Abstract
Previous research shows that choices are influenced by how probabilities are presented, that we value losses more than gains, and that we misunderstand cumulative probabilities over time. These factors are important when designing discrete choice experiments (DCEs) because almost all include some representation of probability over a time period. Contraceptive choice is one of the most common health choices and requires people to trade-off between efficacy, side effects, and modality. We used a DCE to explore whether people chose differently when faced with positive or negative framings of contraceptive effectiveness or valued 1-year or 3-year cumulative risks differentially. We developed a simple eight-task DCE with three attributes: effectiveness, administration frequency, and (non-)hormonal nature. Participants saw effectiveness as either positively or negatively, and with numerically equivalent 1-year and cumulative 3-year effectiveness values. We used mixed multinomial logistic regression models with interaction terms and explored preference heterogeneity. The negative frame increased sensitivity to effectiveness by 18% (p = 0.04) and sensitivity to cumulative effectiveness over 3 years was 10% less than over 1 year (p = 0.01). Preferences were heterogenous with respect to attributes but not framing effects. Attribute framing substantially affected preferences for effectiveness, and decisions around risk presentation should be reported transparently.