Data quality of a computerized medical birth registry

Assessment of the quality of multi-institutional data for population-based epidemiological studies: a case study of atoxoplasmosis in Bali mynahs (Leucopsar rothschildi)

To evaluate the potential to conduct population-based epidemiologic studies using multi-institutional data, the quality of data extracted from the MedARKS system was evaluated with regard to atoxoplasmosis in Bali mynahs (Leucopsar rothschildi). Data extracted for 338 Bali mynahs from 34 institutions were analyzed for completeness, correctness, and consistency and subjectively analyzed for accessibility. Data completeness was ascertained by analysis of missing records, tests, and blank fields. Two hundred and sixty four records had animal enclosure information missing, nine records were missing from the parasitology module, and 85 records did not include tests or results. Data correctness was assessed by evaluation of unclear and inaccurate results. From 2,432 parasitology records, 81 tests (3.3%) were not definitive because of uninterpretable entries. Data consistency was assessed by comparing the problem list to positive tests in the parasitology module and listing of buffy coat smears in the clinical pathology module. Overall, six different terms with 28 values were used to denote "positive for Atoxoplasma sp.," and a substantial number of discrepancies were found between problem lists, parasitology modules, and test findings among these data records. In general, this study showed that the data contained in a computer patient record (CPR) system for zoos has great promise for population-level studies if specific areas are addressed, including 1) reduced reliance on free-text data entry, 2) universal use of a standardized vocabulary, 3) use of methods to identify and track individual animals accurately and easily, 4) integration of data checks and maintenance methods, and 5) concerted use of "centralized" animal medical record information.