Adults with community-acquired bacteremias in two suburban hospitals: factors predicting outcome.

In a suburban hospital, 38 adult patients with community-acquired bacteremic infections and without "rapidly fatal" disease were studied. In univariate analyses, the 31 patients discharged alive had higher initial temperatures, blood pressures, and calcium and albumin values, and lower immature leukocyte counts, and were treated sooner with appropriate antibiotics than those who died. With the exception of time elapsed before appropriate treatment, these factors were used in discriminant analysis to identify patients who would die during hospital admission from bacteremic infection. The resulting model, which had a sensitivity of 71% and a specificity of 68% in the first group of patients, was validated in a second cohort with a sensitivity of 40% and a specificity of 69%. This analysis presents a model that may allow clinicians to assess the risk of death from community-acquired bacteremic infections.