Malaria is a parasitic disease that is a major health problem in manytropical regions. The most characteristic symptom of malaria is fever. Thefraction of fevers that are attributable to malaria, the malaria attributablefever fraction (MAFF), is an important public health measure for assessing theeffect of malaria control programs and other purposes. Estimating the MAFF isnot straightforward because there is no gold standard diagnosis of a malariaattributable fever; an individual can have malaria parasites in her blood and afever, but the individual may have developed partial immunity that allows herto tolerate the parasites and the fever is being caused by another infection.We define the MAFF using the potential outcome framework for causal inferenceand show what assumptions underlie current estimation methods. Currentestimation methods rely on an assumption that the parasite density is correctlymeasured. However, this assumption does not generally hold because (i) feverkills some parasites and (ii) the measurement of parasite density hasmeasurement error. In the presence of these problems, we show currentestimation methods do not perform well. We propose a novel maximum likelihoodestimation method based on exponential family g-modeling. Under the assumptionthat the measurement error mechanism and the magnitude of the fever killingeffect are known, we show that our proposed method provides approximatelyunbiased estimates of the MAFF in simulation studies. A sensitivity analysiscan be used to assess the impact of different magnitudes of fever killing anddifferent measurement error mechanisms. We apply our proposed method toestimate the MAFF in Kilombero, Tanzania.
展开▼