In the field of Health Technology Assessment (HTA), understanding the preferences of patients, healthcare professionals, and other stakeholders is crucial for making informed decisions. One valuable tool for capturing these preferences is the Discrete Choice Experiment (DCE). This article aims to explore what a DCE is, its value in HTA, the main steps involved in implementing a DCE, and best practices when building a DCE. By following these steps, researchers can effectively elicit and analyze preferences, leading to better-informed decision-making processes.

What is a DCE?

A Discrete Choice Experiment (DCE) is a quantitative research method used to assess and measure preferences. It presents participants with a set of hypothetical choices between different health interventions or treatment options, each with different attributes, and asks them to state their preferred option. By analyzing these choices, researchers can determine which attributes are most important to patients and stakeholders and how they weigh the trade-offs between different attributes.

What is the value of DCE?

Patient Preference Studies (PPS) like DCE can be implemented in each step of the medical development process life cycle in various ways. For example, DCE provides valuable insights into patient needs during the discovery phase and can also help with trial design during the clinical development phase. Preferences can also be used to evaluate the value of healthcare interventions and during the HTA phase, stakeholders can use these insights to allocate resources more efficiently and effectively to develop treatments with preferred attributes.

Main steps to implementing a DCE:

Step 1. Define the research question: Clearly articulate the research objectives and the specific preferences to be measured. Identify the target population and relevant attributes that influence decision-making.

Step 2. Design the choice sets-choose the attributes and identify levels: Develop choice sets that represent the alternatives. Define the attributes and their levels based on a thorough literature review, expert input, and stakeholder engagement. Ensure that the combinations of attribute levels are realistic and representative of the decision context. Consider the appropriate number of choices to balance the respondent burden and statistical efficiency.

Step 3. Pilot testing: Before conducting the main study, it is essential to pilot test the DCE design. This helps identify any issues with the questionnaire, refine the attribute descriptions, and ensure that the choice sets are understandable and realistic to respondents.

Step 4. Sampling and data collection: Determine the appropriate sample size and sampling strategy based on the research question and target population. Consider the mode of data collection, such as online surveys, face-to-face interviews, or telephone interviews, based on the target population and available resources. Use the appropriate data collection method to minimize biases and maximize response rates.

Step 5. Data analysis and interpretation: Employ appropriate statistical techniques to analyze the data and estimate preference models. There are three main types of models to choose from. The first is a model to estimate preference weights conditional importance of attributes. The second model identifies groups with similar treatment preferences. The last one is an estimation of willingness to pay. After running the various models, interpret the results in the context of the research question. Provide clear and concise summaries of the findings, including the relative importance of attributes.

Best practices when building a preference study using DCE

• It is important to involve stakeholders, such as patients, caregivers, and healthcare professionals, in the design process to ensure that the research question and the attributes of interest are relevant and meaningful. This can be achieved through focus groups, interviews, or surveys.
• The experimental design should be simple, the number of attributes and levels should be kept to a minimum to avoid overwhelming the participants. Generally, the number of attributes to evaluate is between 5 and 8. The main categories of treatment attributes are: Benefits, Risk, and Treatment Modalities.

• The questions should be written in a way that reduces biases.  For example, there should be neutrality in phrasing the questions and each attribute should show up an equal number of times in the DCE.
• Clear instructions and guidance should be provided to the participants to ensure that they understand the purpose of the study and how to complete the DCE. It is important to explain the concept of trade-offs and the hypothetical nature of the choices.

Discrete Choice Experiments (DCEs) are valuable tools in the field of Health Technology Assessment (HTA) for eliciting and measuring preferences. By following the main steps outlined above and adhering to best practices, researchers can effectively design and implement preference studies using DCEs. The insights gained from DCEs contribute to evidence-based decision making, helping policymakers allocate resources and make informed choices that align with the preferences of patients, healthcare professionals, and other stakeholders.

EvidentIQ can provide support when conducting DCE studies. From the experimental design stage to implementation and reporting, EvidentIQ can customize a solution that can help you effectively execute your patient preference study. They offer best-in-class Real World Evidence (RWE) methodologies, such as DCE, for patient studies in multiple diseases and geographical areas thanks to their direct access to a global patient platform. Patient studies can focus on treatment preference, quality of life, value of health, disease/treatment burden, unmet needs, etc. EvidentIQ can help generate unique Real World Data (RWD) to significantly help life sciences customers support the value story of their product for HTA submission, pricing and reimbursement as well as their scientific communication within the clinical community.

Sources: 

https://www.youtube.com/watch?v=IPIkIXWOJ5g
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8546533/#:~:text=A%20discrete%20choice%20experiment%20

https://yhec.co.uk/glossary/discrete-choice-experiment-dce/

https://bmjopen.bmj.com/content/11/3/e045803

https://systematicreviewsjournal.biomedcentral.com/articles/10.1186/s13643-021-01647-z

I’m Gilda Teissier, Executive Marketing Director at EvidentIQ and I recently met with Lise Radoszycki, COO at Carenity and Global Head of Data Science at EvidentIQ.

Over the past decade, we’ve observed a rise in patients’ power and engagement regarding their own care. 

The concept of Evidence-based medicine has also emerged with the goal of delivering the right care at the right time to the right patient. One of the core concepts of Evidence-based medicine is patient preference.

Industry, regulators, HTA bodies and players are showing more and more interest in the use of patient preference studies. The goal is to help incorporate the patient perspective into clinical drug development, care management, and healthcare decision-making.

In this context, most of the pharmaceutical companies want to be patient-centric, but it’s not always clear how and when to include patient perspective into the product development cycle.

The following interview with Lise Radoszycki, COO at Carenity and Global Head of Data Science at EvidentIQ, will give us an overview of preference study methodologies focusing on Discrete Choice Experiments (DCE), and will share light on how to optimize DCE design to improve the quality on patient studies.

Gilda: First of all, could you explain to us what a preference study is?

Lise: Of course, patient preference studies are based on patient preference information, that reflects what treatment attributes matter to patients, how much these matter to them and how they make trade-offs between treatments attributes.

Describing trade-offs allows us to evaluate compromises done by patients when they select their preferred treatment option. To illustrate this concept, I will ask you 3 questions:

1. Do you prefer to be rich or to be poor? I guess that your answer will be rich. In the same way, if I ask you, do you prefer to be sick or to be healthy? Again, the answer is easy, you will mostly say “healthy.”

2. Now, if I ask you “Do you prefer to be rich and sick or poor and healthy?” The choice is more complicated, and you will need to make some compromises or trade-off.

For each of these 3 types of information, qualitative or quantitative methodologies exist. Qualitative methods are mainly based on focus groups or semi-structured interviews, whereas quantitative methods are mainly based on questionnaires administered directly to patients.

Among the quantitative methods there are two main types of methodologies: the revealed preference methods and the stated preference methods.

Unlike revealed preference techniques that use observations on real choices made by respondents, stated preference methods ask patients how much they value something. This type of methodology allows to evaluate a treatment that is not yet on the market or a drug that targets a restricted population of patients such as a rare disease or a specific disease subtype. Stated preference methods are widely used in healthcare. In this type of method, the subject is asked how much they value something and here is where the Discrete Choice Experiment comes in.

Gilda: What is a Discrete Choice Experiment?

Lise: A DCE is a stated preference method increasingly used in healthcare to elicit patient preferences. In a DCE, participants are asked to state their preferred choice between several competing treatment alternatives. In general, there are two treatment options. The features of these treatments are called attributes. Each attribute is decomposed into levels. Each treatment alternative presented to participants is a combination of treatment attributes and levels. The question asked to patients might be: “If you were offered these two treatments, which one would you choose?”

“In a DCE, participants are asked to state their preferred choice between several competing treatment alternatives.”

Gilda: And which would be the main questions that need to be asked in order to develop a successful DCE?

Lise: For me there are three questions that help create the best study design. You can use these best practices for all patient studies, not just preference studies.

The first question is: Which research question is best suited for the strategic objectives to be achieved by measuring patient preferences?

The second: Which study population would be most appropriate to answer the defined research question?

Finally: What is the most appropriate methodology to answer the research question while considering the operational constraints?

Gilda: If we take this into consideration, it would mean that the DCE is not the right method for every preference study?

Lise: Even if the DCE is considered as a gold standard, it is not always the most appropriate method. Indeed, it is important to highlight that the main guidelines on preference studies published by the NICE, the FDA or ISPOR do not impose a particular methodology, but propose criteria for assessing the quality of patient studies.

For example, if the goal is to evaluate preferences regarding several topics (Qol, care pathway) the DCE is not the most appropriate methodology. Same if you want to evaluate the perception of a specific treatment option testing some pre-defined hypothesis. It is also very important to consider the operational constraints such as the prevalence of the study population, the timelines, the budget allocated to the study, etc.

Gilda: There are 5 key steps to a preference study with DCE methodology, could you tell us more about them?

Lise: Briefly, we could say that the first step is to identify the attributes of the treatment. The second, is to identify and define the levels of each attribute with methods such as literature research, and qualitative interviews. The third step is to create scenarios and choice sets with an experimental design method, which will be used to select a reduced sample of choices (named “fractional factorial” design). The fourth step is Statistical Modeling based on the random utility model. And the final key to a DCE study is the interpretation of results, where it is crucial to identify those that have a negative or positive impact on the preference of a treatment. Patient preference studies generate insightful data that can be used at each step of the medical product life cycle. 

Gilda: Do patient preference studies really make a difference in regulatory decision-making?

Lise: This is a very interesting question. Several Authorities (such as NICE or FDA) invite companies to start a conversation with them about using patient preference information to support their submission. Over the past years, several guidelines and initiatives have been developed to implement and improve quality standards (such as PREFER or IMI).

There is therefore, a real desire to include patient studies in regulatory decision-making. Nevertheless, the level of consideration given to this type of study is very variable and highly depends on the quality of the study.

For example, Janssen included in its dossier a preference study to support Esketamine, a treatment for resistant depression. The goal of this DCE study was to assess patient perspective on benefit-risk trade-offs. In their HTA dossier, outcomes importance was reinforced by the results of the patient study. The advisory committee of the FDA said that it was taken into account in their decision, but we do not know how much influence it really had at the end.

Gilda: Before leaving, could you give us some final tips to a successful DCE study?

Lise: First, it is really important to make sure your participants fully understand the study materials to reduce uncertainty caused by health numeracy/literacy.

For this purpose, you can use a pilot study or a cognitive debriefing to pre-test the questionnaire and the material associated with the study with a limited number of patients.  

The second advice would be to ensure representativeness of the sample and generalizability of results.

Another important thing to remember is that in order to improve the quality of the results, it is crucial to minimize the potential cognitive biases, such as framing bias, anchoring bias, ordering or labelling effects or simplified heuristics.

The final advice that I can give is to think carefully about the statistical analysis plan and data management plan in order to ensure logical soundness and robustness of data analysis.

Gilda: Thank you for your time.

A webinar addressing this same topic was recently conducted by Carenity and Takeda through the Xtalks platform. If you want to watch the replay you can access it by clicking here: ACCESS WEBINAR

For more information about EvidentIQ’s and Carenity’s solutions click here.