Prediction model of RSV-hospitalization in late preterm infants: An update and validation study

Abstract

Background: New vaccines and RSV therapeutics have been developed in the past decade. With approval of these new pharmaceuticals on the horizon, new challenges lie ahead in selecting the appropriate target population. We aimed to improve a previously published prediction model for prediction of RSV-hospitalization within the first year of life.

Methods: Two consecutive prospective multicenter birth cohort studies were performed from June 2008 until February 2015. The first cohort (RISK-I, n=2524, 2008-2011) was used to update the existing model. The updated model was subsequently validated in the RISK-II cohort (n=1564, 2011-2015). We used the TRIPOD criteria for transparent reporting.

Results: 181 infants (n=127 in RISK-I, n=54 in RISK-II) were hospitalized for RSV within their first year of life. The updated model included the following predictors; day care attendance and/or siblings (OR: 5.3; 95% CI 2.8-10.1), birth between Aug. 14th and Dec. 1st (OR: 2.4; 1.8-3.2), neonatal respiratory support (OR 2.2; 1.6-3.0), breastfeeding ≤4 months (OR 1.6; 1.2-2.2) and maternal atopic constitution (OR 1.5; 1.1-2.1). The updated models' discrimination was superior to the original model in the RISK-II cohort (AUROC 0.72 95% CI 0.65-0.78 versus AUROC 0.66, 95% CI 0.60-0.73, respectively). The updated model was translated into a simple nomogram to be able to distinguish infants with high versus low risk of RSV-hospitalization.

Conclusion: We developed and validated a clinical prediction model to be able to predict RSV-hospitalization in preterm infants born within 32-35 weeks gestational age. A simple nomogram was developed to target RSV therapeutics to those children who will benefit the most.

Keywords: Late preterm; Prediction; Respiratory syncytial virus; Updating; Validation.