AbstractObjective:To determine the fraction of hospital deaths potentially associated with nosocomial infection (NI).Design:A matched (1:1) case-control study.Setting:An 800-bed, tertiary-care, teaching hospital.Patients:All patients older than 14 years who were admitted to the hospital between January 1, 1990, and January 1, 1991, were eligible. All 524 consecutive deaths that occurred in the hospital comprised the case group. For each case, a control patient was matched for primary admission diagnosis and admission date.Outcome Measures:The proportion of hospital deaths potentially associated with NI was estimated from the population attributable risk (PAR) adjusted for age, gender, service, severity of illness, length of stay, and quality of the medical record.Results:For stays longer than 48 hours, the PAR for all NIs was estimated to be 21.3% (95% confidence interval [CI₉₅], 16.8%-30.5%). The greatest proportion of deaths potentially associated with NIs was observed in patients with only one infection (PAR, 15.0%; CI₉₆, 10.9%-22.6%) and bacteremia or sepsis (PAR, 7.7%;CI₉₅,4.6%-11.6%).Conclusions:NIs are associated with a large proportion of intrahospital deaths. This information may help clinicians and healthcare managers to assess the impact of programs for the prevention and control of NIs on intrahospital death.