Identifying children with excess malaria episodes after adjusting for variation in exposure: identification from a longitudinal study using statistical count models.
Ndungu FM., Marsh K., Fegan G., Wambua J., Nyangweso G., Ogada E., Mwangi T., Nyundo C., Macharia A., Uyoga S., Williams TN., Bejon P.
BACKGROUND: The distribution of Plasmodium falciparum clinical malaria episodes is over-dispersed among children in endemic areas, with more children experiencing multiple clinical episodes than would be expected based on a Poisson distribution. There is consistent evidence for micro-epidemiological variation in exposure to P. falciparum. The aim of the current study was to identify children with excess malaria episodes after controlling for malaria exposure. METHODS: We selected the model that best fit the data out of the models examined and included the following covariates: age, a weighted local prevalence of infection as an index of exposure, and calendar time to predict episodes of malaria on active surveillance malaria data from 2,463 children of under 15 years of age followed for between 5 and 15 years each. Using parameters from the zero-inflated negative binomial model which best fitted our data, we ran 100 simulations of the model based on our population to determine the variation that might be seen due to chance. RESULTS: We identified 212 out of 2,463 children who had a number of clinical episodes above the 95(th) percentile of the simulations run from the model, hereafter referred to as "excess malaria (EM)". We then identified exposure-matched controls with "average numbers of malaria" episodes, and found that the EM group had higher parasite densities when asymptomatically infected or during clinical malaria, and were less likely to be of haemoglobin AS genotype. CONCLUSIONS: Of the models tested, the negative zero-inflated negative binomial distribution with exposure, calendar year, and age acting as independent predictors, fitted the distribution of clinical malaria the best. Despite accounting for these factors, a group of children suffer excess malaria episodes beyond those predicted by the model. An epidemiological framework for identifying these children will allow us to study factors that may explain excess malaria episodes.