Laboratory tests in evaluation of acute febrile illness in pediatric emergency room patients.

Abstract

The rising costs of health care and the movement for health care reform have focused attention on methods of cost containment. Of routine laboratory and radiologic procedures, complete blood cell count (CBC) and determination of serum electrolyte values rank as high as 2nd and 9th in overall cost. We retrospectively studied use of the clinical laboratory to aid diagnosis of an acute infectious event in a pediatric emergency department population. For 5 months, we reviewed medical records of pediatric patients younger than 15 years brought to the emergency department because of a febrile episode. Of 155 cases reviewed, electrolyte concentrations were determined in 108 patients and CBC in 155. In all patients, either culture or rapid test for streptococcal organisms was performed. In addition, 838 pediatric patients with similar symptoms but who did not undergo laboratory testing were monitored for 100 days. Measures of effectiveness including sensitivity, specificity, positive and negative predictive values, and likelihood ratio were used to correlate specific laboratory findings with antibiotic therapy, serious bacterial disease, and culture positivity. Electrolyte abnormalities were found largely to be dismissed clinically, with the major clinical response consisting of parental education about hydration. The CBC profile was evaluated, with white blood cell count (WBC) indicator limits of > 10,000, > 10,000 but < 15,000, and > 15,000/mm3, and differentiated into absolute neutrophil count, neutrophil percent, and band cell percent. Temperature was evaluated as an independent variable. Insofar as serious bacterial disease and culture positivity, sensitivity was uniformly low (70%), and specificity was only marginably acceptable for WBC > 15,000 (77%). Both positive predictive values and likelihood ratio were low with respect to predicting either serious bacterial disease or culture positivity, emphasizing the limited usefulness of these clinical laboratory measurements. The best hematologic predictors of serious bacterial disease or culture positivity were obtained with automated hematologic analyzers and exceeded manual differential measurement of neutrophil percent and band cell percent. In addition, we correlated the administration of antibiotics with the various hematologic parameters and discovered that WBC > 15,000, regardless of cause, almost uniformly resulted in treatment (positive predictive value, 93.5%; likelihood ratio, 5.60). These findings support the use of automated hematology analyzer-derived measurements and question the use of manual differential counts, unless specific issues are to be addressed. Furthermore, the findings seem to support more reliance on clinical impression and less on laboratory values.