Abstract

Objectives

An increasing number of methods are used to elicit health preference information. It is unclear whether different elicitation methods produce similar results and policy advice. Here, we compared the results from a discrete choice experiment (DCE) and multidimensional thresholding (MDT) that were conducted in the same sample.

Methods

Clinicians (N = 350) completed a DCE and MDT to elicit their preferences for 4 attributes related to the medical management of subarachnoid hemorrhage after aneurysm repair. Preference weights were compared between the DCE and MDT using a complete combinatorial convolution test. Additionally, data from the DCE and MDT were used to compute preference-based net treatment values for 16 hypothetical treatment profiles versus 1000 simulated comparators. The implied treatment recommendations were compared between the DCE and MDT.

Results

Preference weight distributions and median weights did not differ significantly between the DCE and MDT for any attribute: likelihood of delayed cerebral ischemia (medians 0.48 vs 0.40; P = .41), risk of lung complications (medians 0.27 vs 0.30; P = .52), risk of hypotension (medians 0.10 vs 0.11; P = .55), and risk of anemia (medians 0.07 vs 0.07; P = .50). The DCE and MDT produced similar treatment net value distributions (P > .05) and implied the same treatment recommendations in 82.3% of cases.

Conclusions

The DCE and MDT elicited similar preference distributions and produced the same treatment recommendations for most tested cases. However, the share of people supporting the average treatment recommendation differed. More research is needed to determine how these findings would compare with those in other populations (in particular, patients) and applications.