OBJECTIVE: To investigate factors influencing health preference with expectancy-value model (EVM). METHODS: EVM, a model widely used to explore underlying factors of attitudes, was applied to study health preference, which was categorized as attitude in psychology. The factors include attitudinal attributes (AAs) and external variables. AAs are measured in a sum of multiplications of one's subjective probability (expectancy) and perceived value of attributes. In one-to-one interviews, four AAs were identified in focus group discussion, namely, reduction in quality of life (RQoL), burden to family (BTF), dependence on others (DOO) and inability to work (ITW) were assessed using 7-point Likert scales to measure expectancy and value of each attribute. Health preference was measured using visual analogue scales (VAS, range 0-100). Univariate analyses were used to identify external variables (age, gender, ethnicity, education, housing, marital status, and concurrent chronic diseases) that cause significant difference in VAS. Multiple linear regression model (MLR) was used to investigate the explanative power of AAs and possible significant external variable(s) separately or in combination. RESULTS: Twenty-five Chinese and 21 Indians (mean (SD) age: 45.0 (15.55) years, 55.6% female) were interviewed. Ethnicity was identified as the only independent variable causing significant difference in VAS (p<0.05). RQoL, BTF, DOO, ITW were found to explain the variation of VAS of 26.5%, 27.2%, 23.2%, 17.1% respectively in separate MLR models (p<0.05). Combining ethnicity together with the sum of AAs explained up to 27.3% of the variation in VAS, while a model with ethnicity alone only accounted for 10.7% (p<0.05). When MLR was done to examine different predicting power of AAs by ethnicity, ITW failed to predict VAS of Indians (0.3%,p=0.80) while the other 3 AAs moderately explained from 9.1% to 22.9% of the variation (p<0.05). CONCLUSIONS: EVM may be helpful in explaining the variations in health preference and predicting important factors.