To derive and compare relative preferences of physicians and patients for selecting oral antithrombotics in atrial fibrillation (AF).
Elicitation task: Best worst scaling (BWS) choice experiments were constructed from literature review and expert opinion, reflecting four attributes relevant to oral antithrombotics selection in the setting of stroke prevention in AF – frequency of laboratory monitoring, annual risk of stroke, annual risk of major bleed, availability of reversible agent.
Main survey. BWS experimental design was developed using Sawtooth Software. Each physician respondent answered 20 questions and patient respondent answered 16 questions. Patients were recruited through the Atrial Fibrillation Clinic at Vancouver General Hospital and physicians were invited to participate through the local health authority’s email listserv and research broadcast.
Analysis. Relative utilities based on the BWS choice data were derived using the latent class analysis. To determine the difference in preference for each attribute level between physician and patient respondents, the Wilcoxon signed-rank test was performed to assess the difference between the best-worst score for the two groups.
The survey was completed by 33 physicians and 58 patients. Both groups favoured “annual stroke risk of 0%” as the most valued attribute-level with mean utility estimates of 4.58 (p-value <0.001) and 6.5032 (p-value <0.001), respectively. In comparison, “annual stroke risk of 10%” was chosen as the least favourable attribute-level for both physicians and patients with mean utility estimates of -3.66 (p-value <0.001) and -4.86 (p-value <0.001), respectively. Patients preferred “having reversibility agent available” over “laboratory monitoring every year” and deemed “annual stroke risk of 6%” to be a better attribute level than “annual major bleeding risk of 6%”. The reverse was found for the physician respondents in both cases. Wilcoxon signed-rank results revealed significant preference differences between the two perspectives for several of the attribute levels including: “laboratory monitoring every month”, “annual stroke risk of 10%”, “annual bleeding risk of 2%”, and “reversibility agent not available”.
Preferences for stroke prophylaxis differ between physicians and AF patients. This is the first study known to compare valuation between the two perspectives using a BWS choice experiment and provides important insights to clinical decision-making in a patient-centered care model.