Comparing different ways of asking people with type 2 diabetes about their care priorities and preferences
BACKGROUND: There is a growing interest among decision makers in medicine in using stated-preference methods to understand the perspectives of patients and other stakeholders. OBJECTIVES: We addressed evidence gaps in stated-preference methods and demonstrated good practices via applications for typ...
| Main Author: | |
|---|---|
| Corporate Author: | |
| Format: | eBook |
| Language: | English |
| Published: |
[Washington, D.C.]
Patient-Centered Outcomes Research Institute (PCORI)
[2019], 2019
|
| Series: | Final research report
|
| Online Access: | |
| Collection: | National Center for Biotechnology Information - Collection details see MPG.ReNa |
| Summary: | BACKGROUND: There is a growing interest among decision makers in medicine in using stated-preference methods to understand the perspectives of patients and other stakeholders. OBJECTIVES: We addressed evidence gaps in stated-preference methods and demonstrated good practices via applications for type 2 diabetes. We compared stated-preference methods for identifying patient priorities, measuring patient preferences, and analyzing heterogeneity of patient preferences. We assessed the relevance of our findings to patients and stakeholders. METHODS: The Diabetes Action Board (DAB) guided the study design, development, and dissemination. Using survey methodology, we conducted randomized studies to compare Likert and best-worst scaling (BWS) (case 1) in prioritizing the barriers and facilitators of diabetes self-management (ClinicalTrials.gov identifier NCT02637609). Results of Likert and BWS (case 1) measuring priorities for barriers and facilitators of diabetes self-management were highly correlated (rho = 0.97), but the Likert data failed to identify any facilitators of self-management unless a post hoc adjustment was made. Results of the DCE and BWS (case 2) measuring preferences for diabetes medications were also correlated (rho = 0.91), but latent class analysis demonstrated different decision-making heuristics in the 2 preference elicitation methods. We found that simple stratification techniques to demonstrate preference heterogeneity were flawed. We introduced new latent class techniques to reduce overfitting when analyzing preference heterogeneity. During engagement exercises to elicit opinion about qualities of preference studies, the DAB members prioritized ensuring appropriate understanding/interpretation by respondents as desirable and actionable attributes. BWS (case 1) and Likert scales are correlated, but can lead to different policy interpretations. BWS (case 2) and DCE may impact respondents' decision-making styles. Latent class techniques seem to be better than stratification for exploring preference heterogeneity, but many methods questions persist. Respondents preferred the Likert approach to BWS (case 1) but agreed more with the interpretation of the BWS (case 1) results than the Likert results. LIMITATIONS AND SUBPOPULATION CONSIDERATIONS: We have compared several common stated-preference methods (Likert vs BWS [case 1] and BWS vs DCE [case 2]), but many others have yet to be explored. Studies with larger sample sizes may also be needed, especially to further examine techniques to identify preference heterogeneity In the same survey, we also conducted randomized studies to compare discrete-choice experiment (DCE) and BWS (case 2) to measure patient preferences for medications for type 2 diabetes (ClinicalTrials.gov identifier NCT02637622). We also compared stratification and latent class approaches for modeling of preference heterogeneity. We engaged the DAB to discuss the results of these comparisons of methods and to discuss desirable characteristics of preference studies in general. Using a follow-on survey, we informed a sample of original participants and the general population about the study results to assess their beliefs and preferences for different methods that can elicit patient priorities. RESULTS: For the first survey, 1103 people with type 2 diabetes participated. For the follow-on survey, 629 people participated (315 original participants with type 2 diabetes and 314 general population). Follow-on survey participants were more likely to agree or strongly agree that the Likert scale was easier to understand and answer than the BWS (case 1; P < .001). While more than half (63.75%) of the participants said that they preferred the Likert rating scales over the BWS (case 1), more than half of the participants (62.64%) also believed that the BWS (case 1) results for the barriers and facilitators more closely reflected their priorities. Finally, when choosing desirable characteristics of a stated-preference study, most participants (76%) valued quality indicators, but 24% of respondents strongly valued incentives (eg, payments, less time burden). CONCLUSIONS: While there was an acceptable degree of concordance between the stated-preference methods that we compared, we found some key differences. |
|---|---|
| Physical Description: | 1 PDF file (67 pages) illustrations |