Page 89 - 16Neonatal Jaundice_compressed

Neonatal jaundice

bilirubin risk zones (> 95th centile to < 40th centile), sensitivity increased but neither test
could predict hyperbilirubinaemia with more than 98% sensitivity without seriously
compromising specificity (13% for risk factor score, 21% for serum bilirubin risk zone). [EL II]
In the last study, from the USA, the predictive accuracy of clinical risk factors, pre-discharge
14
bilirubin levels expressed as risk zones, and a combination of pre-discharge bilirubin and
additional risk factors was evaluated prospectively. Study methodology and population is
described in detail in Chapter 3 on risk factors. All babies (n = 812) had pre-discharge bilirubin
measured before 52 hours of age with daily transcutaneous bilirubin readings from the forehead
using BiliChek, and these were recorded on the hour-specific nomogram. Bilirubin levels
(transcutaneous bilirubin or serum bilirubin) were also measured on day 3–5 in all babies either
in hospital or at home. If transcutaneous bilirubin readings exceeded the 75th centile or were
≥ 205 micromol/litre, blood samples were taken for laboratory serum bilirubin measurement.
Both the transcutaneous bilirubin and serum bilirubin readings were expressed as risk zones on
the hour-specific nomogram. In cases where both transcutaneous bilirubin and serum bilirubin
levels were measured in the same baby, the serum bilirubin readings were used for the final
analysis. Information on clinical risk factors was extracted from hospital records, and their
association with hyperbilirubinaemia assessed using univariate analysis. The variable most
strongly associated with an increased risk of hyperbilirubinaemia was the pre-discharge bilirubin
level. As this was included in a separate model, the final clinical risk model included five other
factors: gestational age, gender, intended method of feeding, black race and extent of jaundice.
Using logistic regression modelling, the accuracy of three tests was compared for the prediction
of significant hyperbilirubinaemia. In all, 6.4% of babies developed hyperbilirubinaemia,
(bilirubin levels on day 3–5 exceeding or within 17 micromol/litre of the hour-specific AAP
phototherapy treatment thresholds). The predictive accuracy of pre-discharge bilirubin risk zone
assignment was not significantly different from that of multiple risk factors (c-statistic 0.88 versus
0.91). After combining clinical risk factors with pre-discharge bilirubin risk zone assignment, the
only factors that remained statistically significant were gestational age and percentage weight
loss per day. This combination model showed improved predictive accuracy (c-statistic 0.96)
when compared with the pre-discharge bilirubin levels. [EL II]

Evidence summary
Results from two studies with EL II indicate that pre-discharge serum bilirubin plotted on hour-
specific percentile charts (‘nomograms’) shows good accuracy in predicting subsequent
hyperbilirubinaemia. The studies used different threshold values and definitions of
hyperbilirubinaemia. In one study, two consecutive serum bilirubin readings plotted on the
nomogram had greater predictive accuracy than a single measurement. Another study with EL I
indicated that pre-discharge transcutaneous bilirubin plotted on an ‘hour-specific’ nomogram of
bilirubin levels generated from a study of healthy babies could predict hyperbilirubinaemia with
100% sensitivity and 88% specificity. The threshold values for defining hyperbilirubinaemia
were different for the transcutaneous (≥ 75th centile) and serum (≥ 95th centile) bilirubin
levels. Other studies have compared the predictive accuracy of clinical risk index scores with
pre-discharge bilirubin levels. Their results suggest that pre-discharge bilirubin is more accurate
in predicting subsequent hyperbilirubinaemia than clinical risk factors alone, but the best results
are seen when pre-discharge bilirubin measurement is combined with risk factors. A major
limitation of the evidence is that the hour-specific bilirubin nomogram was devised using a
small population of babies in a single city, and that babies with conditions such as ABO
incompatibility were excluded. The nomogram may not, therefore, be applicable to other
populations of newborn infants. Similar nomograms need to be devised for other populations.

GDG translation from evidence
Current evidence suggests that it is possible to identify babies who are likely to develop
significant hyperbilirubinaemia using a pre-discharge assessment. The GDG considered that
assessment of risk factors was important.

Another approach has been based on hour-specific bilirubin estimation. Hour-specific bilirubin
levels were interpreted using a nomogram such as that devised by Bhutani et al., but the
universal application of hour-specific bilirubin estimation could not be relied on as data were

60