We tested three mobile phone data aggregations to maximize the predictive power of spatial disease models. Tracking every individual displacement worsens predictions by fragmenting long-range trips, while aggregated time indicators coupling home locations and most-visited places maximize accuracy.
