Methodological processes in validating and analysing the quality of population-based data: a case study using the Victorian Perinatal Data Collection.

This paper describes methods used and results obtained from a study that measured the accuracy of a routinely collected population-based data set. Data on a random sample of births were extracted from the 2003 Victorian Perinatal Data Collection (VPDC) and compared with information in the original medical record. Accuracy was calculated for 111 items related to diverse aspects of maternity and neonatal health and care. Sensitivity, specificity, positive predictive value and negative predictive value were calculated for dichotomous items. Seventynine items were accurate in at least 97% of cases, 45 of them in at least 99% of cases, and accuracy was below 90% for five items. Very high specificities demonstrate that conditions were rarely reported in error. Lower sensitivities indicate that some events that occurred went unreported on the perinatal form. The excellent results for specifi city indicated that the dataset is appropriate for a conservative analysis of relationships between factors. The lower sensitivities could result in true relationships between factors remaining unidentified. Reasons for discrepancies between the VPDC and the original medical record are described.