INDEPENDENT PREDICTORS OF OCCULT PNEUMOCOCCAL BACTEREMIA (OPB) IN YOUNG FEBRILE CHILDREN. • 797

OPB occurs in 3% of children 3-36 months of age with fever≥39°C without source (FWS) or with otitis media (OM). 6% of these patients who are not treated with antibiotics will develop meningitis. Previous predictive models of OPB have mainly employed univariate statistics. In order to identify independent predictors of OPB, we developed a multiple logistic regression model. 6607 patients (pts) age 3-36 months with fever ≥39°C without source or with OM were prospectively evaluated. We compared the 164 pts with OPB to the 6415 pts whose blood cultures did not grow a pathogen (NEG). All pts were assigned a Yale Observational Score (YOS), and had a blood culture drawn. 5836 patients had a complete blood count, of whom 5478 had a manual differential performed. The following 6 variables were evaluated for their association with OPB: age, temperature (temp), YOS, white blood cell count(WBC), absolute neutrophil count (ANC), and absolute band count (ABC). Variables which were significantly associated with OPB (p<.05) were then entered into a stepwise logistic regression analysis. Variables which retained a significant association with OPB were considered independent predictors of OPB. Pts with OPB were similar in age to NEG pts (mean±S.D.;15.0±7.2 vs 14.2 ±8.2 months, P=.14). Pts with OPB, however, had a higher mean temp (40.0 ±.57 vs 39.8 ±.55°C, P<.001), YOS(7.0±1.8 vs 6.6±1.6, P<.01), WBC (21.4±7.9 vs 12.9±6.5 ×10³/mm³, P<.001), ANC (14.7±6.9 vs 7.3±4.9 ×10³/mm³, P<.001) and ABC(1.9±2.1 vs 1.0±1.2 ×10³/mm³, P<.001) than NEG pts. Logistic regression identified only ANC and temp as independent predictors of OPB: 7.8% of the 1301 pts with ANC≥10,000/mm³ had OPB. 9.1% of the 617 pts with both ANC≥10,000/mm³ and temperature≥40°C had OPB. In contrast, OPB occurred in 0.8% of 4176 pts withANC<10,000/mm³, and in 0.7% of 2713 pts with both ANC<10,000/mm³ and temp <40°C. Table

Table 1

We conclude that the ANC and temperature are independent predictors of OPB in children 3-36 months of age with FWS or fever with OM. Use of these two variables allows for an effective design of algorithms to limit the use of empiric antibiotics in young, febrile children at risk of OPB.