Attribution of reductions in malaria prevalence in Dar es Salaam, Tanzania

摘要

In their Article, Gerry Killeen and colleagues* reanalysed the UMCP data. Although it is reassuring that most of their results confirm those previously reported, their paper raises important concerns. First, the rollout of larviciding followed a stepped- wedge design. It involved sequential crossover of the city's wards from control to intervention such that, ultimately, all wards received the larviciding intervention. This implies that calendar time is a potentially crucial confounder and, following standard practice, must be included in the analyses.° Substituting another covariate for time in the model could grossly overestimate the intervention’s impact.