Bipolar disorder is associated with premature mortality, but evidence is mostly derived from Western countries. There has been no research evaluating shortened lifespan in bipolar disorder using life-years lost (LYLs), which is a recently developed mortality metric taking into account illness onset for life expectancy estimation. The current study aimed to examine the extent of premature mortality in bipolar disorder patients relative to the general population in Hong Kong (HK) in terms of standardised mortality ratio (SMR) and excess LYLs, and changes of mortality rate over time.

This population-based cohort study investigated excess mortality in 12 556 bipolar disorder patients between 2008 and 2018, by estimating all-cause and cause-specific SMRs, and LYLs. Trends in annual SMRs over the 11-year study period were assessed. Study data were retrieved from a territory-wide medical-record database of HK public healthcare services.

Patients had higher all-cause [SMR: 2.60 (95% CI: 2.45–2.76)], natural-cause [SMR: 1.90 (95% CI: 1.76–2.05)] and unnatural-cause [SMR: 8.63 (95% CI: 7.34–10.03)] mortality rates than the general population. Respiratory diseases, cardiovascular diseases and cancers accounted for the majority of deaths. Men and women with bipolar disorder had 6.78 (95% CI: 6.00–7.84) years and 7.35 (95% CI: 6.75–8.06) years of excess LYLs, respectively. The overall mortality gap remained similar over time, albeit slightly improved in men with bipolar disorder.

Bipolar disorder is associated with increased premature mortality and substantially reduced lifespan in a predominantly Chinese population, with excess deaths mainly attributed to natural causes. Persistent mortality gap underscores an urgent need for targeted interventions to improve physical health of patients with bipolar disorder.

To examine: (1) gender-specific determinants of help-seeking for mental health, including health professional consultation and the use of non-clinical support services and self-management strategies (SS/SM) and; (2) gender differences among individuals with unmet perceived need for care.

Analyses focused on 689 males and 1075 females aged 16–85 years who met ICD-10 criteria for a past-year affective, anxiety or substance use disorder in an Australian community-representative survey. Two classifications of help-seeking for mental health in the previous year were created: (1) no health professional consultation or SS/SM, or health professional consultation, or SS/SM only, and; (2) no general practitioner (GP) or mental health professional consultation, or GP only consultation, or mental health professional consultation. Between- and within-gender help-seeking patterns were explored using multinomial logistic regression models. Characteristics of males and females with unmet perceived need for care were compared using chi-square tests.

Males with mental or substance use disorders had relatively lower odds than females of any health professional consultation (adjusted odds ratio [AOR] = 0.46), use of SS/SM only (AOR = 0.59), and GP only consultation (AOR = 0.29). Notably, males with severe disorders had substantially lower odds than females of any health professional consultation (AOR = 0.29) and GP only consultation (AOR = 0.14). Most correlates of help-seeking were need-related. Many applied to both genders (e.g., severity, disability, psychiatric comorbidity), although some were male-specific (e.g., past-year reaction to a traumatic event) or female-specific (e.g., past-year affective disorder). Certain enabling and predisposing factors increased the probability of health professional consultation for both genders (age 30+ years) or for males (unmarried, single parenthood, reliance on government pension). Males with unmet perceived need for care were more likely to have experienced a substance use disorder and to want medicine or tablets or social intervention, whereas their females peers were more likely to have experienced an anxiety disorder and to want counselling or talking therapy. For both genders, attitudinal/knowledge barriers to receiving the types of help wanted (e.g., not knowing where to get help) were more commonly reported than structural barriers (e.g., cost).

Findings suggest a need to address barriers to help-seeking in males with severe disorders, and promote GP consultation. Exploring gender-specific attitudinal/knowledge barriers to receiving help, and the types of help wanted, may assist in designing interventions to increase consultation. Mental health promotion/education efforts could incorporate information about the content and benefits of evidence-based treatments and encourage males to participate in other potentially beneficial actions (e.g., physical activity).

To compare trends in the estimated prevalence of mood and/or anxiety disorders identified from two data sources (self-report and administrative). Reviewing, synthesising and interpreting data from these two sources will help identify potential factors that underlie the observed estimates and inform public health action.

We used self-reported, diagnosed mood and/or anxiety disorder cases from the Canadian Community Health Survey (CCHS) across a 5-year span (from 2003 to 2009) to estimate the prevalence among the Canadian population aged ≥15 years. We also estimated the prevalence of mood and/or anxiety disorders using the Canadian Chronic Disease Surveillance System (CCDSS), which identified cases using ICD-9/-10-CA codes from physician billing claims and hospital discharge records during the same time period. The prevalence rates for mood and/or anxiety disorders were compared across the CCHS and CCDSS by age and sex for all available years of data from 2003 to 2009. Summary rates were age-standardised to the Canadian population as of 1 October 1991.

In 2009, the prevalence of mood and/or anxiety disorders was 9.4% using self-reported data v. 11.3% using administrative data. Prevalence rates obtained from administrative data were consistently higher than those from self-report for both men and women. However, due to an increase in the prevalence of self-reported cases, these differences decreased over time (rate ratios for both sexes: 1.6–1.2). Prevalence estimates were consistently higher among females compared with males irrespective of data source. While differences in the prevalence estimates between the two data sources were evident across all age groups, the reduction of these differences was greater among adolescent, young and middle-aged adults compared with those 70 years and older.

The overall narrowing of differences over time reflects a convergence of information regarding the prevalence of mood and/or anxiety disorders trends between self-report and administrative data sources. While the administrative data-based prevalences remained relatively stable, the self-reported prevalences increased over time. These observations may reflect positive societal changes in the perceptions of mental health (declining stigma) and/or increasing mental health literacy. Additional research using non-ecological data is required to further our understanding of the observed findings and trends, including a data linkage exercise permitting a comparison of prevalence estimates and population characteristics from these two data sources both separately and merged.

People with severe mental illness (SMI) have a high risk of living socially excluded from the mainstream society. Policy initiatives and health systems aim to improve the social situation of people who suffer from mental health disabilities. The aim of this study was to explore the extent of social exclusion (employment and income, social network and social activities, health problems) of people with SMI in Switzerland.

Data from the Swiss Health Survey 2012 were used to compare the social exclusion magnitude of people with SMI with those suffering from severe physical illness, common mental illness and the general population.

With the exception of Instrumental Activities of Daily Living, we found a gradient of social exclusion that showed people with SMI to be more excluded than the comparison groups. Loneliness and poverty were widespread among people with SMI. Logistic regression analyses on each individual exclusion indicator revealed that people with SMI and people with severe physical illness were similarly excluded on many indicators, whereas people with common mental illness and the general population were much more socially included.

In contrast to political and health system goals, many people with SMI suffer from social exclusion. Social policy and clinical support should increase the efforts to counter exclusionary trends, especially in terms of loneliness and poverty.

Cross-national population data from the WHO World Mental Health surveys are used to compare role attainments and role impairments associated with binge-eating disorder (BED) and bulimia nervosa (BN).

Community surveys assessed 23 000 adults across 12 countries for BED, BN and ten other DSM-IV mental disorders using the WHO Composite International Diagnostic Interview. Age-of-onset was assessed retrospectively. Ten physical disorders were assessed using standard conditions checklists. Analyses examined reciprocal time-lagged associations of eating disorders (EDs) with education, associations of early-onset (i.e., prior to completing education) EDs with subsequent adult role attainments and cross-sectional associations of current EDs with days of role impairment.

BED and BN predicted significantly increased education (females). Student status predicted increased risk of subsequent BED and BN (females). Early-onset BED predicted reduced odds of current (at time of interview) marriage (females) and reduced odds of current employment (males). Early-onset BN predicted increased odds of current work disability (females and males). Current BED and BN were both associated with significantly increased days of role impairment (females and males). Significant BED and BN effects on adult role attainments and impairments were explained by controls for comorbid disorders.

Effects of BED on role attainments and impairments are comparable with those of BN. The most plausible interpretation of the fact that these associations are explained by comorbid disorders is that causal effects of EDs are mediated through secondary disorders. Controlled treatment effectiveness studies are needed to trace out long-term effects of BED–BN on secondary disorders.

Unmet needs for mental health treatment are large and widespread, and periods of economic crisis may increase the need for care and the treatment gap, with serious consequences for individuals and society. The aim of this systematic review was to summarise the empirical evidence on the association between periods of economic crisis and the use of mental health care.

Following the PRISMA statement, MEDLINE, Embase, Scopus, Open Grey and Cochrane Database were searched for relevant publications, published between 1990 and 2018, from inception to June 2018. Search terms included (1) economic crisis, (2) use of mental health services and (3) mental health problems. Study selection, data extraction and the assessment of study quality were performed in duplicate.

Seventeen studies from different countries met the inclusion criteria. The results from the included articles suggest that periods of economic crisis might be linked to an increase of general help sought for mental health problems, with conflicting results regarding the changes in the use of specialised psychiatric care. The evidence on the use of mental health care specifically due to suicide behaviour is mixed. The results also suggest that economic crises might be associated with a higher use of prescription drugs and an increase in hospital admissions for mental disorders.

Research on the impact of economic crises on the use of mental health care is scarce, and methodologies of the included papers are prone to substantial bias. More empirical and long-term studies on this topic are needed, in order to adapt mental health care systems to the specific needs of the population in times of economic crisis.

In observational studies, type-2 diabetes is associated with increased risk of dementia; however, the causal nature of this association remains unanswered. In an unselected nationwide study of all Danes, we wanted to test whether type-2 diabetes is associated with dementia subtypes, and to test whether potential associations are of a causal nature.

In the current study of nationwide observational registry data in all Danes above the age of 65 years (n = 784 434) combined with genetic consortia data on 213 370 individuals, we investigated the associations between type-2 diabetes and Alzheimer's disease, vascular dementia, unspecified dementia and all-cause dementia, and whether observational associations were of a causal nature by applying a two-sample Mendelian randomisation strategy. We addressed key biases inherent in Mendelian randomisation approaches.

Important confounders (age, ethnicity, size of community, region, civil status and education level) were captured on all 784 434 individuals and adjusted for in the models. Multifactorial adjusted hazard ratios were 1.13 (1.06–1.21) for Alzheimer's disease, 1.98 (1.83–2.14) for vascular dementia, 1.53 (1.48–1.59) for unspecified dementia and 1.48 (1.44–1.53) for all-cause dementia in persons with type-2 diabetes v. without. Results were similar for men and women. The two-sample Mendelian randomisation estimate for the association between the genetic instrument and Alzheimer's disease was 1.04 (0.98–1.10), consistent with sensitivity estimates, addressing pleiotropy, measurement bias and weak instrument bias.

In a nationwide study of all Danes above the age of 65 years, we show that type-2 diabetes is associated with major subtypes of dementia – with particularly strong associations for vascular dementia and unspecified dementia – the two types of dementia with the most obvious vascular pathologies. Although the present two-sample Mendelian randomisation approach using genetic consortia data suggests that type-2 diabetes is not a direct cause of Alzheimer's disease, we were unable to test the causal nature of type-2 diabetes for vascular dementia and unspecified dementia, because no publicly available genetic consortia data yet exist for these dementia endpoints. The causal nature of type-2 diabetes for dementia with vascular pathologies is pivotal questions to solve for future public health recommendations and therapeutic advice.

Literature suggests an association between loneliness and mortality for both males and females. Yet, the linkage of loneliness to mortality is not thoroughly examined, and need to be replicated with a long follow-up time. This study assessed the association between loneliness and mortality, including associations to gender, in 1363 adult swedes.

This community-based prospective cohort study from the Swedish Lundby Study included 1363 individuals of whom 296 individuals (21.7%) were identified as lonely with use of semi-structured interviews in 1997. The cohort was followed until 2011 and survival analyses were used to estimate the relative risk of death.

Death occurred with an incidence rate of 2.63 per 100 person-years and 2.09 per 100 person-years for lonely and non-lonely individuals, respectively. In crude analysis, loneliness was associated with a significant increased mortality risk of 27% compared with non-lonely individuals [hazard ratio (HR) 1.27; 95% CI 1.01–1.60]. Unadjusted, lonely females had a significant increased risk (HR 1.76; 95% CI 1.31–2.34) and adjusted insignificant increased mortality risk of 27% (HR 1.27; 95% CI 0.92–1.74), compared with non-lonely females. Lonely males were found to have an adjusted significant decreased risk of mortality (HR 0.50; 95% CI 0.32–0.80), compared with non-lonely males.

Findings suggest an association between loneliness and increased risk of mortality and that gender differences may exist, which have not been previously reported. If replicated, our results indicate that loneliness may have differential physical implications in some subgroups. Future studies are needed to further investigate the influence of gender on the relationship.

In recent years several warnings have been issued by regulatory authorities on the risk of electrocardiogram abnormalities in individuals exposed to psychotropic drugs. As a consequence of these warnings, monitoring of the QT interval corrected for heart rate (QTc) has become increasingly common. This study was conducted to measure the frequency of QTc prolongation in unselected psychiatric patients, and to document the associated factors using a cross-sectional approach.

The study was carried out in 35 Italian psychiatric services that are part of the STAR (Servizi Territoriali Associati per la Ricerca) Network, a research group established to produce scientific knowledge by collecting data under ordinary circumstances. During a three-month period, a consecutive unselected series of both in- and out-patients were enrolled if they performed an ECG during the recruitment period and were receiving psychotropic drugs on the day ECG was recorded.

During the recruitment period a total of 2411 patients were included in the study. The prevalence of QTc prolongation ranged from 14.7% (men) and 18.6% (women) for the cut-off of 450 ms, to 1.26% (men) and 1.01% (women) for the cut-off of 500 ms. In the multivariate model conducted in the whole sample of patients exposed to psychotropic drugs, female sex, age, heart rate, alcohol and/or substance abuse, cardiovascular diseases and cardiovascular drug treatment, and drug overdose were significantly associated with QTc prolongation. In patients exposed to antipsychotic drugs, polypharmacy was positively associated with QTc prolongation, whereas use of aripiprazole decreased the risk. In patients exposed to antidepressant drugs, use of citalopram, citalopram dose and use of haloperidol in addition to antidepressant drugs, were all positively associated with QTc prolongation.

The confirmation of a link between antipsychotic polypharmacy and QTc prolongation supports the current guidelines that recommend avoiding the concurrent use of two or more antipsychotic drugs, and the confirmation of a link between citalopram and QTc prolongation supports the need for routine QTc monitoring. The relatively low proportion of patients with QTc prolongation not only suggests compliance with current safety warnings issued by regulatory authorities, but also casts some doubts on the clinical relevance of QTc prolongation related to some psychotropic drugs.

Poor insight is prevalent in patients with schizophrenia and has been associated with acute illness severity, medication non-adherence and poor treatment outcomes. Paradoxically, high insight has been associated with various undesirable outcomes, including low self-esteem, depression and low subjective quality of life (QoL) in patients with schizophrenia. Despite the growing body of studies conducted in Western countries supporting the pernicious effects of improved insight in psychosis, which bases on the level of self-stigma, the effects are unclear in non-Western societies. The current study examined the role of self-stigma in the relationship between insight and psychosocial outcomes in a Chinese population.

A total of 170 outpatients with schizophrenia spectrum disorders were recruited from two general university hospitals. Sociodemographic data and clinical variables were recorded and self-report scales were employed to measure self-stigma, depression, insight, self-esteem and subjective QoL. Structural equation modelling (SEM) was used to analyse the cross-sectional data.

High levels of self-stigma were reported by 39% of the participants (n = 67). The influences of insight, self-stigma, self-esteem and depression on subjective QoL were confirmed by the SEM results. Our model with the closest fit to the data (χ2 = 33.28; df = 20; p = 0.03; χ2/df = 1.66; CFI = 0.98; TLI = 0.97; RMSEA = 0.06) demonstrated that self-stigma might fully mediate the association of insight with low self-esteem, depression and poor subjective QoL. High insight into illness contributed to self-stigma, which caused low self-esteem and depression and, consequently, low QoL. Notably, insight did not directly affect self-esteem, depression or QoL. Furthermore, the association of insight with poor psychosocial outcomes was not moderated by self-stigma.

Our findings support the mediating model of insight relevant to the poor psychosocial outcomes of individuals diagnosed with schizophrenia in non-Western societies, in which self-stigma plays a pivotal role. These findings elucidate the direct and indirect effects of insight on psychosocial outcomes and imply that identifying and correcting self-stigma in people with schizophrenia could be beneficial. Additional studies are required to identify whether several other neurocognitive or psychosocial variables mediate or moderate the association of insight with self-esteem, depression and QoL in patients with schizophrenia. Studies with detailed longitudinal assessments are necessary to confirm our findings.

The clinical outcomes of individuals at clinical high risk of psychosis (CHR-P) who do not transition to psychosis are heterogeneous and inconsistently reported. We aimed to comprehensively evaluate longitudinally a wide range of outcomes in CHR-P individuals not developing psychosis.

“Preferred Reporting Items for Systematic reviews and Meta-Analyses” and “Meta-analysis Of Observational Studies in Epidemiology”-compliant meta-analysis (PROSPERO: CRD42021229212) searching original CHR-P longitudinal studies in PubMed and Web of Science databases up to 01/11/2021. As primary analysis, we evaluated the following outcomes within CHR-P non-transitioning individuals: (a) change in the severity of attenuated psychotic symptoms (Hedge's g); (b) change in the severity of negative psychotic symptoms (Hedge's g); (c) change in the severity of depressive symptoms (Hedge's g); (d) change in the level of functioning (Hedge's g); (e) frequency of remission (at follow-up). As a secondary analysis, we compared these outcomes in those CHR-P individuals who did not transition vs. those who did transition to psychosis at follow-up. We conducted random-effects model meta-analyses, sensitivity analyses, heterogeneity analyses, meta-regressions and publication bias assessment. The risk of bias was assessed using a modified version of the Newcastle-Ottawa Scale (NOS).

Twenty-eight studies were included (2756 CHR-P individuals, mean age = 20.4, 45.5% females). The mean duration of follow-up of the included studies was of 30.7 months. Primary analysis: attenuated psychotic symptoms [Hedges’ g = 1.410, 95% confidence interval (CI) 1.002–1.818]; negative psychotic symptoms (Hedges’ g = 0.683, 95% CI 0.371–0.995); depressive symptoms (Hedges’ g = 0.844, 95% CI 0.371–1.317); and functioning (Hedges’ g = 0.776, 95% CI 0.463–1.089) improved in CHR-P non-transitioning individuals; 48.7% remitted at follow-up (95% CI 39.3–58.2%). Secondary analysis: attenuated psychotic symptoms (Hedges’ g = 0.706, 95% CI 0.091–1.322) and functioning (Hedges’ g = 0.623, 95% CI 0.375–0.871) improved in CHR-P individuals not-transitioning compared to those transitioning to psychosis, but there were no differences in negative or depressive symptoms or frequency of remission (p > 0.05). Older age was associated with higher improvements of attenuated psychotic symptoms (β = 0.225, p = 0.012); publication years were associated with a higher improvement of functioning (β = −0.124, p = 0.0026); a lower proportion of Brief Limited Intermittent Psychotic Symptoms was associated with higher frequencies of remission (β = −0.054, p = 0.0085). There was no metaregression impact for study continent, the psychometric instrument used, the quality of the study or proportion of females. The NOS scores were 4.4 ± 0.9, ranging from 3 to 6, revealing the moderate quality of the included studies.

Clinical outcomes improve in CHR-P individuals not transitioning to psychosis but only less than half remit over time. Sustained clinical attention should be provided in the longer term to monitor these outcomes.

There is limited longitudinal research that has looked at the longer term incidence of depressive symptoms, comparing women with a hysterectomy to women without a hysterectomy. We aimed to investigate the association between hysterectomy status and the 12-year incidence of depressive symptoms in a mid-aged cohort of Australian women, and whether these relationships were modified by use of exogenous hormones.

We used generalised estimating equation models for binary outcome data to assess the associations of the incidence of depressive symptoms (measured by the 10-item Centre for Epidemiologic Studies Depression Scale) across five surveys over a 12-year period, in women with a hysterectomy with ovarian conservation, or a hysterectomy with bilateral oophorectomy compared with women without a hysterectomy. We further stratified women with hysterectomy by their current use of menopausal hormone therapy (MHT). Women who reported prior treatment for depression were excluded from the analysis.

Compared with women without a hysterectomy (n = 4002), both women with a hysterectomy with ovarian conservation (n = 884) and women with a hysterectomy and bilateral oophorectomy (n = 450) had a higher risk of depressive symptoms (relative risk (RR) 1.20; 95% confidence interval (CI) 1.06–1.36 and RR 1.44; 95% CI 1.22–1.68, respectively). There were differences in the strength of the risk for women with a hysterectomy with ovarian conservation, compared with those without, when we stratified by current MHT use. Compared with women without a hysterectomy who did not use MHT, women with a hysterectomy with ovarian conservation who were also MHT users had a higher risk of depressive symptoms (RR 1.57; 95% CI 1.31–1.88) than women with a hysterectomy with ovarian conservation but did not use MHT (RR 1.17; 95% CI 1.02–1.35). For women with a hysterectomy and bilateral oophorectomy, MHT use did not attenuate the risk. We could not rule out, however, that the higher risk seen among MHT users may be due to confounding by indication, i.e. MHT was prescribed to treat depressive symptoms, but their depressive symptoms persisted.

Women with a hysterectomy (with and without bilateral oophorectomy) have a higher risk of new incidence of depressive symptoms in the longer term that was not explained by lifestyle or socio-economic factors.

Self-reported psychotic experiences (SRPE) by individuals from the general population are often unconfirmed by clinical interview and referred to as ‘false-positive’ (FP) SRPE. FP SRPE have been suggested to represent the mildest form of risk along the extended psychosis continuum. However, little is known about their (clinical) outcome and evolution over time. Aims of this study were to prospectively examine, in individuals with FP SRPE, (1) the prevalence of remission, persistence and transition to validated PE at 3-year follow-up; (2) potential baseline psychopathological and psychosocial predictors of persistence of FP SRPE and transition to validated PE; and (3) whether those with persistent FP SRPE and validated PE already differed on psychopathology and psychosocial factors at baseline. We tested the hypotheses that (i) individuals with FP SRPE would be more likely to have SRPE and validated PE at follow-up; and (ii) that FP SRPE would be predictive of lower functioning and more psychopathology and help-seeking behaviour at follow-up.

Baseline (n = 6646) and 3-year follow-up (n = 5303) data of the second the Netherlands Mental Health Survey and Incidence Study (NEMESIS-2), a general population research project on prevalence, incidence, course and consequences of psychiatric disorders was used. Self-report of PE was followed by clinical interview to determine clinical validity. The presence of mood, anxiety and substance use disorders, childhood adversity, help-seeking and functioning as well as PE characteristics (number, frequency, distress and impact) were used in the analyses which included only individuals with complete data for both assessments waves (n = 4683).

At baseline, 454 participants had any FP SRPE; of these 372 participants had complete follow-up data available. Those with baseline FP SRPE were significantly more likely to report SRPE (OR = 3.58; 95% CI 2.38–5.40, p < 0.001) and validated PE (OR = 6.26; 95% CI 3.91–10.02, p < 0.001) at follow-up. Baseline FP SRPE also predicted the presence of mood and anxiety disorders, reduced functioning and help-seeking at follow-up. Several baseline psychopathological, psychosocial and PE characteristics were predictive for the persistence of SRPE. These factors also differentiated groups with FP SRPE or validated PE from those with remitted FP SRPE at follow-up.

‘FP SRPE’ are not truly ‘false’ as they index risk for the development of clinically relevant psychotic symptoms, development of mood and anxiety disorders and reduced functioning. Self-reported PE, even unconfirmed, warrant ‘watchful waiting’ and follow-up over time, especially when they are reported by individuals with reduced psychosocial functioning and general psychiatric problems.

Recovery Colleges are opening internationally. The evaluation focus has been on outcomes for Recovery College students who use mental health services. However, benefits may also arise for: staff who attend or co-deliver courses; the mental health and social care service hosting the Recovery College; and wider society. A theory-based change model characterising how Recovery Colleges impact at these higher levels is needed for formal evaluation of their impact, and to inform future Recovery College development. The aim of this study was to develop a stratified theory identifying candidate mechanisms of action and outcomes (impact) for Recovery Colleges at staff, services and societal levels.

Inductive thematic analysis of 44 publications identified in a systematised review was supplemented by collaborative analysis involving a lived experience advisory panel to develop a preliminary theoretical framework. This was refined through semi-structured interviews with 33 Recovery College stakeholders (service user students, peer/non-peer trainers, managers, community partners, clinicians) in three sites in England.

Candidate mechanisms of action and outcomes were identified at staff, services and societal levels. At the staff level, experiencing new relationships may change attitudes and associated professional practice. Identified outcomes for staff included: experiencing and valuing co-production; changed perceptions of service users; and increased passion and job motivation. At the services level, Recovery Colleges often develop somewhat separately from their host system, reducing the reach of the college into the host organisation but allowing development of an alternative culture giving experiential learning opportunities to staff around co-production and the role of a peer workforce. At the societal level, partnering with community-based agencies gave other members of the public opportunities for learning alongside people with mental health problems and enabled community agencies to work with people they might not have otherwise. Recovery Colleges also gave opportunities to beneficially impact on community attitudes.

This study is the first to characterise the mechanisms of action and impact of Recovery Colleges on mental health staff, mental health and social care services, and wider society. The findings suggest that a certain distance is needed in the relationship between the Recovery College and its host organisation if a genuine cultural alternative is to be created. Different strategies are needed depending on what level of impact is intended, and this study can inform decision-making about mechanisms to prioritise. Future research into Recovery Colleges should include contextual evaluation of these higher level impacts, and investigate effectiveness and harms.