Epidemiological pitfalls using Medicaid data in reproductive health research.

The purpose of this report is to discuss methodologic issues in using Medicaid claims data to conduct epidemiologic analyses in reproductive health. We conducted case-control studies that used Medicaid claims data to evaluate two specific reproductive health questions. Case and control pregnancies were selected from among 106,000 women identified in a Medicaid claims file. Medical record review was conducted for randomly selected cases and controls. Several methodological issues were identified. Women could contribute multiple pregnancies that qualified as either case pregnancies, control pregnancies, or both. The results of the medical record review indicated that 25% of one case group (low-birthweight infants) could not be confirmed, and 70% of the second case group (CNS birth defects) were misclassified. Thirty-five percent of women classified as not having undergone diagnostic ultrasonography based on Medicaid claims data had evidence of having received diagnostic ultrasound during pregnancy on the basis of medical record review. Several problems were encountered with the use of Medicaid billing data to address epidemiologic questions in reproductive health. Although solutions to some of these problems could be identified, others could not be addressed without careful review of the medical records. These limitations may not apply to all state Medicaid databases or other claims data, but they should be carefully considered when planning claims-based analyses of reproductive health issues.