Validation of congenital anomaly coding in Canada's administrative databases compared with a congenital anomaly registry.

BACKGROUND: Congenital anomaly (CA) surveillance provides epidemiologic data that are necessary for health planning. Approaches to CA surveillance vary; however, an increasing number of jurisdictions rely on administrative health databases for case ascertainment. This study aimed to assess the validity of CA coding in three administrative databases compared with a CA registry.
METHODS: A cohort of 5862 live and stillborn infants from Calgary Alberta Canada was created through linking 12 clinical and administrative databases. Diagnostic codes for all health care contacts (hospitalizations, emergency room visits, out-patient physician visits) in the first 3 months of life were examined for relevant International Classification of Disease codes. Sensitivity, positive predictive values, and kappa coefficients were calculated, and data from the Alberta Congenital Anomalies Surveillance System was used as the reference standard.
RESULTS: The ability of administrative data to accurately ascertain CAs varied by data source and the specificity of the diagnosis. Consistently, hospitalization data out-performed other administrative data sources in terms of sensitivity, positive predictive values, and kappa. Kappa scores for CAs easily visible at birth ranged from moderate (0.62 for emergency room visits and 0.65 for out-patient physician claims) to good (0.83 for hospitalization data) depending on the data source.
CONCLUSION: The validity of CA coding in administrative databases compared with a CA registry varies by database used and by CA studied. This has important implications for national surveillance efforts. Condition-specific validity should be assessed locally before use of these data sources for research or planning purposes.