Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004
We applaud the efforts of Drs Friedman and Attia to develop a clinical model to identify children with influenza infection.1 An accurate prediction model may facilitate the timely diagnosis of influenza infection and guide empirical antiviral therapy. Two issues warrant further comment. First, children with viral infections other than influenza were excluded from the predictive model. We believe these children merit inclusion since the clinician does not know beforehand whether these children have influenza infection. It is likely that children with influenza infections have clinical features that are more similar to children with other known viral infections compared with children without identifiable viral pathogens. Thus, the exclusion of these children provides a form of spectrum bias and leads to an overestimate of the specificity of the rule. Even though the absolute effect of including these children in the model may be modest (since it is a small number), excluding all children with infections known post hoc to be caused by other viruses potentially leads to a rule with little clinical utility.
Shah SS, Metlay JP. Clinical Prediction Model for Influenza. Arch Pediatr Adolesc Med. 2004;158(10):1017. doi:10.1001/archpedi.158.10.1018-a