Purpose: To select diagnostic tests that confidently identify febrile infants less than 3 months of age seen at an outpatient facility as being at low risk for serious bacterial infection (SBI).
Data identification: An English-language literature search employing MEDLINE (1966 to 1991), Science Citation Index (1977 to 1991) using key citations, bibliographic reviews of primary research and review articles, and correspondence with authors of recent articles.
Study selection: After independent review by two observers, 10 of 333 originally identified titles were selected on the basis of prespecified selection criteria.
Data extraction: Two observers independently assessed studies by using explicit methodologic criteria for evaluating the quality of studies dealing with diagnostic tests. One reviewer extracted all the data from the articles; the second reviewer checked these data for accuracy.
Results of data analysis: On the basis of prespecified criteria, results were pooled from two studies that used the Rochester criteria, had high methodologic validity, and did not have significant heterogeneity (p = 0.32, Breslow-Day test), to give an estimate of the best negative likelihood ratio (95% confidence interval) for SBI = 0.03; 0 to 0.23).
Conclusion: The negative likelihood ratio of 0.03 allowed us to conclude that after the Rochester criteria for low risk of SBI have been satisfied, the probability of SBI in a febrile infant less than 3 months of age drops from a baseline rate of 7% (or 1 in 14 infants) to 0.2% (or 1 in 500). An expectant approach in these low-risk infants is therefore a reasonable choice.