Article Text

Download PDFPDF

Clinical examination could not accurately predict neonatal jaundice
  1. Max Perlman, MBBS
  1. Hospital for Sick Children and University of Toronto Toronto, Ontario, Canada

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

 QUESTION: How accurate are clinicians in assessing neonatal jaundice?


    Blinded comparison of clinicians' visual observations with serum bilirubin test results.


    Well newborn nursery of a hospital in Houston, Texas, USA.


    122 healthy full term infants (mean age 2 d, 54% boys).

    Description of tests and diagnostic standard

    The infants were observed by 2 clinicians independently (paediatric resident, nurse practitioner, or attending physician) under fluorescent lighting near a window. The clinicians assessed jaundice as being absent, slight, or obvious for prespecified body zones starting at the head and progressing down the body. The record of each clinical assessment was placed in a sealed envelope before the serum bilirubin test was done.

    Main outcome measures

    Agreement between the 2 observers and correlation with serum bilirubin concentrations. The cut point for clinically significant jaundice was a bilirubin concentration of ≥205 μmol/l.

    Main results

    Agreement between observers for jaundice for each part of the body was poor; κ values ranged from –0.06 for soles of the feet to 0.23 for the area of nipple line to umbilicus (0% to 23% agreement beyond chance alone). The absence of visible jaundice below the nipple line predicted that an infant would have a bilirubin concentration <205 μmol/l, but the presence of jaundice below the nipple line did not reliably predict higher bilirubin concentrations. The table shows the sensitivity, specificity, and likelihood ratios.

    Test characteristics for visual observation of jaundice in full term infants*


    Clinical assessment of jaundice by experienced observers was not reliable and could not accurately predict increased serum bilirubin concentrations.


    The study by Moyer et al strongly reinforces existing evidence that visual assessment of neonatal jaundice is unreliable and inaccurate, even when estimates are made of the cephalocaudal progression of jaundice to different body zones. The sensitivity of clinical observation of high bilirubin concentrations (jaundice extending below the nipple line) was greater than the specificity, implying that high serum bilirubin concentrations are less predictable than low concentrations. 1 of the 2 infants with false negative predictions (serum bilirubin concentrations >205 μmol/l whose jaundice stopped above the nipple line) had a serum bilirubin concentration of 279 μmol/l.

    The inclusion criteria and definition of variables of the study may bias the results. The range of gestational ages from 37 to 41 weeks, variation in skin colour of the patients, and range of postnatal ages are clinically relevant. The risks for underestimating jaundice and kernicterus are probably greater for infants of 37 to 38 weeks of gestational age than for those of higher gestational ages, for infants with dark skin tone, and for infants on day 2 rather than on day 7. Day 2 is often the day of discharge from the hospital of mother and infant. Most of the infants we have seen with kernicterus since its resurgence in the 1990s (apparently associated with earlier discharges) had G-6-PD deficiency, had dark skin tone, and were discharged early. A serum bilirubin concentration >205 μmol/l is far above the 90th centile for normal full term infants on day 2, especially early in the day.1

    Of interest was the failure of the clinicians to show a learning curve, consistent with the finding that naive parents made better judgments of the presence or absence of jaundice than did healthcare workers.2 The possibility that variation in predictive accuracy may depend on observer perception, cited by the authors, suggests that clinicians should audit their own predictive performance to detect any consistent tendency to underestimate or overestimate jaundice. The present study will add fuel to the search for non-invasive technology to accurately assess jaundice.


    View Abstract


    • Source of funding: no external funding.

    • For correspondence: Dr V A Moyer, Department of Pediatrics, University of Texas–Houston Health Science Center, 5656 Kelley Street, Houston, TX 77026, USA. Fax +1 713 636 5844.