Abstract WP330: Unreliable Long-term Cost Data Introduces Substantial Uncertainty Into Stroke Cost-Effectiveness Models

Abstract

Background: Cost effectiveness analysis (CEA) models evaluate the value of stroke care methods. Estimates of long-term costs of stroke, a key input, vary substantially. We explored the effect of long-term cost variation on the interpretation of stroke CEA models. Methods: We estimated the lifetime costs and outcomes associated with intravenous tissue-type plasminogen treatment in acute ischemic stroke within the 0-3 hr and 3-4.5 hr windows. Decision analytic model inputs were from previously published literature, and clinical trials. We varied annual long term care costs, using cost groupings (i.e. cost per year at a given level of function) from published CEA models. Low estimates were from the Stroke Treatment Economic Model based on cost in the United Kingdom. High estimates were from the Rochester Stroke Registry from 1987 to1989. Split estimate took cost of minor disability from the low estimate and cost of major disability from the high estimate. We estimated incremental cost effectiveness ratio (ICER) for the base-case and conducted probabilistic sensitivity analyses. Results: The split estimate resulted in the lowest ICERs and was the dominant strategy (improved efficacy, cost savings) in all three conditions (0-3hr, 3-4.5hr, ECASS III) with ICERS -$67,530, -$223,294, and -$39,517 respectively. The low estimate, while still cost effective, increased ICERS substantially to $809, $144, and $18,549 respectively. When low cost estimates were substituted for split cost estimates, the percentage of dominant stimulations dropped by more than half (figure 1). Conclusion: Varying cost estimates led to considerably different conclusions regarding the cost effectiveness of tPA . For a highly cost-effective therapy, these differences do not affect the overall judgement of cost-effectiveness. However, this uncertainty could alter the cost effectiveness of more marginally effective or costly treatments. More reliable long term stroke cost estimates are needed.