OBJECTIVE: The EQ5Dindex is widely used to evaluate health preferences and provide utility estimates. Objective of this study was to evaluate the face-validity of the EQ5Dindex. METHODS: Data used here were the first 40,000 responses in the Health Outcomes Data Repository (HODaR). In addition to survey data HODaR details clinical phenotype. Patients were surveyed with the EQ5Dindex (excluding the VAS), and the SF36. A simultaneous measure quantified health status on an arbitrary scale of 0 to 100 pre-admission and post-discharge. The EQ5Dindex has 243 potential values. RESULTS: Frequency distribution of the EQ5Dindex was tri-modal and difficult to describe in summary statistics. In all, 27 possible values (11%) were responsible for 92% of all observations, 14 possible values had no observations, and 24.7% of returns had an EQ5Dindex of 1.0. There are a number of categories that are rarely used e.g., severe mobility problems and severe self care problems. There was a close correlation between weighted scale and simple addition of responses (R2=0.87). There were 6.8% of responses with an EQ-5Dindex =0.0. There was a low correlation between the EQ5Dindex with the general health question of the SF36 and the arbitrary, continuous valuation of health status above. The ranking of mean estimates was intuitively correct. CONCLUSIONS: The number of theoretical values that are represented was sparse. The EQ-5Dindex distribution results in no easily describable parametric distribution, and the correlation with other general health measures was low. Given that these subjects are hospital treated, too many may have a health status of 1.0, and too many are also in a health status notionally equal to or worse than death. Decisions based on the EQ5Dindex now have enormous health and commercial implications. The EQ5D classifies the right health factors but the sensitivity and scoring methods need urgent revaluation: good but needs improving.