Spatio-temporal modeling of sparse geostatistical malaria sporozoite rate data using a zero inflated binomial model.

The proportion of malaria vectors harboring the infectious stage of the parasite (the sporozoite rates) is an important component of measures of malaria transmission. Variation in time and/or space in sporozoite rates contribute substantially to spatio-temporal variation in transmission. However, because most vectors test negative for sporozoites, sporozoite rate data are sparse with large number of observed zeros across locations or over time in the case of longitudinal data. Rarely are appropriate methods and models used in analyzing such data. In this study, Bayesian zero inflated binomial (ZIB) geostatistical models were developed and compared with standard binomial analogues to analyze sporozoite data obtained from the KEMRI/CDC health and demographic surveillance system (HDSS) site in rural Western Kenya during 2002-2004. ZIB models showed a better predictive ability, identified more significant covariates and obtained narrower credible intervals for all parameters compared to standard geostatistical binomial model.