Pinsent et al, UK malaria Additional File 3: Statistical Model

We assume that the risk of acquiring malaria depends on the daily level of exposure (the EIR) and the duration of stay in a malarial endemic country. As such, we assume that the probability of infection per bite is the same for all individuals in all transmission settings. We used a log-linear model to calculate the hazard of acquiring malaria in which the time of infection was assumed to have occurred at any point during the reported visit to a malaria endemic area and the daily hazard of infection is determined by the EIR (assuming a constant rate and thus ignoring the seasonality in transmission since data on the months of visits was not available). Two covariates were considered both in univariable and multivariable models – the age of the participant (as a categorical variable) and the reported reason for travel. The hazard was calculated as an aggregate measure of the effect of the use of preventative measures, such as bed nets, insect repellent and chemoprophylaxis, because detailed usage statistics were not available for travelers in general. We assume that the risk of acquiring malaria increases with the daily level of exposure in the country visited (the non-seasonal EIR, notated E) and the duration of stay in that country, notated d and use an interval-censored proportional-hazards model for the risk of acquiring malaria for travelers with different purpose of travel and age, i.e. Infection could have occurred at any point during the visit t