OBJECTIVES: A key indicator of population health and well-being is self-rated health. We aimed to compare self-rated health by Americans in 2002 and 2017.

METHODS: Data from two US EQ-5D valuation studies conducted in 2002 and 2017 were combined. In both studies, respondents completed the EQ-5D-3L self-classifier and visual analogue scale (VAS), where health is rated from 0 (worst imaginable health) to 100 (best imaginable health). To take into account differences in cohort characteristics, ordinary least squares regression models adjusted for sociodemographic characteristics, presence of disease, and self-reported problems in EQ-5D dimensions, defined as any/no problems.

RESULTS: The proportion of respondents in 2002 (n=3,728) vs. 2017 (n=1,047) reporting a VAS score of 100 (13.4% vs. 13.0%) or 90-99 (40.0% vs. 41.6%) were similar. No differences in mean VAS scores were observed between respondents in 2017 (84.6 [SD=14.5]) and 2002 (84.4 [SD=16.1]). Adjusting for sociodemographic characteristics and presence of disease had negligible effect. However, upon adjusting for problems on each EQ-5D dimension, mean 2017 VAS scores were significantly higher than 2002 (89.8 vs. 87.6; mean difference=2.2 [SE: 0.43; 95% CI: 1.36 to 3.10]).

CONCLUSIONS: Self-rated health of the general US adult population in 2017 was very similar to 2002, although by adjusting for health problems it was slightly higher in 2017. These results suggest that the “average” adult American rates their health as slightly better compared to 15 years ago. The datasets were highly comparable which was a strength of the analysis, while a limitation is that some unobserved factors may have been important.