ObjectiveTo describe characteristics of patients transferred from outside hospitals to a tertiary medical intensive care unit and to identify patient-level and system-level prognostic factors.DesignRetrospective cohort study.SettingTertiary university hospital.PatientsWe studied 3,347 patients who were transferred to the medical intensive care unit from outside hospitals from January 1990 through September 1999.InterventionsNone.Measurements and Main ResultsData collected included patient demographics, insurance type, discharge diagnoses, length of stay, mortality, admitting service, and distance traveled. The Charlson Comorbidity Score was used to adjust for comorbidity and the diagnostic related group risk level for risk of adverse outcome. Multivariate logistic models of early mortality (<72 hrs) and overall hospital mortality rate were developed. The most common major diagnostic categories included neurologic (10%), respiratory (10%), digestive diseases (10%), and drug overdose (10%). Most patients (70%) were transferred from >60 miles away. Mean medical intensive care unit length of stay was 5.3 days vs. 3.9 days for nontransfer patients. Transfer patients accounted for 49% of medical intensive care unit admits and 56% of intensive care unit patient-days. The overall mortality rate for transfer patients to the medical intensive care unit was 25% (95% confidence interval, 23–26), significantly higher than the 21% (95% confidence interval, 19–22) mortality rate among those admitted directly. Independent prognostic factors for early death (<72 hrs) included male gender, summer season, admitting service, diagnostic related group level, Charlson Comorbidity Score, insurance type, and major diagnostic category. Independent prognostic factors for overall hospital mortality rate included length of stay, medical complication, distance traveled, insurance type, and major diagnostic category.ConclusionsInterhospital transfers to the medical intensive care unit are patients at high risk for mortality and other adverse outcomes. System-level and patient-level characteristics influence both early and overall hospital mortality rates. These variables should be considered when risk stratifying medical intensive care unit patients and in studying outcomes of care.