Discrete choice experiments in pharmacy: a review of the literature



Ms Pradnya Naik-Panvelkar, Faculty of Pharmacy, The University of Sydney, Room S114, A15 Pharmacy, The University of Sydney, Sydney, NSW 2006, Australia.

E-mail: pradnya.naikpanvelkar@sydney.edu.au



Discrete choice experiments (DCEs) have been widely used to elicit patient preferences for various healthcare services and interventions. The aim of our study was to conduct an in-depth scoping review of the literature and provide a current overview of the progressive application of DCEs within the field of pharmacy.


Electronic databases (MEDLINE, EMBASE, SCOPUS, ECONLIT) were searched (January 1990–August 2011) to identify published English language studies using DCEs within the pharmacy context. Data were abstracted with respect to DCE methodology and application to pharmacy.

Key findings

Our search identified 12 studies. The DCE methodology was utilised to elicit preferences for different aspects of pharmacy products, therapy or services. Preferences were elicited from either patients or pharmacists, with just two studies incorporating the views of both. Most reviewed studies examined preferences for process-related or provider-related aspects with a lesser focus on health outcomes. Monetary attributes were considered to be important by most patients and pharmacists in the studies reviewed. Logit, probit or multinomial logit models were most commonly employed for estimation.


Our study showed that the pharmacy profession has adopted the DCE methodology consistent with the general health DCEs although the number of studies is quite limited. Future studies need to examine preferences of both patients and providers for particular products or disease-state management services. Incorporation of health outcome attributes in the design, testing for external validity and the incorporation of DCE results in economic evaluation framework to inform pharmacy policy remain important areas for future research.


Community pharmacy forms a major component of the primary healthcare system in most developed nations. Pharmacists have also become the most accessible and conveniently located points of contact for individuals within the healthcare system.[1, 2] Traditionally, pharmacists have been mainly involved in the dispensing of medications. Increasingly, however, their role has diversified and pharmacists are now involved in the provision of a wide range of healthcare services in the community ranging from drug information provision, health screening, medication management, disease-state management and provision of palliative care.[2, 3] Several large community pharmacy-based studies (including some robust randomised controlled trials) have been conducted globally.[4-14] A substantial number of services targeting disease-state management have demonstrated the potential benefit of such pharmacist-delivered services both clinically and/or economically.[4, 5, 8-15] In fact, some of these pharmacy-based services, such as repeat dispensing, smoking cessation and medication reviews, have also been translated into sustainable services in countries like the UK, often as part of their national pharmacy contracts.[16, 17] However, evidence of improvements in health outcomes from pharmacist-led services is often mixed.[18] This, coupled with the diversity of research approaches and methodologies, makes it difficult to reach an overall conclusion about the impact of pharmacists' healthcare service delivery on patient outcomes.

The concept of ‘patient-centred’ health care has gained an important focus especially in the case of patients with chronic conditions.[19] There is increasing evidence to suggest that understanding a patient's preferences, views and needs, and organising healthcare services to match these aspects together with clinical viewpoints, can lead to improved health and economic outcomes.[20] Previous studies have demonstrated that patient preferences for healthcare services and interventions can impact on their willingness to use services.[21] Thus, investigating the patient perspective can also provide an insight into which health-service aspects are perceived to be of value to patients and can influence their decisions to use/uptake the services, which in turn may reflect on the sustainability and economic viability of these healthcare services.

Measurement of patient satisfaction is one of the most commonly employed methods for eliciting the patients' perspectives for healthcare services as well as pharmacy-based services. However, this technique has several drawbacks including the lack of a consensus regarding a theoretical framework for patient satisfaction, the use of self-developed, non-validated ad-hoc instruments for measuring satisfaction, and issues such as high baseline satisfaction that limit the ability to detect real differences in patients' opinions.[22] Besides these methodological constraints, satisfaction surveys are unable to provide information about the potential value of future services, the aspects/attributes of these services that drive satisfaction levels and the relative importance attached to these aspects/attributes; i.e. information that can provide guidance on the optimal allocation of resources especially in a budget-constrained health system.[23] Further, satisfaction surveys cannot be used to inform economic evaluation and thus are limited in their ability to bring the patients perspective into policy decision making.[24]

Novel preference elicitation techniques such as ‘stated preference methods’, where individuals state what they would choose when offered a product or service, are becoming increasingly popular in the health sector.[23, 25, 26] Stated preferences, unlike revealed preferences, get respondents to make choices based on hypothetical scenarios rather than observing them when making an actual or real-life choice.[23, 25, 26] The last decade has seen an increased use of these methods including conjoint analysis and discrete choice experiments (DCEs) to elicit preferences for healthcare products and services.[25, 27-30] These two methods have a common format in terms of the underlying attributes, use of experimental design methods for instrument design and utilisation of statistical models to determine the importance of each attribute to preferences, although they differ substantially with respect to their theoretical framework as well as preference elicitation.[29] Conjoint analysis involves asking individuals to rank or rate the alternatives provided while DCEs elicit preferences by asking individuals to choose one alternative from those presented.[23]

Discrete choice experiments have their origins in mathematical psychology and have been successfully used in market research, transport economics and environmental economics.[31] Applications in health have been relatively recent since the early 1990s.[25, 29] Within the context of health care these techniques have been successfully applied in several areas such as valuing of patient experience factors, valuing health outcomes, trade-offs between health outcomes and experience factors, job-choices, health provider's preferences for treatments or screening and developing priority setting frameworks.[30]

The DCEs are based on the random utility (RU) framework and assume that a healthcare service can be described by various attributes or characteristics and the extent to which respondents' value the service depends on the level of these attributes.[23, 26] Thus, when offered a choice, respondents choose the alternative that they believe will provide them with the highest value or utility depending on the level and combination of service attributes.[23, 26] The DCE techniques have been used to establish the strength of preferences for healthcare services, to identify which attributes are important to respondents, the relative importance of the different attributes of the service as well as the trade-offs that respondents are willing to make, i.e. choosing one attribute and forsaking another when making a choice.[23, 26] Further, DCEs have also been used in configuring optimal service design, predicting demand and uptake of services under differing scenarios, estimation of willingness-to-pay (WTP) when a monetary/cost attribute is included and informing economic evaluation modelling (for example cost-benefit analysis).[25, 29, 32]

Pharmacy-delivered specialised services are a relatively novel paradigm and are also quite complex in nature. Traditionally, pharmacy practice researchers have often measured patient satisfaction with pharmacy-based services.[22] Measuring patient preferences for such specialised services using techniques such as DCEs can provide important information which can assist in the development of optimal services that patients will use, are willing to pay for, and thus are sustainable and economically viable in the future. An example of a hypothetical DCE design for a pharmacy-delivered specialised asthma service, including possible service attributes and levels, has been illustrated in Figure 1.[33]

Figure 1.

Example of attributes and levels from a comprehensive pharmacy-based asthma service.

Payne and Elliot[23] need to be acknowledged for bringing the DCE technique to the notice of the pharmacy practice community by the publication of their comprehensive review. Their review explains how this technique can be effectively applied in the measurement of preferences for pharmacy services and also identifies applications of DCEs in health care by conducting a systematic search of the literature from January 2003 until May 2004.[23] We believe that publication of this timely review in 2005 may have encouraged the adoption of the DCE technique by several researchers in the field of pharmacy since. Hence, the aim of our study was to conduct an in-depth scoping review of the literature and provide a current overview of the progressive application of DCEs within the field of pharmacy


Databases searched and search strategy

An extensive search of the literature was conducted to identify published English language studies using DCEs within the pharmacy context. The following databases were searched between January 1990 and August 2011: MEDLINE, EMBASE, SCOPUS and ECONLIT. Search strategies were formulated for individual databases using the following keywords: ‘discrete choice’ or ‘discrete choice experiment’ or ‘discrete choice analysis’ or ‘discrete choice modelling’ or ‘conjoint’ or ‘conjoint analysis’ or ‘stated preference method’ AND ‘pharmacy’ or ‘pharmacies’ or ‘community pharmacy’ or ‘pharmacist’ or ‘pharmacy service’ or ‘pharmaceutical service’ or ‘pharmaceutical care’ or ‘pharmaceutical program’ or ‘specialized service’ or ‘cognitive service’ or ‘disease management’ or ‘chemist’.

Inclusion and exclusion criteria

Studies were limited to those that used choice-based techniques, were applicable to pharmacy and were written in English. Reviews, conference papers, commentaries and letters were excluded.

  • ● Choice-based studies: We limited our analyses to utility-based choice studies including discrete choice experiments and conjoint analysis with a choice-based response format. Studies that presented methodological issues or used conjoint analysis with ranking or rating were not included.
  • ● Applicability to pharmacy: This included choice-based studies that elicited (1) patient preferences for pharmacy-delivered products/services, pharmacies and/or pharmacists; (2) pharmacists preferences for products, treatments, services or job-choices; (3) preferences of both, patients and pharmacists; or (4) informed pharmacy policy or the decision-making framework.

Study selection and analysis

Two authors independently reviewed titles and abstracts and all potential articles meeting the inclusion criteria were downloaded/obtained for additional review. The two authors conducted data abstraction independently and in duplicate and reached consensus through discussion about any disagreement. Included papers were organised and analysed for the following:

Methodology: how the included studies conducted different stages of the discrete choice experiment

A DCE is conducted in several stages.[23, 26] Readers are referred to Ryan et al.[26] and Payne and Elliot[23] for a description of the different stages.

Selection of attributes and levels

The first step of a DCE is to identify attributes and levels that adequately describe the service or intervention to be evaluated.

Developing the experimental design and construction of choice sets

The next important step of the DCE methodology is the development of the experimental design, hypothetical scenarios and construction of choice sets. The identified attributes and levels are formed into scenarios. The number of possible scenarios that must be included in the experiment to incorporate the total number of combinations of attributes and levels is called a ‘full factorial design’. A full factorial design that can elicit the preferences for all combinations of attributes and levels can result in an unmanageable number of scenarios.[23, 26] Experimental design methods can help reduce this number by creating smaller fractional factorial designs, e.g. orthogonal designs. These designs enable the estimation of main effects, i.e. the effect of each independent variable on the dependent variable, as well as possible interactions, i.e. when preferences for one attribute depend on the level of another.[30] Orthogonal designs can be obtained from design catalogues, statistical software programs or websites and have the properties of orthogonality (where attributes are statistically independent of each other) and level balance (where levels of attributes appear an equal number of times).[30] Following the development of the experimental design, choice sets need to be constructed, especially when two or more alternatives are present.

Questionnaire design for discrete choice experiment and measurement of preferences

The development of the experimental design is followed by the designing of the DCE questionnaire, pilot testing and data collection.

Data analysis/estimation

Following administration of DCE questionnaires and data collection, the next step is discrete choice modelling within a RU framework to analyse the responses obtained from the DCEs.

The included articles were reviewed and individual details of the DCE methodological steps utilised (including the number of attributes, type of attributes, design type, design plan, design source, method of constructing choice sets, mode of administration of questionnaire, estimation method used and validity tests) were identified and reported.

Pharmacy application: how the included studies were applied to the field of pharmacy

The included studies were then evaluated for their application within the field of pharmacy with respect to the focus of preference (patient, provider, both), focus of study, attributes used, key findings and conclusions. Each paper was also assessed using a ‘checklist of factors to be considered when conducting a DCE’ adapted from Lancsar and Louviere.[25] Please refer to Figure 2 for more details.

Figure 2.

Factors to be considered when designing and conducting a discrete choice experiment (DCE). Adapted from Lancsar and Louviere.[25]


The search generated 243 possible articles. After elimination of duplicates and screening as per inclusion/exclusion criteria (Figure 3),[34] 12 studies were retrieved which were included in the review.[35-46]

Figure 3.

Flowchart describing the selection of studies that were included in the review. Adapted from the PRISMA 2009 flow diagram presented in Moher et al.[34]

Table 1 summarises the background of DCE studies reviewed. The majority of the pharmacy-related DCE studies were conducted in the UK and almost all the studies were published in the last decade, of which 10 were published after 2005. Studies elicited patient preferences or pharmacist preferences or preferences of both for various pharmacy-related products and services. There were no studies that incorporated DCEs into a decision-making framework to inform pharmacy policy.

Table 1. Background of discrete choice experiment studies
ItemCategoryNumber of studies
Country of originCanada1
The Netherlands2
Year of publication1990–2000
Source of preferencePatient/user6
Healthcare provider4
Patient and provider2

Results based on stages of discrete choice experiments utilised

The reviewed studies were examined for the different DCE stages conducted and the results have been reported in Table 2.

Table 2. Discrete choice experiment methodology
ItemCategoryNumber of studies
  1. DCE, discrete choice experiment; NGene Software (ChoiceMetrics Pty Ltd, www.choice-metrics.com); SAS, Statistical Analysis System (Cary, North Carolina, USA); Sawtooth (Orem, Utah, USA); SPEED, Stated Preference Experiment Editor Designer (Hague Consulting Group, The Netherlands); SPSS (Statistical Package for Social Services, IBM, Armonk, New York, USA).
Selection of attributes and levels  
Number of attributes2–31
10 or more1
Attributes coveredMonetary measure11
Health outcome4
Process related9
Provider related4
Experimental design and construction of choice sets  
Design typeFull factorial
Fractional factorial10
Not clear2
Design planMain effects only (includes interactions with covariates)3
Main effects, two-way interactions2
All effects (full factorial)
Not clear7
Design sourceSoftware package/other source 
SPSS (version 11.5)1
No details
Not clear2
Method of creating choice setsOrthogonal arrays 
Single profiles (binary choices)
Random pairing1
Constant comparator pairing3
Foldover with random pairing
D-efficient designs3
Statistically efficient designs (using a priori assumptions of parameter estimates)
Not clear3
DCE questionnaire design and measurement of preferences  
Number of choices per respondentEight or less2
More than 16
Administration of questionnaireMailed questionnaire8
Interviewer administered
Computerised/internet based4
Estimation methods employed  
Estimation methodLogit1
Random effects logit
Random effects probit3
Conditional logit2
Multinomial logit4
Nested logit
Mixed multinomial logit/random parameter logit
Latent class1
Validity testsExternal validity
Internal validity 
Not clearly stated3
Face validity 
Qualitative methods for attribute and level selection7
Pre-testing questionnaires9
Not clearly reported1

Selection of attributes and levels

Table 2 shows the current trends with respect to attribute and level selection within the pharmacy context. The number of attributes in the reviewed studies varied between three and 11 with the majority of the studies including four to seven attributes. Most studies reported the incorporation of qualitative sources (such as interviews and focus groups) in the selection of attributes and levels. All reviewed studies except one[44] included some form of price proxy in terms of co-payment/cost of product or service, change in annual income or increase in health taxes. Nine studies[35-38, 40, 41, 43-45] included some type of time attribute while two studies[45, 46] had a risk attribute. Interestingly, quite a few studies had process-related and provider-related attributes while just three studies had health-outcome attributes.[36, 45, 46]

Experimental design and construction of choice sets

The majority of the studies reviewed used a fractional factorial design (Table 2). Three studies[36, 39, 45] used a main effects design only, while two studies[35, 37] used a main effects plus two-way interaction design. Several studies did not report this important design plan aspect. Software packages were most commonly employed for creating orthogonal arrays the most popular being the Statistical Analysis System (SAS; Cary, North Carolina, USA). Only one study used a catalogue for creating the orthogonal design.[45]

With respect to construction of choice sets (Table 2), the studies reviewed several different approaches such as random pairing (one study[44]), constant comparator pairing (three studies[41, 43, 46]) and foldover (two studies[36, 45]). D-efficient designs were employed by three[35, 37, 40] of the 12 studies while none of the studies used a statistically efficient design with a priori parameter assumptions. As an explanation of these individual DCE-related terms is beyond the scope of this review, the interested reader is guided to Ryan et al.[26] and Payne and Elliot[23] for more details.

Discrete choice experiment questionnaire design and measurement of preferences

Table 2 summarises current practice of DCEs in pharmacy with respect to the DCE questionnaire design and measurement of preferences, i.e. the number of choices that each respondent had to make and the mode of administration of the questionnaire.

The bulk of the studies had nine to16 choices per respondent. With respect to the mode of administration, eight[36, 39-41, 43-46] of the 12 studies were mailed, self-completed questionnaires while the remaining four studies[35, 37, 38, 42] were computer/web based (Table 2).

Estimation methods employed

The reviewed studies showed a trend towards the use of simpler models in analysing DCE data. Generally, random effects probit, conditional logit or multinomial logit (MNL) models were most commonly employed (Table 2). There was a lack of studies investigating other advanced choice models such as the nested logit model, mixed MNL model and the latent class model. Only one study[42] utilised the latent class model for investigating community pharmacists' preferences for patient-centred services and it identified the existence of preference heterogeneity in the study population, clearly important information from a policy point of view. Readers are referred to Ryan et al.[26] and Hensher et al.[31] for a description of individual models discussed above.


Table 2 reports the different aspects of validity tested in the DCE studies reviewed. None of the reviewed studies tested external validity. Internal validity tests, more commonly theoretical validity tests, were conducted by a majority of the studies especially by verifying expected coefficient signs after model estimation. Only one study[44] tested for rationality by including two dominant options.

Face validity was commonly applied to the majority of the pharmacy studies. Seven[35, 37, 38, 40, 43-45] of the 12 studies used qualitative methods to aid attribute and level selection. Pilot testing of the questionnaire was also conducted by the majority of the studies (Table 2).

Results based on application of studies to pharmacy

The reviewed studies were examined on how they were applied to pharmacy and analysed based on an adapted checklist[25] (Figure 2); the results are reported in Table 3.

Table 3. Areas of applicationThumbnail image of

Broadly, DCEs in pharmacy primarily elicited preferences for specific products, therapies and pharmacy-delivered services. Preferences were elicited from: (a) patients, i.e. current or future users of such products/services; (b) pharmacists, i.e. providers of such products/services or (c) both patients and pharmacists (Table 3).

The majority of pharmacy DCEs conducted a valuation of preferences for different aspects of pharmacy products or services. Some also evaluated their WTP by calculating the marginal rate of substitution. Most of the studies did not investigate the existence of preference heterogeneity in the study population. Further, except for two studies investigating preferences for haemophilia therapy,[45, 46] none of the studies examined the match/comparison between patient and pharmacist preferences for the same therapy or service.

Patient preferences

Patient preferences were examined by six of the 12 studies reviewed, of which five looked at preferences for pharmacy services[35-37, 39, 40] while one study investigated preferences for over-the-counter products.[38] Most studies administered the questionnaires to the general population/community users. Only one study[36] specifically recruited a convenience sample of patients from general practice settings. Aspects related to the process of delivering the service were most commonly investigated. These included ‘convenience attributes’ such as distance from home, waiting time, opening hours; ‘quality attributes’ such as certificates of quality and customer satisfaction ratings; ‘marketing attributes’ including discounts, internet service; and ‘healthcare attributes’ such as provision of medication management service. Provider-related attributes were also investigated including source of information and extent of pharmacist interaction. The majority of the studies however, did not include health-outcome related attributes. Almost all the user perspective studies had some form of ‘monetary attribute’ such as cost of service or co-payment on the part of the patient.

Choice of pharmacies and pharmacy services was greatly influenced by monetary attributes in the majority of the studies with a preference for lower cost/co-payments. Patients preferred shorter waiting times,[40] high satisfaction with pharmacy ratings, quality certificates and extended opening hours.[37] A study by Wellman and Vidican,[39] piloting the addition of medication management services in prescription benefit insurance models, demonstrated that respondents placed the greatest value on pharmacist provision of comprehensive medication management services. Several of the studies also indicated the existence of ‘status-quo’ bias (i.e. tendency to prefer their current service) among patients with respect to pharmacy choices.[35-37]

Pharmacist preferences

Four studies examined pharmacist preferences, of which only three elicited preferences with respect to pharmacy services,[41-43] while one was related to preferences for a specific new technology.[44] Grindrod et al.[42] investigated pharmacist preferences for specialised service provision and showed that pharmacists preferred to provide medication and disease-management services but did not have significant preferences for screening services. This was contrary to Scott et al.[43] who investigated pharmacist preferences for extended roles in primary care. Significant predictors of pharmacists' job choices included having an extended pharmacy team, integration with primary and secondary care as well as higher income whereas provision of chronic disease management and health promotion services was not preferred. Using the latent class model, one study[42] showed the existence of preference heterogeneity in pharmacists' preferences with pharmacists falling into three classes, indicating that groups of pharmacists may have different preferences with respect to specialised services provision.

Pharmacist preferences were mainly investigated for process-related aspects such as duration of service, type or level of service provision, setting and integration with primary/secondary care and professional service/job-related aspects including job satisfaction, educational requirements and personal income. In the majority of the studies eliciting preferences for the delivery of specialised pharmacy services (medication or disease management), income was an important attribute, with pharmacists preferring higher incomes.[41-43]

Patient and pharmacist preferences

Only two studies investigated preferences of both patients and providers for haemophilia therapy.[45, 46] These were also the only studies that included health-outcome related attributes in addition to process-related outcomes. While patients and providers showed substantial consensus for some attributes (e.g. cost), preferences for other attributes were quite different. Patients were more focused on process-related attributes as compared to providers.

The relatively few pharmacy DCE studies make it hard to make definitive conclusions about pharmacy services from both the provider and recipient viewpoints. However, a few findings may be highlighted. We found that cost plays an important role with respect to pharmacy delivered services from both the user as well as provider perspective. Further, process attributes are important although studies need to investigate the role of health outcome attributes.


We have conducted a scoping review of the current literature and identified and evaluated studies utilising the DCE methodology within the field of pharmacy. Results indicate that the pharmacy profession has adopted the DCE methodology although the number of studies is quite limited. The DCE methodology has been applied to elicit preferences for different aspects of pharmacy products, therapy or services. In the majority of the studies, preferences for particular products or services were elicited from either users (i.e. patients) or providers (i.e. pharmacists), with just two studies incorporating the views of both (patients and pharmacists). Further, most of the studies examined preferences for process-related or provider-related aspects with a lesser focus on health outcomes.

This is one of the first reviews in the literature which explores how the pharmacy-related DCEs have been designed and conducted and evaluates their progressive application in the pharmacy setting. A strength of our study was that the reviewed studies were thoroughly analysed in terms of their quality and implications. The search strategy was extensive and covered a large number of relevant databases. Further, the study highlights the value of the DCE technique and the need for utilising this technique in pharmacy practice research.

Some limitations also need to be considered. One methodological limitation was reliance on published studies, whereby we may not be accurately representing the state of DCE practice in pharmacy because of issues such as publication lag. Also the search strategy used to identify potential articles for this review was limited to the specific search terms and the databases that we used, which may have affected the articles identified. However, every effort was made to ensure that the search strategy was as comprehensive as possible. Another limitation of our study was the exclusion of the grey literature, which may have led to some relevant papers not being included in our review.

Our review of the literature showed that very few pharmacy-related DCE studies have been published in the last decade. This could be because evaluation of pharmacy products and services has been traditionally done using ‘patient satisfaction’ surveys. Whilst the construct of patient satisfaction is important, clearly there exist some issues and drawbacks with its measurement.[22] Further, measurement of patient satisfaction is limited in terms of the information that can be provided with respect to importance of attributes, trade-offs between attributes, prediction of demand and WTP estimation.[23] Satisfaction surveys also cannot be used in economic evaluation to inform pharmacy policy as they do not incorporate opportunity cost or strength of preference.[24] The DCEs, on the other hand, can overcome all these limitations of patient satisfaction measurement and also have the advantage of being used for economic evaluation and policy making, for example, within cost-benefit analyses.[32] This emphasises the need for moving beyond the commonly used satisfaction instruments and the adoption of DCEs in routine pharmacy practice research.

Overall, pharmacy-related DCEs were consistent with DCEs conducted in general health care with respect to the methodology of designing and conducting the choice experiment.[30] Similar trends between pharmacy-related DCEs and health DCEs were noted for design types and design plans used, the number of choice sets per patients, inclusion of monetary attributes in choice sets and validity tests conducted,[30] Trends, however, differed for aspects related to types of attributes selected and models used for estimation.[30]

Our study found that most of the reviewed studies focused on process attributes or provider attributes with very few health-outcome attributes. This was not the case in the general health DCE literature where the focus has been equally, or perhaps more so, on health-outcome attributes than on process attributes.[30] Also the majority of the pharmacy DCE studies investigated preferences for ‘generic’ pharmacy service provision and included ‘medication/chronic-disease management provision’ as one of the attributes. There was a lack of studies investigating ‘specific’ medication/chronic-disease management services. On the other hand, DCEs in health care more commonly elicit preferences for specific disease screening/management.[47-51] This could be because specialised service provision is better developed in general health services.

Also, it was interesting to note that one of our reviewed studies included 11 attributes in the design. While there are no design restrictions on the number of attributes that can be included in a DCE, often in practice most DCEs in health care have contained fewer than 10 attributes so as to ensure that all attributes are taken into consideration by respondents when making a choice.[52] Increasing the number of attributes in the study design can increase the complexity of design as well as cognitive difficulty of completing a DCE, which can increase response variability.[25] On the other hand, inclusion of fewer attributes can cause omitted variable bias owing to exclusion of key attributes. Rigorous piloting is thus necessary to get the balance of attributes right.[25]

The DCE models that can be estimated from the choice data often depend on the nature of the choice problem as well as the experimental design used.[29] Published literature indicates that while earlier DCEs in health care used the simple logit and probit models, over the last decade they have progressed towards more flexible and advanced econometric models.[30, 53] This is evident from the increasing applications of nested logit, random parameters logit and latent class models which allow for investigation of heterogeneity of preferences.[54-56] The pharmacy DCE studies were, however, restricted to the use of traditional logit or probit or MNL models with only one study utilising the latent class model to investigate pharmacist preferences for specialised services.[42] Probit or logit models or random effects extensions of these models often report the mean preference weights for the sampled population. However, it is likely that individuals or groups of individuals may have different preferences. Accounting for this heterogeneity is thus important and ignoring it may compromise the behavioural realism of the model.[54] The majority of our reviewed studies did not investigate the existence of preference heterogeneity in the study population and generally reported on the mean preference weights. This highlights the need for pharmacy practice researchers to take a structured approach and gain greater understanding of DCE methodology with respect to both the experimental design as well as the estimation models.

Monetary attributes were considered to be important by most patients and pharmacists in the studies reviewed. With respect to pharmacy services, patients showed a preference for lower costs or co-payments while pharmacists preferred higher incomes. On one hand, this information can be used to determine how much patients value pharmacists and pharmacy-based services and the extent to which they are willing to make investments in their health, while on the other hand it can provide insights into pharmacists' job choices and the financial gain they expect in order to deliver the services. This can be useful information at the policy level and in the development of economically viable services.

The majority of reviewed studies elicited patient preferences or pharmacist preferences, with just two studies examining preferences of both. Previous studies have shown that preferences of patients and providers for aspects of drug therapy[57] and screening programmes do differ,[21] thus highlighting the importance of understanding the perspectives of both, patients and providers, for particular products or services. This may be an important area of future research that will help us understand how well providers' views actually reflect patients' preferences, especially for novel specialised services. Also, understanding both perspectives may help identify similarities as well as mismatches, which in turn may help in the design of future optimal services that pharmacists are willing to deliver and patients are willing to use.

Another important observation in the measurement of patient preferences for pharmacy services was the existence of a status-quo bias where respondents tended to favour their current pharmacy or pharmacy service. Previous studies have shown that patients often value services more highly once they have experienced them.[58, 59] Further, the lack of awareness or information about novel alternatives may prompt patients to choose the status quo.[58] As pharmacy delivered specialised services are a relatively new paradigm, this lack of awareness and experience may haveled to patients preferring their current alternative/service. Future services need to overcome this status-quo bias in order to ensure their continual uptake by patients and long-term sustainability.

External validity testing of DCE responses is important, especially as these responses are made in regards to hypothetical choices. However, there have been relatively fewer tests of external validity in health DCEs.[30] One possible explanation may be that these DCEs have been conducted in countries with publicly funded health care where patients have limited choice and usually do not pay at the point of consumption for many of the health services, thereby making external validity tests difficult to conduct.[30] Consistent with health DCEs, none of the reviewed pharmacy-related DCE studies conducted tests of external validity. It is, however, important to note that the community pharmacy setting can offer a unique opportunity to conduct such external validity tests for hypothetical WTP estimates especially because pharmacy patients often pay at the point of consumption for many pharmacy services and interventions.[24, 60] Pharmacy practice researchers need to take advantage of this opportunity and conduct more research in this area of external validity testing.


To summarise, our review shows how DCEs have been designed, conducted and applied within the field of pharmacy. Clearly, more research is needed, beyond the current applications of patient/pharmacist preferences for products and services. The study emphasises the importance of adopting DCEs in pharmacy practice research and the need to move beyond the commonly used satisfaction instruments. Further, inclusion of health-outcome related attributes as well as preference measurement for specific disease-management services needs to be conducted. Testing for external validity and the incorporation of DCE in an economic evaluation framework to inform pharmacy policy remain important areas for future research.


Conflict of interest

The Authors have no conflicts of interest to disclose. The Authors alone are responsible for the content and writing of the paper.


This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Authors' contributions

Pradnya Naik-Panvelkar designed the search strategy, searched the databases, selected studies based on inclusion/exclusion criteria, conducted data abstraction and data synthesis and drafted the manuscript. Bandana Saini assisted in selecting studies based on inclusion/exclusion criteria, data abstraction, data synthesis and critically revised the manuscript. Carol Armour assisted in data synthesis and critically revised the manuscript. All Authors state that they had complete access to the study data that support the publication.