Conjoint Analysis

Conjoint analysis is a survey based statistical technique used to understand how individuals value different attributes of a product, service, or intervention by presenting them with hypothetical scenarios that require trade offs.^[4][5] Participants evaluate profiles composed of varying attribute levels, and their choices reveal the relative importance of each attribute and the preferred combination of features. Developed from work in mathematical psychology and adopted by marketing research, conjoint analysis has become widely used in health economics, health technology assessment, and environmental valuation.

Many research questions involve understanding preferences in situations where multiple factors compete for attention. In health care, patients and clinicians make decisions that involve balancing efficacy, side effects, convenience, and cost. Conjoint analysis quantifies these trade offs in a way that direct questioning cannot, because asking someone "What is most important to you?" often produces socially desirable answers that do not reflect actual decision behavior. By revealing the implicit weights people place on different attributes, conjoint analysis supports patient centered drug development, benefit risk assessment, health policy design, and the development of new services.

Researchers begin by identifying the key attributes and their levels. For a study on diabetes treatment preferences, attributes might include blood sugar reduction (small, moderate, large), injection frequency (daily, weekly, monthly), and out of pocket cost ($50, $150, $300 per month). These attributes and levels are combined using experimental designs (fractional factorial designs are common) to create a manageable number of choice tasks. In a discrete choice experiment, one of the most popular forms of conjoint analysis, participants repeatedly choose between two or more hypothetical treatment profiles. Conditional logit, mixed logit, or latent class models then estimate the preference weight (utility) of each attribute level: an average for the sample with conditional logit, a distribution across respondents with mixed logit, and separate weights for groups of respondents with latent class models.^[5] From these weights researchers calculate willingness to pay, acceptable trade off ratios, and preference heterogeneity across subgroups.^[5]

Use conjoint analysis when you need to understand how people weigh multiple attributes simultaneously. It is appropriate for informing benefit risk assessments of medical products, designing health benefit packages, evaluating patient preferences for treatment characteristics, and estimating the monetary value people place on nonmarket goods (such as environmental benefits or reduced waiting times). Conjoint analysis is also useful in regulatory science, where agencies need to understand public tolerance for specific risk and benefit combinations.

A pharmaceutical company developing a new migraine medication conducts a discrete choice experiment with 500 migraine patients. The attributes include time to pain relief (30 minutes, 60 minutes, 120 minutes), risk of nausea (2%, 10%, 20%), and monthly cost ($20, $60, $100). Participants complete 12 choice tasks, each comparing two hypothetical treatments. Mixed logit analysis reveals that time to pain relief has the highest relative importance (45%), followed by risk of nausea (35%) and cost (20%). The willingness to pay for faster pain relief, in 30 rather than 60 minutes, is estimated at $42 per month, informing the company's pricing strategy and regulatory benefit risk discussions.