Administrative Databases: Friend or Foe in Paediatric Cardiomyopathy

Abstract

Background

Cardiomyopathy (CM) is a rare childhood disease associated with morbidity and mortality. Limited data exist on paediatric CM in Canada. Given the rare nature, single-centre studies are not sufficiently powered to address important questions. Therefore, administrative health data may serve as a resource for the study of childhood CM. The goal of this study was to validate the accuracy of International Classification of Diseases (ICD)-based algorithms to identify paediatric CM in health databases using a clinical registry as the gold standard.

Methods

The clinical registry was compiled from outpatient and inpatient records at the Stollery Children’s Hospital (January 1, 2013, to December 31, 2021). Patients were categorized as having CM or screened without CM. Data were linked to administrative health databases using the patient’s Unique Lifetime Identifier. Algorithms based on the presence of ICD, 10th Revision, codes for CM were then evaluated, and cross-tabulations against the clinical registry were generated. Accuracy, positive predictive value, negative predictive value, sensitivity, and specificity were calculated.

Results

The clinical registry had 90 patients with CM and 249 screened without CM. The algorithms ruled out CM (high negative predictive value) but had variability in the ability to diagnose CM positive predictive value. The algorithm that performed the best was based on a diagnosis of CM in a hospitalization or 2 ambulatory visits.

Conclusions

A combination of inpatient and outpatient databases can be used, with acceptable accuracy, to identify paediatric patients with CM. This finding allows for the use of the identified algorithm for the comprehensive study of paediatric CM in Canada.