Objectives: To determine independent relationships between age and the risk of in-hospital death.
Design: Retrospective cohort study.
Setting: Thirty-eight intensive care units (ICUs) in 28 hospitals in a large Midwest metropolitan region.
Participants: One hundred fifty-six thousand, one hundred thirty-six consecutive admissions to medical, surgical, neurological, and mixed medical/surgical ICUs between March 1, 1991, and March 31, 1997.
Measurements: In-hospital death rates were compared at successive 5-year age intervals, adjusting for gender, diagnosis, admission source, comorbidity, and acute physiology scores. Acute physiology scores were determined using a validated methodology based on abnormalities in 17 physiological measures collected during the first 24 hours of ICU admission.
Results: The adjusted odds of death increased with each 5-year age increment. For example, relative to patients younger than 35, adjusted odds of death in patients aged 40 to 44, 50 to 54, 60 to 64, 70 to 74, 80 to 84, and 90 and older were 1.51, 1.73, 2.38, 2.98, 3.86, and 4.74, respectively. In stratified analyses, age-related increases in the odds of death were somewhat higher in surgical than medical patients or patients with lower severity of illness at admission. Although acute physiology scores had excellent discrimination in all age groups, discrimination decreased with age (e.g., c-statistics of 0.928 and 0.835 in patients younger than 45 and 85 and older, respectively).
Conclusion: Our findings demonstrate incremental increases in the risk of hospital death associated with age that was independent of severity of illness and other prognostic factors. Although the current results may be less biased by differences in treatment goals than studies of general hospitalized patients, the lower discrimination of physiology scores in older patients suggests that unmeasured factors (e.g., functional status, patient preferences for care, differences in physician practices) may be of greater prognostic importance in older than in younger patients.