- The decision-tree model described above was computer derived, based
on data from 3066 febrile infants less than 3 months old who were prospectively
studied in pediatric clinics. Applied retrospectively to the patients
studied, it showed 93.6% sensitivity and 27.3% specificity for identifying
bacteremia and bacterial meningitis.1
- The negative predictive value of the model was 99.6%, while the positive
predictive value was only 2.9%.
- The clinicians who participated in this study followed individualized
approaches to evaluate the infants (a standard approach was applied
only 42% of the time), yet they had a superior sensitivity and specificity
when compared to the standard approach and when compared to the computer
model. (Clinician sensitivity 97.1%, specificity 35.5%; Standard approach
sensitivity 95.2%, specificity 35.2%).
- The prevalence of bacteremia and bacterial meningitis was similar
to that found in studies of inner-city ED populations (bactermia 1.8%
[2.4% of those initially tested] and bacterial meningitis 0.5%)