Quality control of birth defect registry data: a case study.

The California Birth Defects Monitoring Program maintains a population-based registry of children born with congenital malformations. Trained data collectors routinely visit hospitals and genetics centers to identify cases and abstract information. These data are provided to the public health, medical, and lay communities and are used for conducting prevalence and case-control studies. A stratified sample of each data collector's work for one data year was reevaluated to assess the quality of case ascertainment and record abstraction. The sample included data from 109 facilities (37 percent) and 729 abstracts (5 percent). There are three steps in data collection: case-finding, the process of identifying potential cases; culling, the process of reviewing the charts of potential cases to determine which are reportable; and abstracting, the process of recording information from the charts of reportable cases. The probability that a potential case is missed during casefinding is 7 percent for small facilities, 4 percent for medium facilities, and 1 percent for large facilities. The probability that a reportable case is mistakenly classified as not reportable during culling is 3 percent for small and medium facilities and 1 percent for large facilities. The probabilities of incorrectly abstracting selected diagnoses and demographic items are slightly higher (8 percent for small and medium facilities and 6 percent for large facilities) because these are more complex processes than are casefinding and culling. Finally, the overall probability of missing a case from the registry is 3 percent. Therefore, these data indicate that the information collected by this registry are both reliable and complete.