Developing and applying discrete choice experiments (DCEs) to inform pharmacy policy

Introduction. In the United Kingdom (UK), recent policy developments include an increased role for the community pharmacist in the management of drug therapy in primary care. When embarking on such service redesign, it is recognised that patient preferences must be considered. The quality adjusted life years (QALY) approach is the common economic approach to considering preferences in economic evaluations. However, it is limited to consideration of only health outcomes. Other approaches widely used in health services research, such as satisfaction surveys, do not find a place in the economic evaluation framework because they do not incorporate any notion of opportunity cost or strength of preference. Another measure of benefit is required to go beyond the scope of the QALY and place a value on non-health outcomes (e.g. information, reassurance, regret, anxiety) and process attributes (e.g. waiting time at the health care centre, staff involved, waiting time for test results), as well as estimate trade-offs between such aspects of care. One technique adopted over the past decade in health economics to go beyond health outcome is Discrete Choice Experiment (DCEs). This is an attribute-based technique for eliciting preferences. When a price proxy is included, willingness to pay (WTP), a monetary measure of benefit, can be estimated for both marginal changes in service characteristics and changes in the overall service. Whilst the last 15 years has seen an increased use of DCEs in health care, there has been limited application in pharmacy. The Community Pharmacy Medicines Management Project (Medman study) constituted the perfect platform for this PhD thesis. It was a large, randomised, controlled trial (RCT) to evaluate the introduction of a pharmacy-led medicines management service for patients with coronary heart disease (CHD). It allowed comparison of results between traditional outcome measures (such as clinical outcomes, QALYs and satisfaction) and a broader measure of benefit using the DCE. As well as considering the relevance of DCE to policy making, methodological issues in their development were also considered. Aim and objectives. The aims of this PhD programme are: to consider the implications of taking a broader patient centred approach (which goes beyond health outcomes) to evaluating a proposed policy change in community pharmacy; and to develop and apply the DCE methodology to pharmacy. The specific objectives of the thesis are: • to conduct a review of the literature relating to benefit evaluation and its application to the economic evaluation framework in pharmacy; • to apply traditional benefit evaluation approaches, including clinical outcomes, QALYs and a satisfaction survey to a pharmacy setting; • to develop an alternative approach, DCEs, to value benefits, looking at different design approaches when defining multiple choice DCEs and addressing welfare estimation issues when analysing the data; • to apply DCE to pharmacy within the Medman study and to compare the results with these of clinical and QALY approaches within an economic evaluation framework; • to consider the impact on policy when different approaches are applied. Methods. A fully comprehensive review of the health care literature was conducted to identify possible methods for valuing community pharmacy-based services and their integration into economic evaluation. Traditional benefit measures in pharmacy practice were applied to the Medman study and incorporated into an economic evaluation framework. A tool to measure appropriateness of treatment was developed and applied as a proxy for clinical outcome. Health-related quality of life measures (EQ-5D and SF-6D) were also adopted, together with a satisfaction tool developed for the trial. To quantify patients’ preferences for the proposed, extended role of the pharmacist in the management of drug therapy, a DCE approach was employed. Results from patient questionnaires conducted in the main trial were used to inform the attributes of the DCE exercise using a novel approach to attribute definition, namely, factor analysis. The DCE questionnaire was first piloted to allow its readability to be tested, as well as its ability to be completed. The pilot also gave the opportunity to consider any necessary rewording and the appropriateness of the attribute levels. Crucially, it also allowed identification of the most suitable design to employ according to design optimality. Consideration was given to the impact of different DCE design approaches on goodness of fit and WTP. The design with best statistical efficiency (defined according to D-efficiency) and goodness of fit was chosen for the main survey. The main DCE questionnaire was administered to all RCT subjects two years after the start of the Medman trial. Preferences amongst three groups were compared: control (receiving usual care), intervention all (receiving either usual care or Medman service) and intervention still receiving the service (receiving only Medman service). Analysis of the responses compared three different regression models when considering multiple-choice data: conditional logit, nested logit and mixed logit. The selection of the most appropriate models and the impact of their substitution patterns on WTP were addressed. Finally, the integration of different approaches and models of analysis of DCE designs in the economic evaluation framework was considered and their application at the policy level discussed. Results. Findings from the literature review confirmed clinical outcome measures as the most common benefit evaluation in the pharmacy setting (62/103, 60%), followed by satisfaction (39/103, 38%) and health-related quality of life (HR-QOL) applications (33/103, 32%). Only 13% (13/103) of studies presented WTP applications. A limited number of papers inserted health-related outcomes into a full, economic evaluation framework (13/103, 13%). Only 11% (11/103) of the papers included a multiple approach to benefit evaluation comparing clinical, HR-QOL and satisfaction outcomes. Of these, less than half presented similar directions in the results across techniques, while the remaining applications had discrepancies between the different approaches employed. The Medman study offered the perfect platform to compare different approaches and their contribution to economic evaluation. Within the Medman study, standard outcome measures, such as clinical and QALYs, did not demonstrate any benefit to patients receiving the new service compared to the controls. However, patients reported increased satisfaction with the new service. This confirmed the need to identify an alternative measure of benefit, such as DCE, to inform the weights attached to the various dimensions that make up ‘satisfaction’. Results from DCE development in the pilot phase showed that sample size and inclusion of ‘current’ options are likely to influence the statistical properties of the design. The models not incorporating interaction terms had higher statistical efficiency and a better goodness of fit. Resulting WTP differed across the estimated models. Results from the main phase analysis showed that patients most valued the service they experienced (status quo bias). When receiving the Medman service, they preferred it to any other options. The preferred model to analyse the data depended on the group being analysed. WTP estimates were sensitive to the modelling. Main findings showed that, when moving from usual care to their ‘current’ Medman service, intervention groups were willing to pay £19 or £69 (intervention all, intervention receiving the service, preferred model of analysis). The control group preferred their ‘current’ usual care rather than moving to a Medman service (£17, preferred model of analysis). Results from the trial economic evaluation, where a cost minimisation analysis was employed, suggested that this type of medicines management service by community pharmacists, as assessed by short-term health outcomes, did not represent an efficient use of resources for the NHS. The inclusion of DCEs in the evaluation showed that both health outcomes and non-health outcomes need to be considered when introducing a new service. For those who experienced the Medman service, benefits were likely to balance the increase in costs incurred by the introduction of the service. CBA results were sensitive to the DCE models of analysis applied. Conclusions. Alternative approaches to benefit evaluations were compared to inform patient-centred decision making. DCE was demonstrated to overcome the potential limitations of traditional measures. Policy decisions supporting the delivery of a pharmaceutical service should consider economic evaluations with a broader perspective on both health and non-health benefits for patients. Experience influenced how much respondents valued the service. The DCE methodology was developed and sensitivity of the welfare estimates to both the design approaches and the modelling of analysis adopted was demonstrated. Future research should consider issues related to external validity of the DCE stated preferences results