CLINICAL-PREDICTION RULE FOR PULMONARY-INFILTRATES

Objectives: This study was designed to develop a clinical prediction r ule that identifies adult patients with acute respiratory illness who would demonstrate a pneumonic infiltrate on a chest roentgenogram. Mat erials and methods: The demographic, clinical, and chest radiologic da ta of 557 adult patients (derivation set) with acute respiratory illne ss were used to construct and to cross-validate a model using stepwise logistic regression procedure. A prospective sample of 100 patients ( validation set) was used to validate the model. Results: The prevalenc e of pulmonary infiltrates among the derivation and the validation set s was 17% and 18%, respectively. The stepwise model identified documen ted fever (>38 degrees C), rales, colored sputum, dullness to percussi on, absence of asthma, and bronchial breathing as significant factors that independently predict pneumonic infiltrates. The area under the r eceiver operating characteristics was 0.81 (95% CI, 0.77 to 0.85%). Th us, the probability that the logistic rule would correctly identify a patient with pneumonic infiltrates was 81% and the overall sensitivity and specificity of the rule were 78% and 94%, respectively with a tot al error rate of 8.4%. Both cross-validation and the prospective valid ation reaffirmed the robustness of the clinical prediction rule. Concl usion: We concluded that the clinical prediction rule proved useful in identifying adult patients who are having pulmonary infiltrates. The model would be beneficial to assist physicians in deciding whether to order a chest roentgenogram.