OBJECTIVES: The EuroQol-5D3L (=3 levels) is the most widely used health state instrument to calculate QALYs. A major weakness, however, is lack of discrimination as a result of only three levels within each domain. Applying QALYs in pharmacoeconomics may be hampered when expected health gains are significant, but small. We established a formal approach to measure the gain of a 5 instead of 3 level EuroQol system, leaving the domains unaltered. Construct validity, information gain (using a quantification system), and reliability were established. METHODS: Forty-six patients with heterogeneous diseases (n=46) filled out the EuroQol questionnaire and a visual analogue scale (VAS) twice, with a 1-day interval. Patients were randomly assigned to the 3 or 5 level version each occasion, thus creating four combinations of EuroQol pairs between subjects (3-3, 3-5, 5-3 & 5-5). The 5 level wording was discussed within the EuroQol developers group. Using Shannon-Wiener's measure of information, the information content was calculated for each domain separately. Shannon-Wiener's criterion is maximal if all levels are equally used. For determining construct validity, the scores of the 3-5 and 5-3 combinations and VAS scores were used, predicting the use of the 5 levels on the base of VAS scores plus the 3 level score. Test-retest reliability was estimated. RESULTS: Information gain was substantial in each of the 5 domains, varying from 43% to 104% with an average increase of 57 SW-units. Because of skewed distributions and the small sample size, construct validity could be assessed for only part of the EuroQol scale, with confirmation of validity in 67% of the cases. Reliability of the 3-3 and 5-5 combinations was comparable. CONCLUSIONS: The extension of the EQ-5D descriptive to 5 levels is advocated as it substantially improves discriminative power of the EuroQol system in QALY calculations in small outcome changes.