Page 100 - 16Neonatal Jaundice_compressed

Recognition

the reported thresholds of transcutaneous bilirubin levels in the included studies were variable,
results were pooled using the summary ROC curve analysis and the area under the ROC curve
(AROC) was calculated. In order to obtain a baseline test performance value for the various
tests, their sensitivity and specificity were also pooled using the random effects model.

5.1 Visual/clinical examination

5.1.1 Examining the baby

Description of included studies
Seven studies 55-61 have been included in this section. Three studies 57;59;60 were carried out in the
USA and two each in Israel 58;61 and Switzerland. 55;56 All the studies evaluated the correlation of
57
clinical assessment of jaundice by experienced healthcare professionals, while one study also
59
evaluated parental assessment. Six studies 55-58;60;61 were conducted in a hospital setting and one
61
in a community setting. One study was of EL I and the remaining six 55-60 were of EL II.
Review findings
61
In the first study, from Israel, 1129 term and late preterm babies of Jewish (73%) and Arab
(26%) ethnicity were clinically assessed for jaundice by experienced clinicians (five
neonatologists and 17 nurses) in a hospital. All the babies were examined by the observers for
cephalo-caudal progression of jaundice, and they were unaware of the serum bilirubin levels
that were collected simultaneously at the time of visual inspection. The clinical assessment
(called ‘BiliEye’) and serum bilirubin values were grouped into risk zones according to a
nomogram developed by Bhutani et al., and the ability of BiliEye to detect significant
34
hyperbilirubinaemia (defined as zones C and D on the nomogram) was analysed by calculating
the area under the ROC curve.
Although BiliEye and serum bilirubin values were moderately positively correlated (r = 0.75;
P < 0.001), there was generally a poor agreement between the different observers (κ = 0.363)
for the degree of clinical jaundice. Visual assessment also led to a high false-negative rate, that
is, a large number of babies were misclassified into either the lower or higher risk zones and
61.5% (67 of 109) of babies with serum bilirubin in the high-risk zones (zones C and D on the
nomogram) were clinically misclassified as being in the lower risk zones. Moreover, 8.1% (230
of 2857) of babies with clinical estimation determined to be in zone A had serum bilirubin
values in the higher risk zones (zone B, C or D), indicating that BiliEye readings in the low-risk
zone had an NPV of 92% in ruling out serum bilirubin values in the higher risk zones. The area
under the ROC curve plotted for the high-risk zones C and D was 0.82. After adjusting for
postpartum age and gestational age, the best results for the diagnostic accuracy of BiliEye to
detect significant hyperbilirubinaemia were seen when the observations were made after
60 hours of age in babies ≥ 37 weeks gestational age (AROC = 0.93). The results were poor for
observations made before 36 hours of age (AROC = 0.64) and in babies born at less than
37 weeks gestational age (AROC = 0.61). [EL I]
The second study was conducted in an urban public hospital in the USA. The sample
60
population comprised 122 healthy full0term babies with jaundice, with a Rhesus-negative
mother or with a positive DAT. Two observers (paediatric residents, nurse practitioners or
physicians) independently recorded their clinical assessment of jaundice in babies for pre-
specified parts of the body, and serum bilirubin was measured within 1 hour of the assessment.
The clinical assessment included subjective evaluation of jaundice at each site (absent, slight or
obvious), subjective evaluation of the skin tone (light or dark), and estimation of serum bilirubin
level based on clinical appearance. Ethnic origins were not recorded. Results of the clinical
assessment were kept in sealed envelopes until serum bilirubin results were available. Although
there was good agreement between pairs of observers regarding the baby’s skin tone (κ = 0.56),
agreement for jaundice at each site was generally poor (only marginally better than chance),
with the best agreement seen at the ‘nipple line to umbilicus’ site (κ = 0.23, 95% CI 0.09 to
0.38). Linear correlation between the estimated serum bilirubin levels and actual serum
bilirubin levels was poor but statistically significant (r = 0.43 and 0.54 for the two groups of

71