OBJECTIVES: Health related quality of life (HR-QoL) is studied in clinical trials to measure efficacy at the level of individual patients; at the population level HR-QoL reflects health system performance. The purpose of the study was to examine temporal variation in HR-QoL in light of potential bias in study outcomes, and to inform health policy guided by return on recent investment in the UK National Health Service (NHS). METHODS: A total of 23,356 hospitalisations in HODaR, a prospective patient database at Cardiff and Vale NHS Trust, were analysed for self-reported EQ-5D scores recorded routinely post-discharge during the 2002-2006 period. RESULTS: The mean annual EQ-5D for hospitalized population remained stable during the observation period for the general population and in subgroup analysis. In contrast, when records from all years were grouped in categories corresponding to months of survey, differences were significant (p<0.001); the lowest score was observed in May 0.622 [95%CI: 0.609-0.635] and the highest in July 0.674 [0.661-0.687]. When adjusted for age, sex, BMI, exercise, smoking and Charlson comorbidity index, the highest positive monthly effect on HR-QoL was reported in July (0.032, p<0.001) and negative in February (-0.028, p=0.003) and April (-0.019, p=0.0015). For robustness, month effect on all EQ-5D domains was analysed. As expected, in July item scores decreased (increasing HR-QoL) significantly for all the domains. In December and April item scores were higher, but the effect was not significant. EQ-5D scores were highly heterogeneous within individual Health Related Groups and did not reproduce the temporal pattern of the general population. CONCLUSION: Asynchronous measurements of quality of life for different treatment arms can introduce bias to clinical trials. Clinically meaningful changes in HR-QoL may result from monthly effects. In the short-term post-hospitalisation HR-QoL is likely to be insensitive to changes in investment in the NHS.