Title:Measuring spatiotemporal disease clustering with the tau statistic

Abstract:Introduction: The tau statistic uses geolocation and, usually, symptom onset time to assess global spatiotemporal clustering from epidemiological data. We explore how computation and analysis methods may bias estimates.
Methods: Following a previous review of the statistic, we tested several aspects that could affect graphical hypothesis testing of clustering or bias clustering range estimates, by comparison with a baseline analysis of an open access measles dataset: these aspects included bootstrap sampling method and confidence interval (CI) type. Correct practice of hypothesis testing of no clustering and clustering range estimation of the tau statistic are explained.
Results: Our re-analysis of the dataset found evidence against no spatiotemporal clustering p-value $\in$ [0,0.014] (global envelope test). We developed a tau-specific modification of the Loh & Stein bootstrap sampling method, whose more precise bootstrapped tau estimates led to the clustering endpoint estimate being 20% higher than previously published (36.0m,95% bias-corrected and accelerated (BCa) CI (14.9,46.6), vs 30m). The estimated bias reduction led to an increase in the clustering area of elevated disease odds by 44%. We argue that the BCa CI is essential for asymmetric sample bootstrap distributions of tau estimates.
Discussion: Bootstrap sampling method and CI type can bias the clustering range estimated. Moderate radial bias to the range estimate are more than doubled when considered on the areal scale, which public health resources are proportional to. We advocate proper implementation of this useful statistic, ultimately to reduce inaccuracies in control policy decisions made during disease clustering analysis.