"background": "Health system reforms in Ethiopia have focused on expanding district hospital coverage to improve access to care. Evaluating the impact of these policies requires robust quasi-experimental methods, with difference-in-differences (DiD) being a prominent analytical tool. However, the methodological rigour of its application in this context has not been systematically assessed. ", "purpose and objectives": "This review critically evaluates the application of DiD designs in studies measuring the adoption and performance of district hospitals following health system reforms. It aims to appraise model specifications, identification strategies, and the handling of key econometric assumptions. ", "methodology": "We conducted a systematic methodological review of peer-reviewed and grey literature. Studies were analysed for their DiD specification, particularly their approach to the parallel trends assumption, model choice, and handling of potential biases. The canonical two-way fixed effects model is represented as Y{it = \ + \ (Treati \ Postt) + \ + \ +, where \ is the average treatment effect. ", "findings": "A key finding is that over 60% of reviewed studies relied on a simple two-period DiD without testing for pre-intervention trend equivalence. A recurring theme was the omission of robustness checks, such as event-study estimations or the use of heteroskedasticity-robust standard errors clustered at the woreda level, casting uncertainty on the causal interpretation of reported adoption rate increases. ", "conclusion": "While DiD is a valuable method for policy evaluation in this setting, prevalent methodological shortcomings limit the reliability of many existing estimates of reform impacts on hospital adoption rates. ", "recommendations": "Future research must rigorously test the parallel trends assumption, employ dynamic specifications to assess treatment effect heterogeneity, and clearly report inference methods. Sensitivity analyses using alternative estimation strategies are essential. ", "key words": "health policy evaluation, quasi-experimental design, econometrics, parallel trends, health systems strengthening, fixed effects", "
Abebe et al. (Thu,) studied this question.