R Thomson1, D Parkin, M Eccles, M Sudlow, A Robinson. 1. Department of Epidemiology and Public Health, School of Health Sciences, Newcastle Medical School, Newcastle Upon Tyne, UK. richard.thomson@ncl.ac.uk
Abstract
BACKGROUND: Clinical guidelines are needed on whether or not to use anticoagulant therapy to prevent stroke in patients with non-valvular atrial fibrillation. We did a Markov decision analysis to model decision-making with regard to warfarin treatment in patients with atrial fibrillation, and used the model to develop evaluative guidelines. METHODS: The decision analysis involved a systematic literature review supplemented by patients' estimates of the quality of life associated with different states of health, secondary analysis of stroke-registry data, and estimation of service costs; it also incorporated a sensitivity analysis. The derived guidelines were subsequently applied to a cohort of patients with atrial fibrillation. FINDINGS: We constructed decision tables for 12 age and sex groups. For most risk combinations, warfarin treatment would have decreased health-care costs and increased quality-of-life years, although the clinical decision was sensitive to patients' preferences and to the estimate of warfarin's effectiveness. 97% of women with atrial fibrillation older than 75 years, and 69% aged 65-74 would have been recommended for treatment; for men, the corresponding figures would have been 75% and 53%. With the upper quartile for the loss of quality of life associated with being on warfarin treatment (1.00), all but two of the 116 patients without contraindications would have been treated, whereas with the lower quartile (0.92), only 27 of 116 would have been treated. INTERPRETATION: Decision analysis is useful in the incorporation of complex probabilistic data into informed decision-making, the identification of factors influencing such decisions, and the subsequent development of evaluative guidelines.
BACKGROUND: Clinical guidelines are needed on whether or not to use anticoagulant therapy to prevent stroke in patients with non-valvular atrial fibrillation. We did a Markov decision analysis to model decision-making with regard to warfarin treatment in patients with atrial fibrillation, and used the model to develop evaluative guidelines. METHODS: The decision analysis involved a systematic literature review supplemented by patients' estimates of the quality of life associated with different states of health, secondary analysis of stroke-registry data, and estimation of service costs; it also incorporated a sensitivity analysis. The derived guidelines were subsequently applied to a cohort of patients with atrial fibrillation. FINDINGS: We constructed decision tables for 12 age and sex groups. For most risk combinations, warfarin treatment would have decreased health-care costs and increased quality-of-life years, although the clinical decision was sensitive to patients' preferences and to the estimate of warfarin's effectiveness. 97% of women with atrial fibrillation older than 75 years, and 69% aged 65-74 would have been recommended for treatment; for men, the corresponding figures would have been 75% and 53%. With the upper quartile for the loss of quality of life associated with being on warfarin treatment (1.00), all but two of the 116 patients without contraindications would have been treated, whereas with the lower quartile (0.92), only 27 of 116 would have been treated. INTERPRETATION: Decision analysis is useful in the incorporation of complex probabilistic data into informed decision-making, the identification of factors influencing such decisions, and the subsequent development of evaluative guidelines.