An example of a poor use of facility-based data to claim the anti-malaria impact of interventions:
In 2007 the Millennium Village Project published early results related to agriculture, health, and economic development in their three research villages in Kenya, Ethiopia, and Malawi. I am not commenting on their work as a whole but their malaria claims were disproportionate to the evidence (especially given the bias of self-evaluation). The group does note some of their limitations but they also unequivocally state to “have reduced malaria prevalence”. It is an interesting use of the term prevalence, which usually refers to the proportion of parasitemic people in a survey, whereas they used facility based data and measured changes in the proportion of clinical malaria cases against total clinic visits. Regardless, even if prevalence or incidence decreased any credit due is unsubstantiated. Some of the obstacles in interpreting their data include the use of clinical malaria cases (which is problematic as discussed here), no demographic comparisons of facility controls, displaying a reduction in diagnostically confirmed cases without any context of testing trends, and only one year of follow up in the absence of any discussion of pre-intervention trends. To be frank the malaria portion of the paper is rather bad science. I am surprised the Proceedings of the National Academy of Sciences endorsed it.