The aim of the present study was to analyse the role of potential selection processes and their impact when evaluating risk factors for 30-day mortality among patients hospitalised for chronic obstructive pulmonary disease (COPD). A cohort of 26,039 patients aged > or = 35 yrs and hospitalised with COPD were enrolled. A 30-day follow-up was carried out using both the cause mortality register (CMR) and the hospital discharge register (HDR). Individual and hospital factors associated with 30-day mortality were studied using both mortality outcomes. The 30-day mortality rate was 1.21.1,000 patient-days(-1) (95% confidence interval (CI) 1.14-1.29) using the CMR, and 1.06.1,000 patient-days(-1) (95% CI 0.98-1.13) using the HDR. Male patients, the most poorly educated, those who resided outside Rome and those who had more than one hospitalisation in the previous 2 yrs were more likely to die after discharge than when hospitalised. The most frequent cause of in-hospital death was respiratory disease and after discharge, heart disease. Older age, male sex, comorbidities, previous hospitalisations for respiratory failure, and admission to a ward not appropriate to treat respiratory diseases were the most important predictors of 30-day mortality. Using in-hospital 30-day mortality provides a significantly different estimate of the role of specific risk factors.