Sunday, October 18, 2009
Grand Ballroom, Salons 1 & 2 (Renaissance Hollywood Hotel)
Larry D. Lynd, PhD1, Lindsey Colley, MSc1, Camila Guimaraes, PhD1, Carlo A. Marra, PharmD, PhD1, Sabrina Gill, MD1, Scot Simpson, PharmD, MSc2, Graydon Meneilly, MD1 and Regina HC Queiroz, MSc, PhD3, (1)University of British Columbia, Vancouver, BC, Canada, (2)University of Alberta, Edmonton, AB, Canada, (3)University of Sal Paulo, Ribeirao Preto Sao Paulo, Brazil
Purpose:
The purposes of this study were to: 1) use a latent class analysis to model heterogeneity in patient’s preferences for oral, inhaled, and subcutaneous insulin for the management of diabetes, derived using a discrete experiment; and 2) to determine the relative importance of each treatment attribute using a discrete choice experiment (DCE). Methods:
The DCE was compromised of 17 choice sets of 6 attributes. The DCE was designed to determine patients' preferences for route of insulin delivery, the associated risks and benefits of insulin therapy, and their willingness to pay for alternative routes of delivery. In addition to the standard multinomial logit model (MNL), the data were analyzed using a latent class model (LCM) to evaluate if heterogeneity in patients' preferences exists, and to identify characteristics of respondents with different preferences. Ten socio-demographic variables were investigated for inclusion in the final model based on their influence on class membership. Results:
Two hundred and eighty-four respondents (mean age 57 ± 13 years) completed the DCE. The LCM provided a better fit than the simple MNL model therefore only these results are reported. The latent class analysis revealed that there was heterogeneity in preferences, and that there were 5 latent classes. Most parameter values were significant at the 95% confidence level. The class probabilities indicate 38% of the respondents were members of class 1 (control most important), 24% of class 2 (route of administration most important), 15% of class 3 (cost most important), 13% of class 4 (weight is most important) and the remaining 10% of class 5 (hypoglycemia most important). Those individuals in class 1 (control) tended to have lower levels of HbA1c (4-7%) and a high household income (>$50,000). Class 2 individuals (route of administration) tend to have a moderate income ($20-50,000) and were insulin näive. Class 3 individuals (cost) also tend to have a moderate income. Class 5 individuals (hypoglycemia) tend to have high levels of HbA1c (>10%). Conclusions:
The identification of latent classes suggests the existence of heterogeneity in patient's preferences for different routes of insulin delivery and their associated risks and benefit. This underlines the importance of accounting for preference heterogeneity when analyzing data from discrete choice experiments.
Candidate for the Lee B. Lusted Student Prize Competition
