The utility of the general practice research database to examine selected congenital heart defects: a validation study

Abstract

OBJECTIVE: The purpose of this research was (1) to validate that ventricular septal defect (VSD), tetralogy of Fallot (TOF), and coarctation of the aorta (COA) can be studied in the UK General practice research database (GPRD) and (2) to understand which of the available GPRD components (computerized medical records, questionnaires, and maternal/infant free text) provide maximal information about these heart defects. METHODS: Using a practitioner questionnaire, the positive predictive value (PPV) of the computerized medical record for VSD, TOF, and COA were determined. Both infant and maternal free text was examined. Concordance between the infant free text information and questionnaires was calculated. The proportion of infant information captured in the maternal free text was determined. RESULTS: A 93% response rate was achieved. Based on questionnaire responses, an overall PPV of 93.5% was achieved (VSD = 95%, TOF = 90%, COA = 100%). Approximately half of the records contained infant free text information including information on the type and size of VSD, echocardiogram findings, and surgery. Concordance between the infant's free text and questionnaire information occurred in most of the cases (92-100%). The proportion of infant information in the maternal free text was low (4-19%). CONCLUSION: The GPRD computerized medical records are sufficient to assess VSD, TOF, and COA. This study confirms that maternal free text provides a low yield of limited information pertaining to the infants' defect, while the infant free text may provide an additional information usually obtainable from practitioner questionnaires. The information provided by an infant free text may limit the need for practitioner questionnaire validation.