A SPACE-TIME SCAN STATISTIC FOR DETECTION OF TUBERCULOSIS OUTBREAKS IN THE SAN FRANCISCO HOMELESS POPULATION

San Francisco (SF) has the highest rate of TB in the US. Although in recent years the incidence of TB has been declining in the general population, it appears relatively constant in the homeless population. In this paper, we present a spatio-temporal outbreak detection technique applied to the time series and geospatial data obtained from extensive contact and laboratory investigation on TB cases in the SF homeless population. We examine the sensitivity of this algorithm to spatial resolution using zip codes and census tracts, and demonstrate the effectiveness of it by identifying outbreaks that are localized in time and space but otherwise cannot be detected using temporal detection alone.