Observation Scales to Identify Serious Illness in Febrile Children

The pediatrician makes a judgment of the degree of illness (toxicity) of a febrile child based on observation prior to history and physical examination. In order to define valid and reliable observation data for that judgment, data from two previous studies were used to construct three-point scales of 14 observation items correlated with serious illness in those reports. Between Nov 1, 1980 and March 1, 1981, these 14 scaled items were scored simultaneously by attending physicians, residents, and nurses prior to history and physical examination on 312 febrile children aged ≤24 months seen consecutively in our Primary Care Center-Emergency Room and in one private practice. Of these 312 children, 37 had serious illness. Multiple regression analysis based on patients seen by at least one attending physician in the Primary Care Center revealed six items (quality of cry, reaction to parents, state variation, color, state of hydration, and response to social overtures) that were significant and independent predictors of serious illness (multiple R = 0.63). The observed agreement for scoring these six items between two attending physicians who saw one third of the patients ranged from 88% to 97%. The chance corrected agreement levels (κw) for these six items were, with one exception, clinically significant (κw = .47 to .73). A discriminant function analysis revealed that these six items when used together had a specificity of 88% and a sensitivity of 77% for serious illness. Individual scores for each of the six key items were added to yield a total score for each patient. Only 2.7% of patients with a score ≤10 had a serious illness; 92.3% with a score ≥l6 had a serious illness. The sensitivity of the six-item model for serious illness when combined with history and physical examination was 92%. In the population studied, this predictive model, when used prior to history and physical examination, was reliable, predictive, specific, and sensitive for serious illness in febnile children. It was most sensitive when combined with history and physical examination. The model wifi need to be validated on a new population of patients.