Identifying major congenital malformations in the UK General Practice Research Database (GPRD): a study reporting on the sensitivity and added value of photocopied medical records and free text in the GPRD.

BACKGROUND: Postmarketing teratogen surveillance is essential and requires a data source that can reliably capture a wide range of congenital malformations. The UK General Practice Research Database (GPRD) may have the potential to be used for this kind of surveillance.
OBJECTIVE: To assess the extent to which this database can be used to accurately identify major congenital malformations.
METHODS: This study was carried out as part of a broader study to compare data on anticonvulsant use and safety in pregnancy between the GPRD and a pregnancy registry. The study period ran from 1 January 1990 until 31 December 2006. Mother-baby pairs where the mother had a record of epilepsy, seizure or convulsion were identified using the GPRD computerized medical records. Infants of mother-baby pairs who had a record of a major congenital malformation were identified. Full photocopied paper medical records were requested from the infant's general practitioner and where this was not possible any data entries consisting of uncoded comments, so-called 'free text', in the electronic GPRD record were requested from the database provider. This additional information was then reviewed in order to determine the extent to which the congenital malformation diagnoses identified via the computerized records could be confirmed or rejected and then classified as being major or minor.
RESULTS: Within the study population of 3869 live mother-baby pairs, 188 potentially major congenital malformations were identified from the GPRD computerized record relating to 161 unique individuals. Using a combination of photocopied medical records and free text it was possible to verify 160 malformations (85.1%) as the malformation indicated by the computerized records; this ranged from 91.7% of those cases verified using photocopied medical records and 77.9% of cases verified using free text. Of the verified congenital malformations, using a combination of computerized data, photocopied medical records and free text, it was possible to classify 78.1% as being major and 15.0% as minor, and this percentage was found to be the same for those cases reviewed by photocopied records and those where free text was used. The proportions of malformations that could be verified and those that could be classified as major or minor were found to vary by malformation class.
CONCLUSIONS: The GPRD can be used to ascertain a wide range of congenital malformations. In many cases, when a malformation is identified in the GPRD via the computerized medical records, the malformation is likely to exist. However, in this study a small proportion of identified cases had to be excluded because they had been coded incorrectly or diagnostically ruled out. Therefore, depending on the congenital malformation of interest, verification of such malformations using photocopied medical records or free text is generally recommended.