Data available from the standard hospital discharge database (SDO) allow us to explore differences in health conditions according to different indicators of socioeconomic status (SES). We analysed all the patients aged 30-59, discharged from the S. Giovanni Battista (Molinette) hospital (the main general hospital in Turin, Italy) during three years (1996-1998) (n = 49949). Three health indicators were used as outcomes: a) emergency admission; b) severity of illness (according to the "All Patient Refined DRGs" subclasses); c) hospital mortality. Patients were compared for each outcome according to two different SES indicators: a) level of education; b) employment status. Logistic regression models (both conditional and unconditional) were used to adjust for several potential confounders. Patients with lower education (up to 5 years of schooling), compared to those with 13 or more years of schooling, showed a higher probability of being admitted through the emergency ward (29.1% vs 23.3%), with an odds ratio (OR) = 1.56-95% confidence interval (95% CI) = 1.45-1.68; of being classified in higher severity subclasses of illness (23.3% vs 17.7%, OR = 1.14; 95% CI = 1.07-1.22) and of dying in hospital (2.3% vs 1.6%). However, after adjustment for other prognostic factors (as severity of illness and specific expected mortality), this association disappeared (OR = 1.05, 95% CI = 0.84-1.32). Similar, but somewhat stronger, associations were observed when comparing the unemployed versus the employed. The corresponding figures (ORs; 95% CI) were 1.57 (1.42-1.74) for emergency admission; 1.31 (1.18-1.45) for severity of illness and 1.55 (1.10-2.16) for hospital mortality. In conclusion, this study showed that SES differentials in health are clearly measurable through routine hospital information systems, and documented that patients of low SES, particularly unemployed, experienced a delayed access to hospital, were admitted in poorer general health conditions and had a more unfavourable prognosis.