Publications
List of publications, posters and presentations
Orme ME, Perard R. Generalised multinomial logit models for assessing scale heterogeneity in patient preferences: an example using HIV treatment preferences (596)
ABSTRACT:
Background: Patient preference weights may be subject to ‘scale’ heterogeneity whereby some individuals make more random choices compared to others. Generalised multinomial logit (GMNL) models have recently been developed to deal with such heterogeneity (Fiebig, 2010). Objectives: To compare GMNL, scale heterogeneity (SMNL) and mixed logit (MIXL) models. Methods: Data were collected from HIV patients participating in a prospective survey and discrete choice experiment (DCE). The DCE consisted of 12 hypothetical choices of two drug scenarios with five different attributes. For each scenario, patients chose the option they preferred. The analysis was conducted in STATA v13.1. Willingness to pay (WTP) in GBP was used as a metric to quantify the strength of preference for each attribute. Model fit was assessed by Bayes and Akaike information criteria. Results: The sample consisted of 278 UK respondents. 72.6% were men who have sex with men and 14.7% were female, median age was 44 (range 21-66) years. The median time since diagnosis was 8 (0-30) years and median duration of treatment 5 (0-27) years. 36% were on a single tablet regimen (STR), 9% were taking five tablets or more per day. The models provide similar treatment attribute rankings though the range of WTP was more extreme for some models than others. The best fitting model was the GMNL with random correlated attributes followed by GMNL with random correlated attributes with two respondent-level covariates (disease duration and on STR). The SMNL model provided a better fit than the MIXL model with fixed attributes and less extreme WTP than the MIXL model with random attributes. Conclusion: GMNL models provide a flexible framework for analysing preference data taking into account scale heterogeneity.
Presented at the 12th Annual Meeting of the HTAi, 15-17 June 2015, Oslo, Norway.
Citation: HTAI 2015