Respiratory syncytial virus (RSV) accounts for the majority of lower respir
atory tract infections (LRTI) in infants and young children. A chest x-ray
is frequently performed in infants with LRTI caused by RSV. The aim of this
study was to develop and validate a prediction model to estimate the proba
bility for a normal chest x-ray in children with RSV infection. For this pu
rpose, easy obtainable diagnostic parameters were used. This prediction mod
el may be applied to decide which patients do not require a chest x-ray. Th
e data of 287 children admitted with RSV infection or diagnosed as such in
the outpatient department of the Sophia Children's Hospital between 1992-19
96 were studied. The derivation set comprised 232 patients (1992-1995), and
the validation set contained 55 patients (1995-1996). A chest x-ray was de
signated as normal when atelectasis, hyperinflation, or pulmonary infiltrat
es were absent. In order to develop a prediction model, patient history and
clinical and laboratory variables were consecutively entered into a logist
ic regression model according to the diagnostic workup that was practiced a
t the time. Variables with P less than or equal to 0.10 were retained in th
e model. The predictive accuracy of the multivariable models was examined u
sing the area under receiver operating curve (ROC-area).
In 202 (87%) patients from the derivation set, a chest x-ray was performed,
A normal chest xray could be predicted by increasing age, increasing birth
weight, presence of rhinitis, absence of retractions, and increasing arter
ial oxygen saturation. The ROC-area was 0.80 in the derivation and validati
on sets. This prediction model was transformed into a score chart.
In conclusion, a normal chest x-ray can accurately be predicted, using a mo
del including easily obtainable patient characteristics, and clinical and l
aboratory variables. This model may be useful tool in deciding whether or n
ot to perform a chest x-ray in patients with RSV infections. (C) 2001 Wiley
-Liss.