Who needs a blood culture? A prospectively derived and validated prediction rule

From Clinfowiki
Revision as of 14:20, 28 October 2008 by DeanSittg (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Introduction and Objectives: Clinically accurate and cost effective utilization of any diagnostic or therapeutic modality is essence of good health care. With this aim, this study tried to see how one could optimize the use of blood culture in diagnosing bacteremia (positive blood culture). The authors prospectively collected data for an observational cohort of consecutive patients whose blood cultures were obtained in the Emergency Department (ED) at a tertiary care hospital in Boston. This data was analyzed to try and create retrospective rules which could improve ordering of blood cultures in a manner that it improved the positivity of outcome without compromising the sensitivity of the test and hence, derive easily applicable clinical decision rules to that effect.

Methods and Results: 3730 out of 3901 patients whose blood cultures were drawn in the ED over a year were studied. All characteristics of history, physical examination, lab data and other comorbities (covariates) and outcomes were collected. Two thirds of these patients were assigned to a derivation set and the rest were assigned to a validation set. “Bacteremia” was present in 8.3% of patients in the derivation set and 8.0% in the validation set, suggesting that the two groups were comparable. All the covariates were analyzed by univariate analysis and a multiple logistic regression model was created. This showed that 13 covariates were identified as independent predictors of bacteremia. Further, patients could be divided into three groups based on the number of these risk factors (each given 1 point). This stratification showed that the incidence of bacteremia was 0.6%, 6.8% and 26%, in patients with low risk (0-1 points), moderate risk (2-5 points) and high risk (> 6 points), respectively. This point system was then applied to the validation group and it showed that bacteremia in this group was 0.9%, 9.1% and 15% in the low, moderate and high risk subgroups, respectively, and this was statistically similar (p=0.27).

Discussion and Conclusion: Careful analysis of the clinical profile of patients could help optimize the use of blood cultures to diagnose bacteremia for patients presenting to the ED. The authors were able to define major and minor criteria (details in article) and suggested that blood cultures be sent only for patients with either one or more major criteria or with at least two minor criteria. This could increase the yield of the test by limiting it to patients with moderate or high risk. They estimated eliminating the use of blood cultures in the low risk population could result in 27% reduction in cost (Annually $16,000 in cost or $124,000 in charges).

Implications in Health Informatics: Working with this approach to the use of Clinical Decision Support Systems could turn out to be a big asset to Health Informatics in Medicine. During computerized physician order entry, the system could calculate the risk score of the patient with the data available and could prompt the ordering physician regarding the risk score of the patient to enable optimal use of the test.

Ira Bhargava