Electronic health care databases are increasingly being used to investigate the epidemiology of congenital anomalies (CAs) although there are concerns about their accuracy. The EUROlinkCAT project linked data from eleven EUROCAT registries to electronic hospital databases. The coding of CAs in electronic hospital databases was compared to the (gold standard) codes in the EUROCAT registries. For birth years 2010-2014 all linked live birth CA cases and all children identified in the hospital databases with a CA code were analysed. Registries calculated sensitivity and Positive Predictive Value (PPV) for 17 selected CAs. Pooled estimates for sensitivity and PPV were then calculated for each anomaly using random effects meta-analyses. Most registries linked more than 85% of their cases to hospital data. Gastroschisis, cleft lip with or without cleft palate and Down syndrome were recorded in hospital databases with high accuracy (sensitivity and PPV ≥ 85%). Hypoplastic left heart syndrome, spina bifida, Hirschsprung's disease, omphalocele and cleft palate showed high sensitivity (≥ 85%), but low or heterogeneous PPV, indicating that hospital data was complete but may contain false positives. The remaining anomaly subgroups in our study, showed low or heterogeneous sensitivity and PPV, indicating that the information in the hospital database was incomplete and of variable validity. Electronic health care databases cannot replace CA registries, although they can be used as an additional ascertainment source for CA registries. CA registries are still the most appropriate data source to study the epidemiology of CAs.