21.2  Study designs to include

Public health and health promotion are broadly-defined activities that are evaluated using a wide variety of approaches and designs. No single method can be used to answer all relevant questions about all public health and health promotion problems and interventions. If the review question has been specified clearly then types of study designs needed to answer it should automatically follow (Petticrew 2003). A preliminary scoping search will also help to identify the types of study designs that may have been used to study the intervention. The criteria used to select studies should primarily reflect the question or questions being answered in the review, rather than any predetermined hierarchy (Glasziou 2004). The decisions about which type(s) of study design to include will influence subsequent phases of the review, particularly searching, assessment of risk of bias, and analysis (especially for meta-analyses).

 

Randomized trials provide a useful source of evidence of effectiveness, although their results may have limited generalizability (Black 1996). For many health promotion and public health interventions randomized trials may not be available, due to issues including feasibility and ethics. Cluster-randomized trials are increasingly adopted within the field of public health; some interventions require their application at the cluster level (Donner 2004). These trials can contribute valuable evidence if a sufficient number of units are randomized to ensure even distribution of potential confounders among groups: see Chapter 16 (Section 16.3).

 

For some questions, non-randomized studies may represent the best available evidence (of effectiveness). Reviewing non-randomized evidence can give an estimate of the nature, direction and size of effects. Demonstrating the patterns of evidence drawn from different study designs may lead to the development of subsequent study designs (including randomized trials) to test the intervention. Studies generating qualitative data may also be relevant to other kinds of questions beyond effectiveness questions. For example, data may be gathered on the preferences of the likely recipients of the interventions and the factors that constrain or facilitate the successful outcome of particular interventions. Research is ongoing into the differences between randomized and non-randomized studies of public health and health promotion interventions (for example the UK Methodology Programme). Chapter 13 discusses general issues on the inclusion of non-randomized studies in Cochrane reviews, and Chapter 20 addresses qualitative studies.