ABSTRACT
Dif cult to Randomize .................................................................. 160 12.4.4 Pros and Cons of Choosing to Do a Nonrandomized Study .......160 12.4.5 How to Design a Nonrandomized Study to Estimate
the Bene ts of an Intervention ..................................................... 162 12.5 Conclusion .................................................................................................. 163 References ............................................................................................................. 164
From the point of view of a health policy maker, evaluation studies should have two key properties. First, they should be designed to protect against bias (“validity”). Second, they should be designed so that the ndings are relevant to, and can be applied to, typical healthcare situations (“applicability” or “external validity”) [1]. It can sometimes be dif cult to design a study that is both highly valid and highly applicable and different study designs may be appropriate to evaluate subtly different research questions, de ned by the population, intervention, comparator, and outcome(s) of interest (“PICO”) [1].
