Efficiency and Accuracy of Disease Monitoring Systems: Application of Capture-Recapture Methods to Injury Monitoring

Abstract

Capture-recapture methods were employed to determine the most accurate and efficient approaches to monitor adolescent injuries. Multiple sources were used to ascertain cases of adolescent injuries that occurred between September 1 and December 31, 1991, in a single school district in metropolitan Pittsburgh, Pennsylvania. Eliminating the duplicate cases between the sources revealed 144 verified injuries; 127 (88.2%) were identified by student monthly recalls, 33 (22.9%) by daily attendance records, 58 (40.3%) by medical excuses, and 72 (50.0%) by a 4-month student recall. Capture-recapture analyses were undertaken to assess potential dependencies between the sources, to estimate the degree of underascertainment in the population, and to evaluate the efficiency of the individual sources and the combinations between them. It was estimated that 91% of the cases in the population were ascertained when all four methods of case finding were utilized. Furthermore, the analysis indicated that accurate injury estimates could be achieved using combinations of only two or three of the sources. An analysis of the efficiency of the methods of ascertainment revealed a trade-off between effort (the number of hours needed to identify cases) and the precision (coefficient of variation) of the injury estimates. Capture-recapture analysis not only provided an approach to evaluate and adjust for undercount but also offered a formal means to evaluate the most efficient combination of the sources to maximize completeness while minimizing effort. The use of these techniques has the potential to evaluate and improve injury surveillance as well as other disease monitoring systems. Am J Epidemiol 1995;142:1069–77.