LAUSR

To meet the transmission reduction goals of the End TB strategy, there is a growing interest in identifying and targeting case-finding efforts to tuberculosis“hotspots,” geographic regions of active transmission. Collecting and interpreting spatial and pathogenic genetic information, the most reliable evidence of active transmission, is prohibitively resource-intensive under routine conditions in high-burden settings. Many countries maintain case-notification registers under routine conditions, representing an attractive source of data to investigate for transmission. However, notification data are imperfect. Areas of high incidence may reflect other underlying patterns, and individual-level covariate information and other information that may aid in its interpretation, such as baseline census data or other healthcare utilization data, is often unavailable. Despite imperfections, the accessibility of notification data demands further investigation. We examined notification data from 2005 to 2007 in a South American, high-burden setting where the household address of each case was geocoded. Subsequent investigation of notification data in the same setting from 2009 to 2012 additionally provided pathogen genetic evidence from all culture-positive cases suggesting regions of active transmission of tuberculosis. We investigated a disease mapping modeling approach leveraging only age-specified tuberculosis notification data to suggest hotspots of active tuberculosis transmission. Given the absence of baseline population data at a comparable spatial resolution, we aggregated the point-referenced cases reported to the Peruvian National Tuberculosis Program from 2005 to 2007 within two of Lima’s four health districts into a grid with 400 m × 400 m cells. We used Bayesian hierarchical spatial modeling methodology to model the proportion of children cases of the total number of adult and child cases in each cell. Where the modeled proportion of child cases is higher than expected, we suggest that case notification is driven primarily by active transmission. This method identified several grid cells in which the proportion of child cases is higher than expected. The location of these grid cells was found to approximate the location of active transmission evidenced by a later genotyping study. This evidence suggests that age-specified notification data, with all its limitations, may be sufficient to suggest hotspots of active transmission of tuberculosis. We additionally provide the first spatial evidence to support the long-cited belief that with respect to tuberculosis transmission, childhood cases may truly be “the canary in the coal mine.”