Predicting the likelihood of lower respiratory tractUreaplasmainfection in preterms

Abstract

ObjectiveTo develop predictive models ofUreaplasmaspp lower airway tract infection in preterm infants.MethodsA dataset was assembled from five cohorts of infants born <33 weeks gestational age (GA) enrolled over 17 years (1999–2016) with culture and/or PCR-confirmed tracheal aspirateUreaplasmastatus in the first week of life (n=415). Seventeen demographic, obstetric and neonatal factors were analysed including admission white blood cell (WBC) counts. Best subset regression was used to develop three risk scores for lower airwayUreaplasmainfection: (1) including admission laboratory values, (2) excluding admission laboratory values and (3) using only data known prenatally.ResultsGA and rupture of membranes >72 hours were significant predictors in all 3 models. When all variables including admission laboratory values were included in the regression, WBC count was also predictive in the resulting model. When laboratory values were excluded, delivery route was found to be an additional predictive factor. The area under the curve for the receiver operating characteristic indicated high predictive ability of each model to identify infants with lower airwayUreaplasmainfection (range 0.73–0.77).ConclusionWe developed predictive models based on clinical and limited laboratory information available in the perinatal period that can distinguish between low risk (<10%) and high risk (>40%) of lower airwayUreaplasmainfection. These may be useful in the design of phase III trials of therapeutic interventions to preventUreaplasma-mediated lung disease in preterm infants and in clinical management of at-risk infants.