Data from Health Information Systems (HIS) have been used in recent years to evaluate effectiveness and quality of care. We analyzed in-hospital mortality occurring within 30 days following operation among 1984 patients (age > or = 35 yrs, males 84%) who underwent Coronary Artery Bypass Graft (CABG) surgery in seven hospitals in Rome in 1996. Data were extracted from the Lazio HIS run by the Regional Health Authority. The HIS database includes up to four diagnoses and surgical procedures (ICD-9); the following variables were considered a priori risk factors: type of ischemic disease, comorbidities, and others surgical interventions during the same admission. Logistic regression was performed in order to evaluate the association between hospital and risk of mortality after adjusting for potential confounders (ORs and 95% CI). The overall in-hospital mortality was 4.7% (ranged from 0.0 to 14.7%). Predictors of outcome included: older age, acute myocardial infarction, chronic myocardial ischemia, other heart diseases, chronic renal diseases, peripheral vascular diseases, other heart and vascular interventions. Statistically significant variability in mortality was observed across hospitals; taking hospital A as reference, hospitals D and E showed the highest risks (OR = 6.36 and OR = 3.12, respectively). We conclude that the observed differences in mortality rates among hospitals cannot be explained by differences in case mix, nor by varying coding styles. They are likely to reflect differences in currently unknown aspects of patient care.