: residential and inpatient peer-run programmes; peer support programmes focusing exclusively on areas other than overall mental health recovery (e.g. employment, physical health or drug and alcohol use); and interventions led by mental health professionals.
When studies included more than one eligible intervention, we combined them for analysis. When studies included multiple comparison groups, we included all comparisons that allowed us to isolate the effects of peer-provided interventions (e.g. peer support with treatment as usual (TAU) compared with TAU) or that directly compared peer support with another intervention (e.g. peer-led versus professionally-led services). Other groups were not analysed (see Additional file 1).
We searched the Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, Embase, Medline, preMedline, and PsycINFO from inception to January 2013, combining synonyms for: severe mental illness; peer support; and randomised controlled trial, using the AND command (see Additional file 1). The search was updated in July 2013 as part of a broader search for trials of interventions for psychosis. Reference lists of reviews, included studies, and excluded studies were searched for additional citations. We contacted study authors and other experts. Two authors independently screened abstracts (BH and HI) and resolved differences with a third author (BLE).
Studies were assessed using the Cochrane Collaboration Risk of Bias Tool. Two authors rated each study for risk of bias due to: sequence generation; allocation concealment; blinding of participants, assessors and providers; selective outcome reporting; and incomplete data. Risk of bias for each domain was rated as high (seriously weakens confidence in the results), low (unlikely to seriously alter the results) or unclear. Discrepancies were resolved through discussion.
Data were extracted independently by two reviewers (HI or BLE, and BH). We extracted data regarding outcomes at all-time points, study characteristics (setting, number randomised, duration), inclusion criteria and participant demographics, and characteristics of the interventions. We also contacted all authors to request missing data on participant characteristics and outcomes, and to enquire about unpublished studies.
For continuous outcomes, we calculated the standardised mean difference, Hedges g, and weighted studies using the inverse of variance. For dichotomous outcomes, we calculated risk ratios (RR) and combined studies using the Mantel-Haenszel method. All outcomes are reported with 95% confidence intervals (CI) using random-effects models.
Missing data were noted for each outcome. When dropout was not reported, we contacted the authors. When analyses were reported for completers as well as controlling for dropout (for example, imputed using regression methods), we used the latter.
Statistical heterogeneity was assessed by visual inspection of forest plots, by performing the Chi test (assessing the P value) and by calculating the I statistic, which describes the percentage of observed heterogeneity that would not be expected by chance. If the p value was less than 0.10 and I exceeded 50%, we considered heterogeneity to be substantial. When subgroup analyses were conducted, differences between groups were tested using Chi. Meta-analysis was conducted using RevMan and a summary of results was prepared using the GRADE system, which is a structured assessment of confidence in the evidence for individual outcomes attending: threats to internal validity, inconsistency, indirectness, imprecision, and reporting bias. For example, results with I > 50% were downgraded for study quality. The GRADE system rates confidence in the evidence from each analysis of pooled data as high, moderate, low or very low.
Methods
We evaluated the effects of peer-provided interventions on objective outcomes including hospitalisation and employment, and on self-reported outcomes including symptoms of mental health problems, quality of life, recovery, hope, empowerment, and satisfaction with services. The review protocol was pre-specified. This review of previously reported studies required no ethical approval or additional consent from participants.
Eligibility Criteria
Types of Trials. Randomised controlled trials (RCTs) including cluster RCTs and factorial RCTs were included. Published and unpublished trials were eligible.
Types of Participants. Studies were included if participants were adults with severe mental illness. We included participants with schizophrenia spectrum or bipolar disorder, or studies with mixed populations of people using secondary mental health services. We excluded studies including only participants with unipolar depression or personality disorders. Peers were using or had used secondary mental health services.
Types of Interventions. Included interventions were community-based peer support designed to facilitate recovery from severe mental illness. We included studies of peer support in addition to other interventions if the effect of peer support could be isolated. We excluded: residential and inpatient peer-run programmes; peer support programmes focusing exclusively on areas other than overall mental health recovery (e.g. employment, physical health or drug and alcohol use); and interventions led by mental health professionals.
When studies included more than one eligible intervention, we combined them for analysis. When studies included multiple comparison groups, we included all comparisons that allowed us to isolate the effects of peer-provided interventions (e.g. peer support with treatment as usual (TAU) compared with TAU) or that directly compared peer support with another intervention (e.g. peer-led versus professionally-led services). Other groups were not analysed (see Additional file 1).
Types of Outcome Measures
Hospitalisation
Employment
Overall psychiatric symptoms
Symptoms of psychosis
Depression and anxiety
Quality of Life
Recovery (self-rated)
Hope
Empowerment
Satisfaction with services
Search Strategy
We searched the Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, Embase, Medline, preMedline, and PsycINFO from inception to January 2013, combining synonyms for: severe mental illness; peer support; and randomised controlled trial, using the AND command (see Additional file 1). The search was updated in July 2013 as part of a broader search for trials of interventions for psychosis. Reference lists of reviews, included studies, and excluded studies were searched for additional citations. We contacted study authors and other experts. Two authors independently screened abstracts (BH and HI) and resolved differences with a third author (BLE).
Assessment of Bias
Studies were assessed using the Cochrane Collaboration Risk of Bias Tool. Two authors rated each study for risk of bias due to: sequence generation; allocation concealment; blinding of participants, assessors and providers; selective outcome reporting; and incomplete data. Risk of bias for each domain was rated as high (seriously weakens confidence in the results), low (unlikely to seriously alter the results) or unclear. Discrepancies were resolved through discussion.
Data Management
Data were extracted independently by two reviewers (HI or BLE, and BH). We extracted data regarding outcomes at all-time points, study characteristics (setting, number randomised, duration), inclusion criteria and participant demographics, and characteristics of the interventions. We also contacted all authors to request missing data on participant characteristics and outcomes, and to enquire about unpublished studies.
Statistical Analysis
For continuous outcomes, we calculated the standardised mean difference, Hedges g, and weighted studies using the inverse of variance. For dichotomous outcomes, we calculated risk ratios (RR) and combined studies using the Mantel-Haenszel method. All outcomes are reported with 95% confidence intervals (CI) using random-effects models.
Missing data were noted for each outcome. When dropout was not reported, we contacted the authors. When analyses were reported for completers as well as controlling for dropout (for example, imputed using regression methods), we used the latter.
Statistical heterogeneity was assessed by visual inspection of forest plots, by performing the Chi test (assessing the P value) and by calculating the I statistic, which describes the percentage of observed heterogeneity that would not be expected by chance. If the p value was less than 0.10 and I exceeded 50%, we considered heterogeneity to be substantial. When subgroup analyses were conducted, differences between groups were tested using Chi. Meta-analysis was conducted using RevMan and a summary of results was prepared using the GRADE system, which is a structured assessment of confidence in the evidence for individual outcomes attending: threats to internal validity, inconsistency, indirectness, imprecision, and reporting bias. For example, results with I > 50% were downgraded for study quality. The GRADE system rates confidence in the evidence from each analysis of pooled data as high, moderate, low or very low.
