Evaluation of prenatal care information on birth certificates.

Misclassification frequently leads to bias in epidemiological studies, and causes concern for perinatal epidemiologists interested in using birth certificates as a data source. We used a maximum likelihood method to estimate the classification probabilities (conditional probabilities that indicate the probability of classification in a particular category, given the person's true category) of two data sources for a three-category outcome of prenatal care. The probability that women receiving adequate or inadequate care were correctly classified was estimated to be greater than 90%. The probability was much lower (< 35%) that women receiving intermediate care were correctly classified. The misclassification women from the intermediate category resulted in poor predictive values (< 70%) of women classified as receiving either adequate or inadequate care. Because of these findings, we combined the adequate and intermediate categories to form a two-category classification system. This revision resulted in higher positive predictive values (> 90%) with only a slightly lower classification probability (> 85%) for the combined category. We conclude that the degree of accuracy for a two-category classification of prenatal care based upon birth certificate information is acceptable, but we question the accuracy of indices of prenatal care with more than two categories.