When working as a psychiatrist people often tell me their main worries relate to things like where they’ll live, how to pay the bills, how to get out the house and meet people, or to have something meaningful to do in the day. All of these social factors can be made much more difficult by mental illness, cause significant distress, and if they aren’t addressed can contribute to a future crisis.
Ideally we’d prevent these social determinants from happening in the first place – see past Elf blogs on problems and solutions (Guo and Higson-Sweeney, 2025; Roberts and Hemming, 2025). But once social inequalities are there, how can we support people to address them?
Reviews from the last few years have looked at the evidence base for interventions to improve people’s social circumstances (Bell, 2023; Barnett et al., 2022; Killaspy et al., 2022). Both Barnett and Killaspy concluded that the strongest evidence is for IPS (Individual Placement and Support) in gaining employment, and Housing First in tackling homelessness.
Greenburgh and colleagues take a step further (Greenburgh et al., 2025), and ask what works for whom? Do these interventions work for everyone, or are some groups left behind? Are there social interventions that we haven’t explored? Does existing research apply to marginalised groups?
Without understanding these questions we risk increasing health inequalities, if some people in need of support don’t benefit from the interventions available.
Methods
The authors conducted a systematic review, which means they searched for all the existing literature on social interventions for people diagnosed with severe mental illness or common mental disorders. This included interventions to help with housing, money, work, and social isolation. They aimed to explore which topics have been studied, to summarise the gender, ethnicity and socioeconomic position of people included in research, and see if interventions worked better or worse for certain groups.
Usually systematic reviews follow a predictable format, but this paper has a couple of quirks. They used a patchwork design, which included all the studies from the two previous reviews (Barnett et al., 2022; Killaspy et al., 2022), topped up with a fresh search of papers published between 2020-24.
Results
Studies included
- The combined searches included results from 266 studies – a lot of work! 165 were from existing studies and 101 from the updated search.
- Studies looked at interventions in the following areas:
- Employment (34%)
- Social connectedness and social skills (24%)
- Housing (19%)
- Community support (11%)
- Family (7%)
- Education (4%)
- Offending (3%)
- Debt and Finance (1%)
- and Trauma and Victimisation (1%)
- It appears that since 2010 there has been an increase in studies exploring employment, social connectedness / social skills, and housing.
- The studies were conducted in 34 countries, with 98 studies in the USA, and 65% in an urban setting.
Who was involved in the studies?
- Overall, reporting about marginalised groups was poor.
- Gender – most studies reported only male/female genders. The average inclusion of women was 43% (i.e. less than 50%). 10 studies (4%) reported on individuals who identified as non-binary, transgender, gender-free or ‘other’.
- Ethnicity / race #1 – 53% studies reported on this, but methods varied significantly. 47% did not include any data on ethnicity, race, nationality, migration status, indigeneity, heritage or related indicators. Language used in different countries varies significantly.
- Ethnicity / race #2 – The authors note that 87% of studies in the UK or USA did report on ethnicity and/or race, and in the paper they give pooled statistics on the representation of different groups in these studies. They couldn’t do this for other studies as the categories used weren’t consistent.
- Socioeconomic position – 78% studies reported some sort of measure, most commonly education, but others included social class, occupation and financial situation. Because this was measured in so many different ways the authors couldn’t combine results.
- Substance use – 20% studies excluded people with any history of substance use or abuse. 69 studies reported the proportion of people with current substance abuse or dependence.
- Physical health problems – 5 studies excluded people based on physical health problems. 27 studies reported the proportion of people with physical health problems.
Who did the interventions work for?
- Only 20 papers (8%) reported results for different groups based on gender, ethnicity and/or race, or socioeconomic position.
- No studies reported results for people in multiple different marginalised groups.
- Of the studies that did report these results, more than half did not report enough information to interpret them (for example just saying results were not significant – mini-rant on this later!)
- In the small number of studies that did separate results by social group, some showed people with a marginalised socio-economic position benefitted least (but this is weak evidence).
Conclusions
Overall, this systematic review finds that studies of social interventions rarely discussed context, had huge variation in the way data was collected and reported, and that data on marginalised groups was often not reported at all.
Most research was conducted in the urban global north, and most interventions focused on housing, employment, social connectedness and community participation. The authors comment that there are important areas with very little research (such as financial support).
Ethnicity was poorly reported, and most studies that did report this looked at a predominantly white population, so results may not apply to other groups. There was very little representation of non-binary genders. People with a marginalised socio-economic position might benefit least from interventions, but the evidence for this is weak. And there was little inclusion of co-occurring conditions, substance-use or physical ill health.
So, in summary… What works for whom? We don’t know.
Strengths and limitations
This review looked at an important topic and was reported clearly. The authors worked closely with people with lived experience, who are named as co-authors on the paper, which is great to see.
The patchwork search strategy, which combined results from different studies, means some studies could have fallen through gaps, as criteria were slightly different for each of these reviews. It also means different methods were used for assessing the quality of the studies included, as each paper used different tools, so results might not be directly comparable. Overall it’s a pragmatic decision, and one that I think was reasonable – systematic reviews can be really time-consuming, so it doesn’t make sense to repeat someone else’s work.
The inclusion criteria for the review also required a diagnosis, so the authors may have missed studies that did not use a diagnostic framework. They also only looked at English language studies.
There are a number of other limitations of the types of studies included in this review (not necessarily criticisms of this review, but the state of the field).
The review assessed inclusion based on complex concepts of social constructs (such as ethnicity, race, socioeconomic position). Definitions of these vary significantly, and are reported in different ways, which makes comparison between studies difficult. The authors give the example of France, where researchers only report nationality as a binary – French national or not. I’m reminded of the US government’s attempt to restrict the use of words like ‘trans’ or ‘Hispanic’ in research funding bids and government websites (Malhi et al., 2025).
In discussing sub-group analysis, the authors comment that a number of results were not reported, or only described as ‘not significant’. What does this mean! The idea of results being ‘significant’ or ‘not significant’ has long been rubbished, but remains standard practice (Amrhein et al., 2019). A ‘non-significant’ result (based on an arbitrary p-value of 0.05) doesn’t necessarily mean there is no effect, it might be that the study wasn’t able to identify an effect that truly exists. This is particularly relevant to sub-groups with small sample sizes, but showing the results and acknowledging their uncertainty (for example with confidence intervals) is still useful in giving an idea of what might be going on.
Implications for practice
We know lots about the social determinants of mental health, and we know that these might provide one of the best opportunities to intervene and improve people’s quality of life (Kirkbride et al., 2024). We also have information on which interventions might be helpful (like IPS and Housing First), but it is essential that interventions benefit the full range of people who experience mental illness, otherwise we will just widen inequalities, and the most marginalised will be left behind. Greenburgh and colleagues show us that we need to do a lot better to work towards that.
Clinicians should think about how they can support people with social interventions, but be aware that the evidence might not apply to all groups. Services can think about how to build this into routine practice, and how they can contribute to improving the evidence base (for example by collecting and reporting data).
This study was part of a wider programme of work at King’s College London, which included another review exploring targeted social interventions for marginalised communities. This found limited evidence for some interventions, but also stated that we need better evidence to inform policy decisions (Baldwin et al., 2025). To conclude their wider piece of work the authors suggest some ways that social and economic factors can be identified and addressed within mainstream mental health services, based on their experience of doing this in South London (Baldwin et al., 2026). Their suggestions include advocacy for funding, locating services in community hubs, closely monitoring data on key social outcomes, and collaborating with others in a wider network.
Beyond employment and housing there are lots of social interventions that have not been researched. There are probably multiple reasons for this, including the way that funding is prioritised. Social interventions can be difficult to define and to test in trials, so work on social interventions (especially those addressing complex social processes) is less common than more testable interventions. This leads to a continued focus on individual risks and solutions, rather than the broader causes of the causes.
The evidence that does exist might not apply to the global majority population, as it appears to be based on people with white ethnicity and no co-occurring conditions, and ignores intersectionality. It might apply, but given the lack of reporting, we just can’t say if this is the case or not. Trial evidence is only directly applicable to the people included in trials, so if we want an inclusive population health strategy we need to think carefully about who is represented in research, with transparent reporting of this.
For those who are trying to do research in this field the authors give a gentle prod – do better trials, share data, pre-specify analyses, and help us understand what works for whom.
This paper is a reminder of how little we know, and a motivator for future work on inclusive social interventions.
Statement of interests
Jonathan Monk-Cunliffe has no conflicting interests to declare.
Editor
Edited by Laura Hemming.
Links
Primary paper
Anna Greenburgh, Helen Baldwin, Hannah Weir, Zara Asif, Dionne Laporte, Mark Bertram, Achille Crawford, Gabrielle Duberry, Shoshana Lauter, Brynmor Lloyd-Evans, Cassandra Lovelock, Jayati Das-Munshi & Craig Morgan (2025) What works for whom: a systematic review of inequalities in inclusion and effectiveness of social interventions for mental ill- health. Soc. Psychiatry Psychiatr. Epidemiol. https://doi.org/10.1007/s00127-025-02984-3
Other references
Amrhein V, Greenland S, McShane B. (2019) Scientists rise up against statistical significance. Nature 567, 305–307.
Baldwin H, Greenburgh A, Iskander-Reynolds A. et al (2026) Using theory of change to better address social and economic needs in mental health services. Int. J. Ment. Health Syst. 20, 5.
Baldwin H, Greenburgh A, Weir H. et al (2025) Targeted interventions to improve the social and economic circumstances of people with mental ill-health from marginalised communities: a systematic review. Psychol. Med. 55, e217.
Barnett P, Steare T, Dedat Z. et al (2022) Interventions to improve social circumstances of people with mental health conditions: a rapid evidence synthesis. BMC Psychiatry 22, 302.
Bell A. Improving social circumstances for people with mental health difficulties. The Mental Elf, 10 Jan 2023.
Guo X, Higson-Sweeney N. Tackling social determinants will reduce the global mental health burden: mega-blog of current prevention strategies. The Mental Elf, 27 Jun 2025.
Killaspy H, Harvey C, Brasier C. et al (2022) Community-based social interventions for people with severe mental illness: a systematic review and narrative synthesis of recent evidence. World Psychiatry 21, 96–123.
Kirkbride J, Anglin D, Colman, I. et al (2024) The social determinants of mental health and disorder: evidence, prevention and recommendations. World Psychiatry 23, 58.
Malhi GS, Adlington K, Al-Diwani A. et al (2025). The value of mental science: we publish what matters. Br. J. Psychiatry 227, 429–433.
Roberts T, Hemming L. Rethinking How We Evaluate Social Determinants of Mental Health. The Mental Elf, 14 Mar 2025.