Using Phase-Based Costing of Real-World Data to Inform Decision–Analytic Models for Atrial Fibrillation

Background Atrial fibrillation (AF) poses a significant economic burden. An increasing number of interventions for AF require cost-effectiveness analysis with decision–analytic modeling to demonstrate value. However, high-quality cost estimates of AF that can be used to inform decision–analytic models are lacking. Objectives The objectives of this study were to determine whether phase-based costing methods are feasible and practical for informing decision–analytic models outside of oncology. Methods Patients diagnosed with AF between 1 January 2003 and 30 June 2011 in Ontario, Canada were identified based on a hospital admission for AF using administrative data housed at the Institute for Clinical Evaluative Sciences. Patient observations were then divided into phases based on clinical events typically used for decision–analytic modeling (i.e., minor stroke/transient ischemic attack [TIA], moderate to severe ischemic stroke, myocardial infarction, extracranial hemorrhage [ECH], intracranial hemorrhage [ICH], multiple events, death from an event, or death from other causes). First 30-day and greater than 30-day costs of healthcare resources in each health state were estimated based on a validated methodology. All costs are reported in 2013 Canadian dollars (Can$) and from a healthcare payer perspective. Results Patients (n = 109,002) with AF who did not experience a clinical event incurred costs of Can$1566 per 30 days, on average. The average 30-day cost of experiencing a fatal clinical event was Can$42,871, but the cost of dying from all other causes was much smaller (Can$12,800). The clinical events associated with the highest short-term costs were ICH (Can$22,347) and moderate to severe ischemic stroke (Can$19,937). The lowest short-term costs were due to minor ischemic stroke/TIA (Can$12,515) and ECH (Can$12,261). Patients who had experienced a moderate to severe ischemic stroke incurred the highest long-term costs. Conclusions Real-world Canadian data and a phase-based costing approach were used to estimate short- and long-term costs associated with AF-related major clinical events. The results of this study can also inform decision–analytic models for AF.