The type of interventions implemented and their subsequent success or failure are highly dependent on the social, economic and political context in which they are developed and implemented (see example in Figure 21.6.a). A problem in reviewing public health and health promotion interventions is how to disentangle ‘intervention’ effects from effects that should be more appropriately called ‘program by context interactions’ (Hawe 2004). Traditionally, outcomes have been attributed to the intervention. However, the outcomes noted in studies may in fact be due to pre-existing factors of the context into which the intervention was introduced. Hence, context should be considered and measured as an effect modifier in studies (Eccles 2003, Hawe 2004). Such contextual factors might relate to aspects of the program’s ‘host organization’. Broader aspects of context might include aspects of the system within which the host organization operates. Some investigators would also argue that context factors also pertain to the characteristics of the target group or population. For many years these aspects have been acknowledged (but not clearly specified) when decision makers have argued that results of evidence reviews from other countries do not apply in their own country.
Use of the term ‘context evaluation’ became more prevalent in health promotion after the review by Israel and colleagues (Israel 1995). However the systematic investigation of context-level interactions as part of the design of randomized trials of community or organizational-level interventions is almost unknown (Eccles 2003, Hawe 2004). Instead, aspects of context have been explored as part of the more developed field of sustainability research or research on program institutionalization: see Section A related and growing multidisciplinary research field is the implementation and integration sciences that are leading researchers more into the complexity of the change processes that interventions represent (Ottoson 1987, Bauman 1991, Scheirer 1994). At the present time, quantitative studies lag behind qualitative analyses of context.
Systematically disentangling context effects from intervention effects in anything other than a study set up for this purpose is extremely difficult. Whilst some programs have been transferred from one context to another and benefits have been observed (Resnicow 1993), others have not (Lumley 2004). Cluster-randomized designs may be expected (in theory) to even out important aspects of context, provided that the sample size is sufficient. However, few investigators at present measure or report on any aspect of context that might be important to our assessment. We also note recent calls for a greater focus on external validity (Glasgow 2006, Green 2006). Working together, journal editors and researchers are encouraging more examination of, and reporting on, aspects of intervention context (Armstrong 2008). This should be reflected in the content of future Cochrane reviews.