OBJECTIVE: To compare the performance of a generic measure (EuroQoL EQ-5D) and condition-specific measures (GHQ-12, HADS and SCL-90) in measuring mental health outcomes. METHODS: Patients from 3 separate studies were identified - dental patients attending for treatment (n=600) ; psychiatric patients taking part in an RCT of psychotherapy (n=104) ; patients in an observational study in rheumatoid arthritis (RA) (n=233). Patients receiving dental treatment also complete the GHQ-12 (a measure of psychiatric ‘caseness’). Psychiatric patients in the interventional study were assessed using SCL-90 at baseline and again at 3- and 6-months. RA patients completed HADS at baseline, 3- and 12-months. All patients completed EQ-5D on one or more occasions. The performance of EQ-5D was tested in terms of its capacity to discriminate between groups with known differences in mental health status, and to detect significant clinical change over time. RESULTS: Dental patients who screened positive on the GHQ-12 differed significantly in terms of their EQ-5D from patients who did not. RA patients with well separated HAD scores were found to differ significantly in terms of EQ-5D. In the psychotherapy study, patients with poor levels of general health status on admission, reported significant improvement; above average patients reported significant deterioration. Correlation between EQ-5D and SCL-90 subscales was significant. CONCLUSIONS: EQ-5D detects significant differences in health status between patients with known mental health problems, and is sensitive to changes in patients with diagnosed mental illness. However, the anxiety/depression dimension may have different sensitivity to its two components.