A mechanistic spatiotemporal model for drug resistant infections

Abstract

AbstractWhen drug resistance is suspected to be in a region, patients in the region are sampled and the suspicion is confirmed. This biased sampling limits our ability to capture underlying dynamics, meaning strategies to lengthen the lifespan of drugs are reactionary, not proactive.Testing for drug resistant infections is becoming easier and cheaper, therefore we should revisit sampling decisions. We present a hierarchical mechanistic Bayesian model, and apply it to a simulated dataset, where we sample between 5% and 30% of the population in a biased and unbiased manner. We show that unbiased spatiotemporal data on the presence of drug resistant infections, combined with our model, highlights underlying dynamics.Our mechanistic model is more accurate than a generalised additive model with space and time components. Moreover, highlighting underlying dynamics creates novel strategies that lengthen the lifespan of drugs. In low to middle income countries, generally, drug resistance emerges into a population from hotspots such as treatment centres (perhaps the use of sub-standard drugs), or major transport hubs, and then resistance spreads throughout the population. Using our model, we rank resistance hotspots, enabling resources to be targeted - such as verifying the quality of drugs at a particular health care centre.