How Suitable Are Registry Data for Recurrence Risk Calculations? Validation of Diagnoses on 1,593 Families With Congenital Heart Disease

• Published in: World journal for pediatric & congenital heart surgery Vol. 7; no. 2; p. 169
• Main Authors: Ellesøe, Sabrina Gade, Jensen, Anders Boeck, Ängquist, Lars Henrik, Hjortdal, Vibeke Elisabeth, Larsen, Lars Allan, Brunak, Søren
• Format: Journal Article
• Language: English
• Published: United States 01.03.2016
• Subjects: Data Accuracy; Denmark; Genetic Predisposition to Disease; Heart Defects, Congenital - diagnosis; Heart Defects, Congenital - genetics; Humans; Odds Ratio; Registries - standards; Risk Assessment; congenital heart disease (CHD); adult congenital heart disease; cardiology; statistics
• ISSN: 2150-136X
• DOI: 10.1177/2150135115615786

Congenital heart disease (CHD) occurs in approximately 1% of all live births, and 3% to 8% of these have until now been considered familial cases, defined as the occurrence of two or more affected individuals in a family. The validity of CHD diagnoses in Danish administrative registry data has only been studied previously in highly selected patient populations. These studies identified high positive predictive values (PPVs) and recurrence risk ratios (RRRs-ratio between probabilities of CHD given family history of CHD and no family history). However, the RRR can be distorted if registry data are used indiscriminately. Here, we investigated the consequences of misclassifications for the RRR using validated diagnoses on Danish patients with familial CHD. Danish citizens are assigned a civil registration number (CPR number) at birth or immigration, which acts as a unique identifier in the Danish registries, thus enabling connection of information from several registries. Utilizing the CPR number, we identified Danish patients with familial CHD and reviewed each patient's file. We compared diagnoses from the registries with those manually assigned, which enabled calculation of the PPVs of diagnoses in the Danish registries, and from this, we deduced the false discovery rate (FDR). To measure the consequences on the RRR, the FDR was applied to a simulated data set with true RRR values of 2 and 10. We validated diagnoses of 2,442 patients from 1,593 families. Of these, 874 patients were misclassified corresponding to an FDR of 36%. Applying this FDR on the simulated data sets, we found that the RRR decreased from 2 and 10 to 1.4 and 5.1, respectively. Lastly, we estimated that 11% of all cases with CHD were familial. We found that approximately one of nine of all cases with CHD are familial, and we also found that 36% of individuals with CHD in administrative medical registries are misclassified, which distort the RRR in simulated scenarios.