Development of a predictive index for picornavirus infections.

Abstract

To create a clinical prediction index that aids in the diagnosis of picornavirus respiratory infections, we analyzed patients from 5 clinical trials designed to evaluate the efficacy of an antiviral treatment for respiratory infections. Logistic regression was used to determine which baseline symptoms and patient characteristics best predicted picornavirus infection. Parameter estimates were then used to create a predictive index for estimating the probability of picornavirus infection on the basis of cold symptoms. The presence at baseline of rhinorrhea (odds ratio [OR], 2.73), nasal congestion (OR, 1.63), and sore throat (OR, 1.37) increased the likelihood of picornavirus infection; the presence of myalgia (OR, 0.71) and fever (OR, 0.59) decreased the likelihood. The positive and negative predictive values of the model were 61.5% and 64.4%, respectively. The model was simplified for clinical use by creating a whole-number index: the lowest possible score (-3) indicates a 15% chance of picornavirus infection and the highest (7) indicates a 69% chance of picornavirus infection.