The impact of violations of expected utility theory on choices in the face of multiple risks. Gonzalez Sepulveda, J. M., Van Houtven, G., Reed, S. D., Webster, S., & Johnson, F. R. (2024). Journal of Choice Modelling, 53.

The use of preference information to infer risk tolerance has increased in recent years as a means to inform benefit-risk evaluations in regulatory and medical decision-making. However, a framework for measuring tolerance for multiple uncertain outcomes has not been formalized when choices do not comply with Expected Utility Theory (EUT). We developed a formal analytical framework for measuring preferences through choices under uncertainty with multiple risks. Based on the analytic framework, we find that violations of EUT can lead to interaction effects between uncertain outcomes, not just nonlinearities in the disutility of risks. Our framework also implies that measures of risk tolerance derived from utility, such as maximum-acceptable risk, must consider all relevant risks jointly if their combined effect on choices is expected to violate the EUT. Somewhat reassuringly, however, we find that cross-outcome effects are expected to be negligible when the probabilities of other outcomes approach certainty. Finally, we identify a simple test that can help evaluate whether preferences for one uncertain outcome are affected by other uncertain outcomes.

To pool or not to pool: Accounting for task non-attendance in subgroup analysis. Gonzalez, J. M., Johnson, F. R., & Finkelstein, E. (2024). Journal of Choice Modelling, 51.

Pooling data from different subgroups offers the advantages of shrinking standard errors and simplifying characterization of the data structure. The ability to pool data also facilitates meta-analysis, which enables the evaluation of consensus among multiple studies and informs the transfer of benefits to new choice settings. Testing for poolability requires accounting for differences in response variance or scale among subgroups. This is commonly done by assuming a single scale factor within each subgroup of interest. This assumption may not hold for many subgroups, especially when task non-attendance is present. We use data from a prior DCE study to show that task non-attendance, and by extension, the assumption of a single scale factor across subgroups, can lead to inaccurate conclusions when determining poolability. To address this concern, we propose a latent-class/random-parameters Logit (LCRP) model specification that accommodates task non-attendance or other causes of scale differences within subgroups and directly tests for poolability.

Self-stigmatization and treatment preferences: Measuring the impact of treatment labels on choices for depression medications. Gonzalez Sepulveda JM, Townsend M, Waters HC, Brubaker M, Wallace M, Johnson R. (2024). PLoS One, 19(9).

We developed a discrete-choice experiment (DCE) survey instrument that asked respondents to make choices between hypothetical treatments for major depressive disorder (MDD). The choice questions mimicked the information presented in product inserts and required systematic tradeoffs between treatment efficacy, treatment type, and indication. We calculated how many patients were willing to forgo efficacy to avoid treatments with information associated with self-stigmatization, and how much efficacy they were willing to forgo. We also evaluated the impact of contextualizing the treatment information to reduce self-stigmatization by randomizing respondents who received additional context. Product-label treatment indication can potentially lead to patient self-stigmatization, as shown by patients' avoidance of treatments that are also used to treat schizophrenia. Although the effect appears to be relatively small, the results suggest that label effects may be pervasive.

What Can Discrete-Choice Experiments Tell Us about Patient Preferences? An Introduction to Quantitative Analysis of Choice Data. Johnson FR, Adamowicz W, Groothuis-Oudshoorn C. (2024). Patient.

This paper provides an introduction to the statistical analysis of choice data, using example data from a simple discrete choice experiment (DCE). It describes the layout of the analysis dataset, the types of variables contained in the dataset, and how to identify response patterns in the data that indicate data quality. Model-specification options include linear models with continuous attribute levels, as well as non-linear models with continuous and categorical attribute levels. The advantages and disadvantages of conditional logit, mixed logit, and latent-class analysis are discussed and illustrated using the example of DCE data. Readers are provided with links to various software programs for analyzing choice data. References are provided on topics for which there currently is limited consensus and on more advanced techniques to guide readers interested in exploring choice-modeling challenges in greater depth. Supplementary materials include the simulated example data used to illustrate modeling approaches, together with R and MATLAB code to reproduce the estimates shown.

Getting it right with discrete choice experiments: Are we hot or cold? Ozdemir, S., Gonzalez, J. M., Bansal, P., Huynh, V. A., Sng, B. L., & Finkelstein, E. (2024). Soc Sci Med, 348, 116850.

In 2019, the US Food and Drug Administration issued a warning that symptomatic relief from claudication using paclitaxel-coated devices might be associated with an increase in mortality over 5 years. We designed a discrete-choice experiment (DCE) to quantify the trade-offs that patients would accept between a decreased risk of clinically driven target-vessel revascularization (CDTVR) and an increased mortality risk. Our work serves as a model for future collaborative efforts, involving patients, clinicians, researchers, regulatory partners, and industry, to further define patient preferences and care goals.

Quantifying patients' preferences on tradeoffs between mortality risk and reduced need for target vessel revascularization for claudication. Reed SD, Sutphin J, Wallace MJ, et al. (2024). Vascular Medicine, 29(6):675-683.

Patients making benefit-risk tradeoffs must consider the magnitude of risks, which requires considering probabilities or percentages. Good risk-communication practices are essential to ensure patients' understanding; however, prior research has not evaluated risk-communication approaches in the context of discrete-choice experiment questions, where multiple risks are presented for alternative treatment profiles. We designed a six-arm randomized study to compare different approaches to presenting probabilistic treatment risks and evaluated their impact on internal validity across discrete choice experiments (DCE) questions.

Using separate single-outcome risk presentations instead of integrated multi-outcome formats improves comprehension in discrete choice experiments. Wallace MJ, Weissler EH, Yang J-C, Brotzman L, Corriere MA, Secemsky EA, Sutphin J, Johnson FR, Gonzalez JM, Tarver ME, Saha A, Chen AL, Gebben DJ, Malone M, Farb A, Babalola O, Rorer EM, Zikmund-Fisher BJ, Reed SD. (2024). Med Decis Making.

Evaluating poolability by considering subgroup means, as is usually done, can provide an incomplete assessment of differences in preferences across several groups or samples. Comparing means only can overlook the fact that subgroups can share the same set of preference types, even if distributed differently. We developed a novel approach to control for distributional differences in preferences across subgroups. In a study of family caregivers across five countries, we found that our approach afforded us greater flexibility to accommodate respondents from multiple countries within the same model of preferences. This, in turn, helped us shrink standard errors of the preference estimates without increasing the study sample size.

Use of patient preferences data regarding multiple risks to inform regulatory decisions. Montano-Campos JF, Gonzalez JM, Rickert T, Fairchild AO, Levitan B, Reed SD. (2023). MDM Policy Pract, 8(1):238146832211487.

Medication non-adherence remains a significant challenge in healthcare, impacting treatment outcomes and the overall effectiveness of medical interventions. This article introduces a novel approach to understanding and predicting medication non-adherence by integrating patient preferences with patients' beliefs, efficacy expectations, and perceived costs. Existing theoretical models often fall short in quantifying the impact of barrier removal on medication adherence and struggle to address cases where patients consciously choose not to follow prescribed medication regimens. In response to these limitations, this study presents an empirical framework that aims to provide a quantifiable model for predicting non-adherence at both individual and population levels under various scenarios.

Method for calculating the Simultaneous Maximum Acceptable Risk Threshold (SMART) from discrete-choice experiment benefit-risk studies. Fairchild AO, Reed SD, Gonzalez JM. (2023). Med Decis Making, 43(2):227–38.

Conventional estimates of maximum-acceptable risks (MARs) evaluate only one adverse-event risk at a time, but patients are often exposed to multiple treatment-related adverse-event risks. To account for this, we developed Simultaneous Maximum Acceptable Risk Thresholds (SMART) to represent combinations of maximum risk levels that would be jointly acceptable for specific treatment benefits.

A quantitative framework for medication non-adherence: Integrating patient treatment expectations and preferences. Muiruri C, van den Broek-Altenburg E, Bosworth H, Cené C, Gonzalez J. (2023). Patient Prefer Adherence, 17:3135–45.

We extended methods for SMARTs to produce SMART contours that account for uncertainty associated with acceptable tradeoffs between benefits and two adverse-event risks. In this paper, we apply this approach to three previously published patient-preference studies.

Patient-preference diagnostics: Adapting stated-preference methods to inform effective shared decision making. Gonzalez Sepulveda JM, Johnson FR, Reed SD, Muiruri C, Hutyra CA, Mather RC III. (2023). Med Decis Making, 43(2):214–26.