Objective: To determine the accuracy of predischarge visual assessment of jaundice for estimating bilirubin concentration and predicting risk of significant neonatal hyperbilirubinaemia.
Design: Prospective cohort study.
Setting: Well Baby Nursery at the Hospital of the University of Pennsylvania.
Patients: 522 term and late preterm newborns.
Interventions: Nurses used a 5-point scale to grade the maximum cephalocaudal extent of jaundice prior to discharge.
Main outcome measures: (1) Correlation between jaundice grade and bilirubin concentration. (2) Predictive accuracy of jaundice grade for identifying infants who developed significant hyperbilirubinaemia, defined as a bilirubin level that at any time after birth exceeded or was within 1 mg/dl (17 micromol/l) of the American Academy of Pediatrics-recommended hour-specific phototherapy treatment threshold.
Results: Nurses' assessment of jaundice extent was only moderately correlated with bilirubin concentration and was similar in black and non-black infants (Spearman's rho = 0.45 and 0.55, respectively (p = 0.13)). The correlation was particularly weak among infants <38 weeks' gestational age (rho = 0.29) compared with infants > or = 38 weeks' gestation (rho = 0.53, p = 0.05). Jaundice extent had poor overall accuracy for predicting risk of significant hyperbilirubinaemia (c-statistic = 0.65) but complete absence of jaundice had high sensitivity (95%) and excellent negative predictive value (99%) for ruling out the development of significant hyperbilirubinaemia.
Conclusions: Clinicians should not use extent of cephalocaudal jaundice progression to estimate bilirubin levels during the birth hospitalisation, especially in late preterm infants. However, the complete absence of jaundice can be used to predict with very high accuracy which infants will not develop significant hyperbilirubinaemia.