OBJECTIVES: In population health surveys, self-reported health status is usually assessed by a single question. In spite of good face validity, such a measure may be susceptible to differential item functioning (or measurement bias). This study aimed to examine the differences in health status measured by the self-assessed health status (SAHS) question and the EQ-5D visual analog scale (EQ-VAS) between Hispanics, blacks, and others (all other ethnicities/races as one group) in the US.  METHODS: We used the data collected in the US Valuation of the EQ-5D Health States study where a representative sample (N=4048) of the non-institutionalized adult population was surveyed in 2002. Hispanics and blacks were oversampled for the survey. In face-to-face interviews, all respondents answered the SAHS question (“overall, how would you rate your health?”) using a 5-point Likert-type scale (excellent/very good/good/fair/poor) and the EQ-5D which includes a vertical, 0-100 VAS for assessing global health. RESULTS: Excellent/very good health was reported by 64.4%, 65.1%, and 70.8% of Hispanics, blacks, and others, respectively. Mean (standard deviation) VAS scores for the Hispanics, blacks, and others were 84.7 (0.8), 83.5 (1.0), and 83.4 (0.6), respectively. In an ordinary least-square model, Hispanics (regression coefficient: 0.8) and blacks (regression coefficient: 0.2) were not different in VAS scores compared to others after adjusting for socio-demographics, smoking status, and presence of chronic conditions (both P>0.05). In contrast, Hispanics (odds ratio: 0.51) and blacks (odds ratio: 0.64) were less likely to report excellent/very good health than others in a logit model with adjustment of other factors (both P<0.001).          CONCLUSIONS: Hispanics, blacks and other ethnicities in the US respond similarly to the EQ-VAS but differently to the SAHS question. Caution should be exercised when interpreting self-reported health status of culturally diverse populations as assessed by single-item measures.