BackgroundInequitable access to primary health care (PHC) remains a problem for most western countries. Failure to scale up effective interventions has been due, in part, to a failure to share the logic and essential elements of successful programs. The aim of this paper is to describe what we learned about improving access to PHC for vulnerable groups across multiple sites through use of a common theory-based program logic model and a common evaluation approach. This was the IMPACT initiative.MethodsIMPACT’s evaluation used a mixed methods design with longitudinal (pre and post) analysis of six interventions. The analysis for this paper included four of the six sites that met study criteria. These sites were located in Canada (Alberta, Quebec and Ontario) and Australia (New South Wales). Using the overarching logic model, unexpected findings were reviewed, and alternative explanations were considered to understand how the mechanisms of each intervention may have contributed to results.ResultsEach site addressed their local access problem with different strategies and from different starting points. All sites observed changes in patient abilities to access PHC and provider access capabilities. The combination of intended and observed consequences for consumers and providers was different at each site, but all sites achieved change in both consumer ability and provider capability, even in interventions where there was no activity targeting provider behaviors.DiscussionThe model helped to identify, explore and synthesize intended and unintended consequences of four interventions that appeared to have more differences than similarities. Similar outcomes for different interventions and multiple impacts of each intervention on abilities were observed, implying complex causal pathways.ConclusionsAll the interventions were a low-cost incremental attempt to address unmet health care needs of vulnerable populations. Change is possible; sustaining change may be more challenging. Access to PHC requires attention to both patient abilities and provider characteristics. The logic model proved to be a valuable heuristic tool for defining the objectives of the interventions, evaluating their impacts, and learning from the comparison of ‘cases’.