Meta-analysis and evidence synthesis for social work

Evidence synthesis in social work

Social work research asks what helps people and communities facing difficulty and how services can be organized to support them. Topics include child protection and foster care, parenting programs, family preservation, domestic abuse services, homelessness and housing support, services for older adults, disability support and community development. The field uses a wide range of methods. Reviews have synthesized trials of parenting programs and intensive family support, and qualitative studies of the experiences of care leavers, parents and workers.

The evidence has features that call for care. Interventions are complex, delivered by professionals and shaped by relationships. Randomization is possible in some settings but often ethically or practically difficult. Outcomes such as child safety rely on administrative records and professional judgments. Voices of people using services are central, and qualitative studies carry much of the evidence on experience. Our methods follow systematic review and meta-analysis practice, adapted to these features. This page builds on the general guidance for social sciences. The service describes research and does not give casework advice.

Outcomes and who defines them

Reviews should record who defined the outcomes and whether service users were involved. Outcomes that matter to families may differ from those chosen by agencies, such as feeling respected and supported. Administrative outcomes like removal may increase when services improve detection, so a rise in recorded referrals is not necessarily a negative result. Different outcome types should be analyzed separately and summarized with attention to what each captures.

Complex interventions and fidelity

Programs such as intensive family preservation, multisystemic therapy and parenting groups have multiple components delivered by workers with varied training, to families with varied needs. Effects depend on the quality of the delivery and the fit with local services. Reviews should record the components, intensity, training and supervision, and implementation fidelity where measured, and examine whether outcomes differ by these features. Evaluations in research conditions, with intensive support to workers, may overstate effects in routine practice. Where a program has a protocol and a trained workforce, outcomes seen in effectiveness trials are more relevant than those from efficacy settings, and the review should separate them.

Child welfare and family support

Reviews of parenting programs have found improvements in child behavior and parenting skills, mostly on parent-reported scales, with smaller or less certain effects on official records of maltreatment. Parent-reported outcomes are open to bias because parents know they received the program. Reviews of family preservation services have found mixed effects on preventing out-of-home placement. Studies of foster care and kinship care compare groups that differ in many ways, so selection is a serious problem. Long-term follow-up is limited. A review should present these limits and avoid causal language for comparisons that cannot support it. Safeguarding decisions in individual cases belong to qualified professionals.

Qualitative and participatory evidence

Qualitative studies of the experiences of service users, carers and workers carry knowledge that trials cannot. They describe how services are experienced, what makes support helpful and what causes harm. Methods such as thematic synthesis and meta-ethnography combine themes across studies, and confidence in the resulting findings can be rated with approaches such as GRADE-CERQual. The qualitative evidence synthesis and mixed-methods review pages describe these methods. Participatory research, in which people with lived experience take part in the design and interpretation, is an important tradition, and reviews can involve service users in setting the question and interpreting findings. We describe how this is done and recognize that it takes time and resources.

Homelessness, housing and community interventions

Studies of housing-first programs, supported employment and community development use randomized and quasi-experimental designs. Housing-first approaches have shown improved housing stability compared with treatment-first approaches in several trials, with smaller or mixed effects on other outcomes. Outcomes depend on the local housing market, and findings from one city may not transfer. Community programs are hard to evaluate because the community is the unit, and the number of communities is small. A review should record the unit of analysis and be cautious about precision.

Older adults, disability and long-term care

Research on services for older adults and people with disabilities covers care coordination, home support, falls prevention and carers' support. Outcomes include independence, institutionalization, carer burden and costs. Trials are often small, follow-up is limited and loss to follow-up through illness or death is common, which complicates analysis. Reviews should record how missing data were handled, and examine whether findings hold when plausible assumptions about missing data change.

Bias, publication and developer effects

Programs evaluated by their developers tend to show larger effects than those evaluated independently. Small positive studies are more likely to be published. We compare sources and designs and search grey literature, including agency reports and theses, which hold a large part of the evidence in social work.

Mental health, substance use and integrated services

Social workers often support people with mental health problems or substance use difficulties, and research covers case management, peer support, motivational approaches and integrated care. Outcomes include symptoms, hospital use, housing, employment and quality of life. Studies of case management differ widely in model and intensity, and effects on symptoms are small on average, with larger effects on service engagement. Reviews should classify models by their components and avoid treating "case management" as a single intervention. Clinical aspects of treatment belong in the clinical psychology and psychiatry reviews, and this page confines itself to service-level questions.

Peer support and lived-experience roles are growing areas with evidence that is varied in quality. Reviews can describe outcomes and experience, and should record the training and supervision of peers.

Workforce, supervision and organization

Studies of the social work workforce examine caseloads, supervision, burnout, turnover and the quality of decisions. Associations between workload and outcomes are usually based on cross-sectional surveys and administrative data. Reviews should state that these designs show association, not cause. Research on decision-making in child protection includes studies of structured decision tools. These have been examined for reliability and predictive accuracy, using the methods for prediction models, with the cautions about fairness and consequences noted elsewhere on this site. Our role is to describe the research on these tools, not to evaluate their use in any agency, and any such evaluation would require access to data and governance arrangements that are outside this service. Workforce findings from one jurisdiction also depend on its legal framework and funding, which differ greatly between countries.

Common pitfalls we look for

• Treating administrative outcomes as direct measures of family wellbeing.
• Relying on parent-reported outcomes without noting bias.
• Interpreting comparisons of care settings as causal.
• Ignoring fidelity and implementation.
• Leaving out qualitative evidence from people using services.
• Generalizing across housing markets and legal systems.

Planning a social work synthesis

We help define the population, intervention, comparator and outcomes with attention to what matters to service users, plan searches in Social Services Abstracts, Social Care Online, ASSIA, PsycINFO, MEDLINE, ERIC and Campbell Library, and set up coding of program components, fidelity, design, outcome definitions and context. See the systematic review service for scope and process.

An invented example of a parenting-program effect

Suppose a review of a parenting program finds a pooled standardized difference of 0.40 on parent-reported child behavior across 25 invented studies, 0.15 on observer-rated behavior, and no clear effect on recorded child protection referrals. A reader would see that the program appears to help on the measure closest to the parent's own view, less on an independent measure, and that evidence for an effect on safety is not there. The review would not call the program ineffective, because referral records are insensitive and follow-up short, but it would report that the evidence for safety outcomes is uncertain and describe what qualitative studies say about how families experienced the program.

Coding and transparency

Coding frames record the population, setting, country, service model, intervention components and dose, worker training, fidelity, design, outcome type and source, who reported, follow-up, attrition and funding. Qualitative studies are appraised with a tool for qualitative research and coded for context, participants and method. Two reviewers work independently on a sample, and the coded data are shared with the final report.

Ethics and respectful language

Work in social care concerns vulnerable people. Reviews use published aggregate data, avoid identifying individuals or families and use language that respects the people described. Where a review involves people with lived experience, we describe how their contributions were recognized and how they could influence the work.

How we support research projects in this area

Boundaries of this service

A social work synthesis describes published evidence on interventions and experiences. It does not provide casework advice, risk assessment of any family or individual, safeguarding decisions or legal opinion. Evidence comes from specific services and places, and many studies have short follow-up and rely on reports by those who received or delivered the service.

Frequently asked questions