A two-phase study of the reliability of computerized morbidity data.

This paper reports the results of a detailed examination of morbidity data collected at the Department of Family Medicine, Brown University Program in Medicine. The reliability study was conducted in two phases, spanning five years, with two distinct aims: (1) to assess the extent of consistency between morbidity data stored in the computer and those data abstracted from medical records, and (2) to determine the impact on data reliability of a standardized, precoded encounter form. This form was implemented between the first and second phases of the study. It was found that morbidity data stored in the computer represents 80 to 85 percent of identifiable morbidity in the medical records. In addition, approximately 8 percent of the computer stored health problems were either incorrectly coded or not found in the medical records. The data reliability improved between the two phases, which points to a positive impact of the encounter form in this setting. Several important methodological issues related to identification and coding of health problems surfaced during the study. These issues are discussed and are relevant to all researchers working with computerized morbidity data in family practice.