OBJECTIVES: Although discrete choice experiments are increasingly being used in health care applications to elicit preferences, their use has been limited in diseases with low prevalence because it is challenging to design a statistically and clinically meaningful study with a small sample of respondents. This study used best-worst case scaling to enhance the design of a discrete choice experiment eliciting physician preferences related to the care of patients being treated for hemophilia with inhibitors. METHODS: Thirty hematologists provided data on factors having an impact on their treatment decisions by completing a survey instrument via face-to-face interviews at a scientific meeting. To increase the amount of useful information obtained from each respondent, best-worst scaling was used. Specifically, each choice task was structured so that respondents provided input on and ranked three scenarios from most to least preferred. This innovative method had not previously been used in discrete choice modeling. RESULTS: With the increase in data due to the applied best-worse scaling method, an aggregate multinomial logit model established stable parameter estimates while obtaining a ‘consensus' view of hematologists' preferences. In substantive terms, the time required to stop bleeding was the most important factor affecting treatment decisions [relative importance (RI) = 16.3%]. Physicians also preferred treatments that resulted in quick pain relief [RI=12.9%]. CONCLUSION: This example indicates that best-worst case scaling can effectively be used in discrete choice experiments involving aggregate multinomial logit modeling. This method can enhance and increase the use of discrete choice experiments to elicit preferences from relatively small numbers of physicians or patients with rare diseases, or when few respondents are available.