Abstract O.02: Validation of Kawasaki Disease Incidence Assessment as Derived from Health System Administrative Databases vs. Active Retrospective Surveillance in Ontario, Canada

    loading  Checking for direct PDF access through Ovid

Abstract

Introduction: Historically, 2 methods have been used to determine the incidence of Kawasaki disease (KD): active or passive surveillance, or the use of administrative databases. Given the increasing regulatory requirements, mainly around patient privacy, periodic retrospective surveillances have become increasingly challenging. Administrative databases are not curated datasets and doubts have been cast on their accuracy.

Methods: The Hospital for Sick Children has been conducting retrospective triennial surveillances of KD since 1995 by contacting all hospitals in Ontario and manually reviewing all cases through chart review, reconciling inter-hospital transfers and multiple readmissions. We queried the Canadian hospital discharge database (Canadian Institute for Health Information) for hospitalizations associated with a diagnosis of KD between 2004-9. The administrative dataset was manually reviewed; patient national health number, institution and dates of admission/discharge were used to identify inter-hospital transfers, readmission and follow-up episodes.

Results: The Canadian hospital discharge database reported 1,685 admissions during the study period (281±44 per year) for Ontario. Manual review of the dataset identified 219 (13%) as inter-hospital transfers (56, 26%), readmissions (122, 56%), admissions for follow-up of coronary artery aneurysms (14, 6%) or hospital admissions not related to KD (27, 12%). When these admissions were removed, the total number of incident cases for the study period was 1,466 (244±45 per year). The retrospective triennial surveillance identified 1,373 KD cases during the same period (229±33 per year). The Canadian hospital discharge database overestimated the number of cases in all 6 years by an average of 6.7±5.9%. The overestimation likely comes from patients who were originally diagnosed with KD but in whom the diagnosis of KD was subsequently excluded (historically ~5-6%).

Conclusions: Reliance on administrative data to determine incidence of KD is possible and accurate; data should be manually reviewed to remove non-incident cases and estimates should be adjusted to reflect the expected proportion of patients in whom the diagnosis of KD will be subsequently excluded.

Related Topics

    loading  Loading Related Articles