The purpose of this study was to analyze causes of early readmission to the surgical intensive care unit (SICU), to determine whether readmission can be predicted or prevented, and to compare outcomes of patients readmitted to the SICU with patients not requiring readmission. All patients admitted to the Cedars-Sinai SICU from January 1, 1996, to December 31, 2001, were included. Clinical data was prospectively collected in an on-line computer system. The charts of all early readmission patients were retrospectively reviewed. SICU and hospital outcomes were abstracted from a computerized data warehouse. During the study period, 10,840 patients were admitted to the SICU including 97 (0.89%) early readmissions. SICU admission APACHE II and SAPS I scores, SICU and hospital length of stay, and mortality were significantly higher in readmitted patients compared to patients not requiring readmission. The majority of early SICU readmissions were due to respiratory and neurologic deterioration. Upon review, 62 per cent of all readmissions met appropriate SICU discharge criteria and were not predictable while only 5 per cent of SICU discharge were felt to have been premature. Patient outcomes are adversely affected by early readmission to the SICU. Careful neurologic assessment, meticulous attention to respiratory care transfer orders, and prompt respiratory therapy on floor care may significantly decrease the need for early readmission to the SICU.