October 1, 1990Annals of Internal Medicine
Predicting Bacteremia in Hospitalized Patients
Authors
Discussion
Loading...
Member takes
Implication
Prospective cohort study demonstrates clinical predictors accurately stratify bacteremia risk in hospitalized patients, indicating potential to reduce unnecessary blood cultures and antibiotics.
Key Points
- To develop and prospectively validate a clinical prediction model for bacteremia in hospitalized patients to identify low-risk subsets and reduce the misinterpretation of false-positive blood cultures.
- Prospective cohort study evaluating 1,516 blood culture episodes drawn from patients at a single urban tertiary care hospital.
- The sample was partitioned into a derivation cohort (n=1,007 episodes) and a validation cohort (n=509 episodes) to identify and test independent multivariate predictors.
- Independent multivariate predictors of true bacteremia were temperature ≥38.3°C, rapidly or ultimately fatal disease, shaking chills, intravenous drug abuse, acute abdomen, and major comorbidity.
- In the low-risk cohort lacking these predictors, the model misclassification rate was 1% (4 of 303) in derivation and 2% (3 of 155) in validation, with a positive predictive value of only 14%.
- The derivation cohort demonstrated an overall true-positive rate of 7% (74 of 1,007) and a false-positive rate of 8% (81 of 1,007), whereas the model-defined high-risk subset had a true bacteremia rate of 16% (41 of 264).
Cite This Study
David W. Bates (1990) studied this question.