This is the final article in the series on the epistemology of patient safety research, and considers the selection of study design and end points during the planning of an evaluation. The key message of this series is that "one size does not fit all'': the nature of the evaluation will depend on logistical and pragmatic constraints, a priori assessment of the probability of benefits and harms, the plausible scale of these effects and the target audience for the results. This paper also discusses the advantages of mixed method designs. The strength of any conclusions can be increased if different end points concur and the authors therefore advocate assessment of the effect of the intervention on different end points across the generic causal chain linking structure, process and outcome. The use of both qualitative and quantitative methods is also advocated to help explain findings, generate theory and help contextualise results. We show how a bayesian framework can be used to synthesise evidence from a number of different sources and why this approach may be particularly appropriate for the evaluation of patient safety interventions.