OBJECTIVES: To compare discrete choice experiment (DCE) and best-worst scaling (BWS) in eliciting lung cancer treatment preference measurement.

METHODS: A face-to-face survey was conducted among lung cancer patients recruited using a quota sampling in 2 tertiary hospitals in China. Participants completed sociodemographic information, 13 DCE and 13 BWS choice tasks in random order, and the questions about the difficulty of the tasks. Each profile in the choice tasks was comprised of six attributes, namely, overall survival (OS), risk of severe adverse effects, severity of pain, appetite, physical functioning, and monthly cost. A generalized multinomial logit model was used to estimate attribute level preference weights, which were further used to compute the relative importance (RI) for each attribute. The preference weights were rescaled to calculate the correlation coefficient between the two tasks. The completion time, difficulty of the tasks, and RIs were compared between DCE and BWS.

RESULTS: A total of 161 patients completed the survey, with 29.8%, 5.6%, 30.4% and 34.2% at stage Ⅰ, Ⅱ, Ⅲ and Ⅳ, respectively. The majority were male (69.6%) and the mean age was 58.2 years. Respondents took longer to complete DCE than BWS tasks (25 min vs. 21min, P<0.001). However, the DCE task was easier to understand (74.5% vs. 66.5%) and answer (64.0% vs. 57.8%), and thus preferred by more respondents (49.7% vs. 26.1%), compared with BWS. DCE and BWS generated similar attribute rankings with OS and physical functioning ranked the first two with the RIs of 38% and 24% in DCE, and 34% and 17% in BWS, respectively. Rescaled attribute level preference weights of DCE and BWS had strong correlation of 0.881 (P=0.000).

CONCLUSIONS: DCE and BWS generated similar attribute ranking and their preference weights correlated strongly. However, DCE was easier to complete and thus preferred over BWS by lung cancer patients in China.