Preference heterogeneity and choice of cardiac rehabilitation program: Results from a discrete choice experiment
Introduction
During the last two decades, the discrete choice experiment (DCE) has gained attention as an approach for measuring preferences within the area of health care; see Ryan and Gerard [1] for a recent overview of the application of DCE in health care. Although the understanding of the DCE has increased considerably, health economists employing the method are continuously confronted with new issues that need to be considered—issues related to the design, the theoretical foundation and the modelling of the DCE. One such issue concerns the use of more generalised discrete choice models such as the random parameter logit (RPL)—also termed mixed logit1. RPL generalizes standard logit by allowing the parameters associated with the observed variables (i.e. attributes in DCEs) to vary randomly across individuals. Moreover, RPL allows modelling of repeated choices by the same individual in the data which is often the case in DCEs [2]. Recent advantages in computer speeds and in the understanding of simulation techniques to approximate integrals have facilitated the application of the structure; however, there still remain unresolved issues in using this technique. Use of RPL to model DCE data (and data in general) have been widely applied in transportation and environmental economics leading to a considerable increase in model fit compared to standard logit; e.g. [3], [4], [5], [6], [7], [8]. In contrast, the application of RPL to DCEs in health economics has so far been limited [9]; examples of studies include [10], [11], [12]. This, however, is likely to change in the coming years as the awareness of the model increases and as RPL becomes an integrated part of standard software. Other RPL applications within health economics include [13], [14], [15].
The objective of this paper is to examine patients’ choice of cardiac rehabilitations activities from DCE data using a RPL specification. We estimate a RPL model with policy relevant person specific interactions. The obtained results from the RPL model are compared with the results from a standard logit model. We highlight the added information that the RPL model provides and discuss the policy relevant interpretations that may arise. Finally we discuss some of the potential pitfalls of the RPL that researchers need to recognise.
The data used in this paper is based on a DCE study examining Danish cardiac patients’ preferences for the offer of various cardiac rehabilitation activities. The aim of the DCE survey was to provide information to health care professionals and health care decision makers on former cardiac patients’ preferences for cardiac rehabilitation (CR). The DCE was part of a larger study performed in cooperation between Research Centre for Prevention and Health, Copenhagen County and three cardiology departments in the university hospitals of the Copenhagen County1.
Participation in a cardiac rehabilitation program is usually offered to patients who have suffered coronary artery disease. CR programs have been found to be an effective secondary prevention of coronary heart disease [16], [17], [18], [19], [20], [21]. Common to the previous CR studies has been an implicit aim of maximizing the individual's health. From a welfare economic perspective, however, anything that affects an individual's preferences for health services is relevant to the measurement of utility and should, therefore, be included in the valuation of the benefits of the health service. In the welfare economic approach, the applied decision rule involves maximizing the individual's utility2.
Referral of patients to CR is made by health professionals and patient participation in CR is of course optional. From a welfare economic perspective the latter implies that the patient will not attend rehabilitation if the utility derived from the rehabilitation program is less than or equal to zero. Factors that might influence the patients’ utility function include factors related to the rehabilitation program as well as person specific characteristics. Hence, the individual's utility not only depends on the health effects of the intervention but equally on other factors such as preferences for the activities per se. This implies that the patient might consider not attending a rehabilitation activity even if it has a positive effect on his health. Focusing entirely on health effects is, thus, misleading as patients’ preferences may affect participation and consequently final health outcomes.
Section snippets
The discrete choice experiment
The CR programs in the DCE choice task were described by way of five possible activities/attributes); see Table 1 for an overview of attributes. The attributes and their descriptions were chosen such that they in the best possible way resembled the actual CR programs performed at the hospitals in Copenhagen County. The five attributes produced a full factorial design of 32 alternative rehabilitation programs. The alternatives were paired using an experimental design with SAS that maximizes
Results
Of the 742 former cardiac patients who received the questionnaire, 551 patients returned the questionnaire of which 511 of the respondents had answered at least one of the choice questions. This led to a response rate of 69%.
The estimation results for the RPL model at convergence and the corresponding standard binary logit model is reported in Table 2. For the RPL model we find that the loglikelihood index (reported as adjusted pseudo R2) increases considerably compared to the standard logit.
Discussion
The group meetings attribute is found to be insignificant in the standard logit model. Since the logit model assumes homogeneity in preferences (i.e. constant parameter values across individuals) the subsequent interpretation must be that group meetings are not deemed important to patients in choice of rehabilitation. Relying solely on the estimate from the standard logit, the policy recommendation may be not to offer group meetings at all. Turning to the RPL model we observe that the mean
Conclusion
The present paper focused on the elicitation of preferences for the offer of various CR activities at the hospitals in Copenhagen County using the RPL approach. The paper provides information on patients’ preferences for CR for decision makers at the clinical level as well as the funding level that can be used to enhance patients’ participation in CR.
Our findings indicate that preferences for rehabilitation activities in general are low among patients aged 76 and that smokers value physical
References (38)
- et al.
Valuing wetland attributes: an application of choice experiment
Ecological Economics
(2003) - et al.
Willingness-to-pay for reducing fatal accident risk in urban areas: an internet-based web page stated preference survey
Accident Analysis and Prevention
(2004) A mixed logit based valuation of frequency in civil aviation from SP-data
Transportation Research Part E-Logistics and Transportation Review
(2006)- et al.
Using stated preference discrete choice modeling to evaluate health care programs
Journal of Business Research
(2004) - et al.
Economic evaluation of cardiac rehabilitation soon after acute myocardial infarction
American Journal of Cardiology
(1993) - et al.
Exercise-based rehabilitation for patients with coronary heart disease: systematic review and meta-analysis of randomized controlled trials
American Journal of Medicine
(2004) - et al.
Using discrete choice experiments to value health care programs: current practice and future research reflections
Applied Health Econonomic and Health Policy
(2003) - et al.
Mixed logit with repeated choices: households’ choices of appliance efficiency level
Review of Economics and Statistics
(1998) Recreation demand models with taste differences over people
Land Economics
(1998)- et al.
Households’ willingness to pay for water service attributes
Environmental & Resource Economics
(2005)