Natural experiment methodology for research: a review of how different methods can support real-world research

Abstract

In particular research domains, the randomized control trial (RCT) is considered to be the only means for obtaining reliable estimates of the true impact of an intervention. However, an RCT design would often not be considered ethical, politically feasible, or appropriate for evaluating the impact of many policy, programme, or structural changes common in public health research. As such, researchers must use alternative yet robust research methods for determining the impact of such interventions. The evaluation of natural experiments (i.e. an intervention not controlled or manipulated by researchers), using various experimental and non-experimental design options can provide an alternative to the RCT. The following review highlights (a) the importance of evaluating natural experiments; (b) design considerations associated with evaluating natural experiments; (c) methods for reducing bias in natural experimental studies; and (d) the potential benefits of targeted systems to enable natural experiments in emerging priority domains moving forward.