A systematic review in public health synthesizes evidence on population-level interventions, health policy effectiveness, and determinants of health to inform decision-making at organizational, community, and governmental levels. Public health reviews differ from clinical systematic reviews because they evaluate complex interventions delivered across diverse settings, populations, and implementation contexts, requiring methodological approaches that go beyond standard Cochrane clinical trial synthesis.
Why Public Health Evidence Synthesis Demands Different Methods
The Medical Research Council (MRC) framework for complex interventions (Skivington et al., 2021) recognizes that public health programs involve multiple interacting components, operate across organizational levels, and produce outcomes shaped by context. A school-based obesity prevention program, for example, involves dietary education, physical activity components, parental engagement, and cafeteria policy changes, all operating simultaneously within specific socioeconomic and cultural environments.
This complexity means that the standard clinical question format often falls short. While the read about pico framework works well for drug trials, public health reviews frequently use PICOS (adding Setting) or SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type) to capture the full scope of the question. Our PICO/SPIDER framework generator supports both formats.
The Campbell Collaboration, the social science counterpart to Cochrane, maintains a library of systematic reviews on education, crime and justice, social welfare, and international development. Their methodological standards complement Cochrane's clinical focus and provide guidance specific to public health and social policy reviews.
Unlike clinical trials with well-defined endpoints, public health outcomes are often distal and diffuse. A tobacco control policy review might measure quit rates, initiation rates, secondhand smoke exposure, healthcare utilization, and economic productivity, each requiring different data sources and synthesis approaches.
Searching the Public Health Literature
Public health evidence is scattered across medical, social science, education, and policy databases. A comprehensive search strategy must cast a wider net than a typical clinical systematic review.
Core databases include PubMed/MEDLINE, Embase, and the Cochrane Library. But stopping here misses critical public health evidence found in CINAHL (nursing and allied health), PsycINFO (behavioral interventions), ERIC (education-based health programs), Social Science Citation Index, and Global Health (international public health, particularly low- and middle-income countries).
Grey literature is disproportionately important in public health reviews. Government reports, WHO technical documents, NGO evaluations, and conference proceedings often contain implementation evidence not published in peer-reviewed journals. Searching OpenGrey, the WHO IRIS repository, CDC WONDER, and organizational websites is standard practice.
The search strategy development process for public health reviews requires balancing sensitivity (finding all relevant evidence) against the practical reality that broader searches generate thousands of records. A well-constructed strategy for a public health topic might retrieve 5,000 to 15,000 records, compared to 1,000 to 3,000 for a focused clinical question.
Handling Complex Interventions in Synthesis
Pooling results from complex public health interventions requires careful judgment about what is meaningfully combinable. The Cochrane Handbook (Chapter 12) provides guidance on synthesizing complex interventions, emphasizing the importance of developing a logic model before starting the review.
A logic model maps the hypothesized causal pathway from intervention components through mediators to outcomes. For a community-based diabetes prevention program, this might include: health literacy sessions (component) leading to dietary knowledge (mediator) leading to HbA1c reduction (outcome), with socioeconomic status and healthcare access as moderators.
Harvest plots and effect direction plots are visual synthesis tools particularly useful when meta-analysis is not feasible due to intervention heterogeneity. These methods, endorsed by the Cochrane Public Health Group, display the direction and magnitude of effects across studies without requiring standardized effect sizes.
When meta-analysis is appropriate, public health reviews frequently use random-effects models because between-study variation is expected and clinically meaningful. Meta-regression can explore whether intervention characteristics (duration, intensity, delivery mode) explain heterogeneity, and our practical forest plot generator visualizes these pooled estimates.
