Is visual assessment of jaundice reliable as a screening tool to detect significant neonatal hyperbilirubinemia?

Objective To assess the reliability of visual assessment of bilirubin levels (BiliEye) in newborns as a screening tool to detect significant neonatal hyperbilirubinemia. Study design 5 neonatologists and 17 nurses estimated 3,532 BiliEye in 1,129 term and late preterm (>= 35 weeks) infants before discharge from the nursery, at 62 +/- 24 hours. Total serum bilirubin (TSB) levels were measured concomitantly. Results Mean TSB and BiliEye were 6.7 +/- 2.9 mg/dL (range, 0.4-18.2 mg/dL) and 6.6 +/- 3.2 mg/dL (range, 0.0-17.2 mg/dL), respectively, with good correlation (Pearson's r = 0.752, P < .0001), but other measures of agreement were poor. 61.5% of the 109 babies with TSB levels in high-risk zones were clinically misclassified. The area under curve (AUC) of the receiver-operating characteristics plotted for these high-risk zones was 0.825, but became low for early discharge (:536 hours; AUC = 0.638) and late preterm (35-37 weeks; AUC = 0.613). There was significant interobserver variation (low weighted kappa, 0.363). Conclusions Although there was good correlation between BiliEye and actual TSB level, visual assessment was unreliable as a screening tool to detect significant neonatal hyperbilirubinemia before discharge. Babies with TSB levels within high-risk zones may be clinically misdiagnosed as low- risk, resulting in inadequate follow-up.