GROUP CLUSTERING OF DCE-ELICITED PREFERENCES PREDICTS ADHERENCE TO ASTHMA PREVENTER MEDICATION

Purpose: Patients’ decisions about whether or not to take prescribed medication (i.e., to adhere) are shaped not only by their knowledge and beliefs about their condition and its treatment options, but also by what they value in these domains.  Such decisions often involve trade-offs, and so economic choice methodology has been used as a means to elucidate preference-based decisions. However, individuals’ value systems differ and this particularized information is obscured in the group-level parameters generated by standard choice tasks. Heretofore, the dominant approach to assessing the heterogeneity of preferences has been straightforward rating or ranking tasks, which cannot educe trade-offs.  This study represents an attempt to extract variations in preference parameters within a patient group from discrete choice (DCE) methodology and to assess their contributions to adherence decisions.

Method: 140 patients with asthma were asked to select which hypothetical medication they would choose from among eight choice sets that varied along seven attributes (Long Term Efficacy, Short Term Efficacy, Immediate Relief, Number of Inhalers, Steroid Dose, Administration Time, and Side Effects). They also rated the importance of each of these attributes on 10-point Likert-type rating scales.

Result: Data from the DCE were subjected to a latent cluster analysis which suggested four distinct groups of patients: Those whose choices are based on (1) long term benefits, (2) medication side effects, (3) a trade-off between side effects and efficacy and (4) all attributes equally. Based on stepwise regression analyses, membership in the group valuing  long-term outcomes predicted an additional 8% of the variance (ΔR² = .08, F = 4.0, p < .001)  in patient-reported adherence to asthma preventer medication above and beyond that accounted for asthma knowledge and beliefs alone.  Notably, none of the rating scale items, including the one tapping long term effectiveness, correlated with adherence.

Conclusion: We have demonstrated an approach to elucidating patient variations in preferences from standard DCE methodology. Moreover, we have shown that these preferences better predict behaviour than do preferences gleaned from the more standard rating scales  that lack  the capacity to capture the trade-offs inherent in most real-world decisions.