RESUMO
PURPOSE: To describe long-term mortality and hospital readmissions of patients admitted to Brazilian intensive care units (ICU). METHODS: Retrospective cohort study of adult patients admitted to Brazilian hospitals affiliated to the Public Healthcare System from 10 state capitals. ICU patients were paired to non-ICU patients by frequency matching (ratio 1:2), according to postal code and admission semester. Hospitalization records were linked through deterministic linkage to national mortality data. Primary outcome was mortality up to 1 year. Other outcomes were mortality and readmissions at 30 and 90 days and 3 years. Multiple Cox regressions were used adjusting for age, sex, cancer diagnosis, type of hospital, and surgical status. RESULTS: We included 324,594 patients (108,302 ICU and 216,292 non-ICU). ICU patients had increased hospital length of stay [9 (5-17) vs. 3 (1-6) days, p < 0.001] and mortality (18.5 vs. 3.6%, p < 0.001) versus non-ICU patients. One year after discharge, ICU patients were more frequently readmitted to hospital (25.4 vs. 17.4%, p < 0.001) and to ICU (31.4 vs. 7.3%, p < 0.001) than controls. Mortality up to 1 year was also higher for ICU patients (14.3 vs. 3.9%, p < 0.001). A significant interaction between surgical status and mortality was found, with adjusted hazard ratios (HRs) up to 1 year of 2.7 [95% confidence interval (CI) 2.5-2.9] for surgical patients, and 3.4 (95%CI 3.3-3.5) for medical patients. The risk for death and readmission diminished over time up to 3 years. CONCLUSIONS: In a public healthcare system of a developing country, ICU patients have excessive long-term mortality and frequent readmissions. The ICU burden tended to reduce over time after hospital discharge.