Study instrument

We first devised an English-language version of the survey, drawing heavily from Wright et al.10 We similarly used 5-point Likert-scale responses, and 23 items on career-choice predictors from Wright et al.10 As we wanted students to consider specific characteristics of GP, we added 7 items on work settings; 7 items on preferred age/gender groups; 10 items on attitudes about GP, and 14 items on attitudes about career and lifestyle. Demographic information was linked from the student’s unique identifier. As not relevant for China, we did not include two items previously used by Wright et al (adequate income to eliminate debt, sooner deal with medical than social problems).10