PND35 COMPARISON OF ANALYIC HIERARCHY PROCESS AND CONJOINT ANALYSIS METHODS IN ASSESSING TREATMENT ALTERNATIVES IN STROKE REHABILITATION

OBJECTIVES: There has been increasing interest novel HTA methods that will incorporate patient preferences in a more transparent and scientifi cally valid way. The fundamental problem of the assessment of benefi ts in HTA is the identifi cation, ranking and valuation of multiple health care outcomes. We used two multi-criteria methods to rank and value fi ve different treatments in stroke rehabilitation. Analytic Hierarchy Process (AHP) stems from operations research and is increasingly being used in health care to weigh patient-reported endpoints. Conjoint analysis (CA) is a stated preference method that often takes the discrete choice format. In CA, hypothetical scenarios are used to generate part-worth utilities for attributes. METHODS: To determine the clinical decision context and related criteria, a paper-and-pencil questionnaire was conducted among a sample of Dutch physiatrists united in a stroke interest group. From the lists of criteria (e.g. clinical benefi t, impact of treatment) an expert panel defi ned the AHP decision structure as well as the conjoint analysis survey format. Finally, the complete questionnaire including the AHP and CA survey was sent out to 184 patients with ankle-foot impairments. Eventually, 89 patients completed both surveys. RESULTS: On average, the prediction of preferred treatment on a group level is similar for both AHP and CA. However. on an inidividual level there seems to be more variation in treatment preference. Using AHP weights, a vast majority preferred soft-tissue surgery where most patients preferred orthopedic shoes if CA weights were used. This may have been caused by labelling effects of the attributes. CONCLUSIONS: Both methods have there pros and cons in ranking and valuing patient-reported endpoints. Of the methods AHP is relatively easy to apply. In prediction of overall outcome, both methods perform equally. However, for individual treatment preference we observed some differences. It may be concluded that the decision structure, framing and labelling of the treatment attributes are more important than the specifi c elicitation method used.