Validation of an Online Mapping Methodology to Locate Village of Residence of Tuberculosis Patients in Mombasa

Abstract

University of Washington Abstract Validation of an Online Mapping Methodology to Locate Village of Residence of Tuberculosis Patients in Mombasa Ruth Wangui Deya Chair of the Supervisory Committee: Professor R. Scott McClelland Departments of Medicine, Epidemiology and Global Health BACKGROUND In Kenya, the dual TB-HIV epidemic has led to increasing morbidity and mortality especially in urban areas. National TB programs need to identify communities with the highest TB burden and to employ aggressive prevention measures such as intensive case finding and contact tracing in order to interrupt ongoing disease transmission. We evaluated the accuracy of an online mapping method to locate TB patients' village of residence using addresses provided in TB registers in Mombasa, Kenya. METHODS To validate the spatial accuracy of a health-worker-based online mapping procedure, we compared whether latitude and longitude point locations of TB patients matched those collected using a Global Positioning System (GPS) device during a home visit. Patient, clinic, and village-level characteristics associated with correct location in village of residence were evaluated using logistic regression. RESULTS Between July 2012 and July 2013, 246 participants who met study inclusion criteria were mapped using both methods. The online method located 140 cases (56.9%) in the correct village of residence, 80 (32.5%) and 26 (10.5%) were incorrectly allocated to a neighboring and non-neighboring village, respectively. The mean error distance between the two methods was 225.2m (±296.2) for all cases, 106.5m (±92.5) and 382m (±386.9) for those correctly located and incorrectly located, respectively. Type of clinic, clinic caseload, village status, and density were significantly associated with being correctly mapped to the village of residence in bivariate analysis. In multivariate analyses, medium caseloads (201 - 400 cases versus 0 - 200 cases per year; adjusted OR 2.75 95% CI 1.37 - 5.53; p 0.004); moderate number of households per village (1,001 - 2000 versus 0 - 1000 households; aOR 0.42 95% CI 0.21 - 0.85; p 0.016); high village density (>30,000 versus 0-15,000 per km2; aOR 0.43 95% CI 0.19 - 0.94; p 0.035) and attending municipal clinics (aOR 0.45 95% CI 0.20 - 0.99; p 0.048) were significantly associated with correct village location. DISCUSSION This online mapping tool correctly allocated almost 60% of TB cases in their village of residence while majority of misclassifications were cases located in neighboring villages. New mapping tools such as online Google Maps could be improved and employed in surveillance of TB cases for timely diagnosis, follow-up and treatment completion. Tuberculosis programs in dense urban areas with high caseload clinics may benefit from improving the quality of addresses obtained in their TB registers in order effectively implement intensive case finding and contact tracing activities.