Sie sind auf Seite 1von 11

DOI:10.1542/peds.103.1.

6
1999;103;6-14 Pediatrics
Vinod K. Bhutani, Lois Johnson and Emidio M. Sivieri
Newborns
Subsequent Significant Hyperbilirubinemia in Healthy Term and Near-term
Predictive Ability of a Predischarge Hour-specific Serum Bilirubin for
This information is current as of March 21, 2006
http://www.pediatrics.org/cgi/content/full/103/1/6
located on the World Wide Web at:
The online version of this article, along with updated information and services, is
reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.
Village, Illinois, 60007. Copyright 1999 by the American Academy of Pediatrics. All rights
trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove
and publication, it has been published continuously since 1948. PEDIATRICS is owned, published,
PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly
by on March 21, 2006 www.pediatrics.org Downloaded from
Predictive Ability of a Predischarge Hour-specific Serum Bilirubin for
Subsequent Significant Hyperbilirubinemia in Healthy Term
and Near-term Newborns
Vinod K. Bhutani, MD; Lois Johnson, MD; and Emidio M. Sivieri, MS
ABSTRACT. Objective. To assess the predictive abil-
ity of a universal predischarge serum bilirubin measure-
ment to screen for risk of subsequent significant hyper-
bilirubinemia in the direct Coombs negative healthy
term and near-term newborn during the first postnatal
week.
Methods. Total serum bilirubin (TSB) levels were ob-
tained at the time of the routine metabolic screen in all
term and near-term newborns cared for in the Pennsyl-
vania Hospital Well Baby Nursery (n 13 003). Postnatal
age (in hours) at the time of TSB measurement was
recorded. A percentile-based bilirubin nomogram for the
first week was constructed from hour-specific predis-
charge and postdischarge TSB values of newborns (n
2840; median BW 3230 g and median gestational age
39 weeks) who met classification criteria for healthy new-
borns (excluding those with a positive direct Coombs test
or those requiring phototherapy before age 60 hours) and
who were enrolled in a hospital supervised home or
outpatient follow-up program. The accuracy of the pre-
discharge TSB as a predictor of subsequent degree of
hyperbilirubinemia was determined.
Results. The study patients in the nomogram were
racially diverse. Nearly 60% were breastfed. Predis-
charge, 6.1% of the study population (172/2840) had TSB
values in the high-risk zone (>95th percentile) at 18 to 72
hours; of these, 39.5% (68/172) remained in that zone
(likelihood ratio [LR] 14.08, sensitivity 54%; speci-
ficity 96.2%, probability 39.5%). Predischarge, 32.1%
of the population (912/2840) had TSB values in the inter-
mediate-risk zone. In a clinically significant minority of
these newborns (58/912 or 6.4%), the postdischarge TSB
moved into the high-risk zone (LR of this move: 3.2 from
the upper-intermediate zone and .48 from the lower-
intermediate risk zone). The predischarge TSB in 61.8%
of the newborns (1756/2840) was in the low-risk zone
(<40th percentile) and there was no measurable risk for
significant hyperbilirubinemia (LR 0, sensitivity
100%; specificity 64.7%; probability 0%).
Conclusions. An hour-specific TSB before hospital dis-
charge can predict which newborn is at high, intermediate
or low risk for developing clinically significant hyperbil-
irubinemia (specifically defined as TSB levels >95th per-
centile for age in hours). Risk designation and subsequent
increases or decreases of in TSB can be easily monitored on
an hour-specific percentile based predictive bilirubin no-
mogram. A predischarge TSB measured as a universal pol-
icy would facilitate targeted intervention and follow-up in
a safe, cost-effective manner. In conjunction with bilirubin
practice parameter of the American Academy of Pediatrics,
it could reduce the potential risk for bilirubin-induced neu-
rologic dysfunction. Pediatrics 1999;103:614; hyperbiliru-
binemia, jaundice prediction, kernicterus, universal bilirubin
screen, hour-specific bilirubin nomogram.
ABBREVIATIONS. TSB, total serum bilirubin; GA, gestational
age; BW, birth weight; G-6-PD, glucose-6-phosphate dehydroge-
nase; AAP, American Academy of Pediatrics; LR, likelihood ratio;
ROC, receiver operating characteristic; BIND, bilirubin-induced
neurologic dysfunction.
N
eonatal hyperbilirubinemia remains a public
health concern as documented by recent re-
ports of kernicterus in otherwise healthy
term and near-term newborns born and cared for in
the United States.
17
Kernicterus in such newborns is
preventable, provided excessive hyperbilirubinemia
for age is promptly identified and appropriately
treated.
817
With the intent to facilitate such identifi-
cation and treatment, universal screening for severity
of bilirubinemia before hospital discharge may pre-
dict that extraordinary segment of the neonatal pop-
ulation that is at risk for excessive hyperbiliru-
binemia during the first week after birth.
This article will present data on the predictive
value of a routine predischarge total serum bilirubin
(TSB) measured at the time of the universal meta-
bolic screen. This value is plotted on an hour-specific
bilirubin nomogram that describes the range of TSB
levels observed in a population of racially diverse,
healthy, term and near-term newborns during the
first postnatal week. The percentile curves for TSB
values, in mg/dL, are marked on the nomogram,
which shows that TSB levels of 8 mg/dL (137
M/L) at about 24 hours, 14 mg/dL (239 M/L) at
about 48 hours and 17 mg/dL (290 M/L) at about
84 hours age are above the 95th percentile for post-
natal age in hours. Such levels of hyperbilirubinemia
have been deemed significant and are generally con-
sidered to require close supervision, possible further
evaluation, and sometimes intervention if brain dam-
age is to be prevented without resort to exchange
transfusion.
1821
The current nomogram appears to
identify which newborns are at low, intermediate or
high risk for reaching bilirubin levels above the 95th
percentile in the first week after birth. It will thereby
From the Section on Newborn Pediatrics, Pennsylvania Hospital, University
of Pennsylvania Health System, Philadelphia, Pennsylvania.
Received for publication Apr 30, 1998; accepted Jul 13, 1998.
Reprint requests to (V.K.B.) Newborn Pediatrics, Pennsylvania Hospital,
800 Spruce St, Philadelphia, PA 19107.
PEDIATRICS (ISSN 0031 4005). Copyright 1999 by the American Acad-
emy of Pediatrics.
6 PEDIATRICS Vol. 103 No. 1 January 1999
by on March 21, 2006 www.pediatrics.org Downloaded from
allow for a more targeted, safe, and cost-effective
follow-up than is currently available.
The appeal of the predictive nomogram is its sim-
plicity and immediate applicability. The develop-
ment of more accurate noninvasive technologies for
measuring bilirubin in the skin will facilitate its use.
Concurrent estimation of bilirubin production
by measuring exhaled carbon monoxide (COSTAT,
Natus, Inc, Palo Alto, CA) should enhance its role
as a predictive and interventional tool.
22,23
PATIENTS AND METHODS
Subjects
The population base for developing the nomogram consisted of
all 17 854 live births that occurred at Pennsylvania Hospital dur-
ing 1993 to 1997. Of these, 13 003 healthy term and near-term
newborns were eligible for discharge by day 1 or 2 (vaginal births)
or by day 3 (cesarean births). All these newborns had predischarge
TSB level obtained at the same time as the routine metabolic
screen. In some, earlier TSB values were obtained for clinical
reasons. Newborns who had postdischarge TSB levels obtained
over the next 1 to 6 days in a hospital supervised follow-up
program were eligible for inclusion in the nomogram.
Inclusion Criteria
Term or near-term appropriate for gestational age (GA) new-
borns, as defined by a birth weight (BW) 2000 g for 36 weeks
GA or BW 2500 g for 35 weeks GA.
Exclusion Criteria
Admission and treatment in the intensive care nursery for
neonatal illness or, positive direct Coombs test. All newborns
whose mothers had blood type O, were Rh-negative, or had a
positive indirect Coombs test were evaluated for blood type and
direct Coombs test. TSB values measured after the initiation of
phototherapy were excluded from the nomogram but were doc-
umented and recorded. TSB values not measured at the hospital
laboratory were excluded but were replaced by a repeat, hospital-
based measurement close in time. In addition, after the initial
analysis, newborns who required phototherapy before age 60
hours to control unexplained rapidly rising TSB levels were ex-
cluded from the predictive nomogram (rationale listed in Discus-
sion). During this study period there was no predischarge screen-
ing for glucose-6-phosphate dehydrogenase (G-6-PD) deficiency.
Initial Bilirubin Assessment
Newborns studied in 1993 and 1994 (n 1042) included those
healthy newborns who were electively discharged on day 1 in
accordance with managed care recommendations. All these con-
secutively discharged newborns, with universal TSB measure-
ments between age 20 to 28 hours and with no major blood type
(ABO) or Rh isoimmunization (negative direct Coombs test) were
prospectively followed to assess the ability of the predischarge
bilirubin to predict degree of subsequent hyperbilirubinemia.
24
Based on these preliminary observations, the Section on Newborn
Pediatrics at Pennsylvania Hospital recommended universal bili-
rubin screening for all newborns at the time of the routine meta-
bolic screen. Newborns studied during 1994 to 1997 were usually
discharged on day 2 and included newborns discharged on day 3
after cesarean section delivery.
Follow-up
Outpatient follow-up was according to a Pennsylvania Hospital
supervised early discharge protocol. Subsequent TSB levels were
usually obtained within 24 to 48 hours after discharge and as
needed thereafter. This follow-up was offered to all parents and
private pediatricians and was generally accessible by insurance
coverage that allowed home nursing care. TSB samples were
obtained in the hospital outpatient department or by the home
care nurse and transported to the hospital laboratory for analysis.
Additional follow-up either involved a repeat TSB sample or a
visual inspection at physicians discretion. Resolution of hyperbi-
lirubinemia was confirmed at about age 10 days, usually through
contact with the private pediatrician. This hospital-based moni-
toring was performed in cooperation with private pediatricians
and sought to minimize tests, investigations, and interventions.
All bilirubin samples were obtained after parental consent.
Bilirubin Assay
Serum bilirubin assay was performed by the 2,5-dichlorophe-
nyldiazonium tetrafluoroborate (DPD) diazo method (Hitachi,
747) in the Pennsylvania Hospital Ayer Clinical Laboratories.
Standard (bed) National Institute of Standard Technology guide-
lines were followed to maintain the accuracy and precision of the
technique.
2325
The coefficient of variation for the hospital labora-
tory was targeted for 6%. During the study period, each of the
actual variance values, assessed every 3 months, ranged from 2%
to 3%.
Interventions
Hospital-based or home-based phototherapy was initiated at
the discretion of the pediatrician. Unless modified for specific
clinical reasons, guidelines for phototherapy published by the
American Academy of Pediatrics (AAP) were used.
19,20
As stated
earlier, postphototherapy TSB values were excluded from the
nomogram. Pediatricians and nursing staff recommended breast
milk feedings as instructed, supervised, and monitored by two
full-time lactational consultants. Early and frequent breast feed-
ings were encouraged. The ability to latch and the adequacy of
intake were monitored and documented. Onset of jaundice or
need for phototherapy did not lead to cessation of breastfeedings.
In the event of presumed inadequate intake, feeding techniques
were evaluated by joint consultation with the mother, lactational
consultants, and the pediatric team. Supplements were used as
needed and breast milk production was enhanced with mechani-
cal pumping.
Clinical and Demographic Risk Factors for
Hyperbilirubinemia
The following data were recorded: BW, GA, gender, racial
background, history of family members or siblings with jaundice,
maternal blood type, blood type, and direct Coombs of the infant
(when tested). In addition, mode of delivery (especially use of
vacuum or forceps assistance), presence of bruising or hemato-
mas, and history of possible Gilberts disease or suspected G-6-PD
deficiency were recorded on the log sheet used for outpatient
follow-up. Also recorded were the type of feeding, use of lactation
consultants, assessment of breast milk intake, type and amount of
formula supplements, stooling pattern (frequency, amount, and
color), and timing and duration of phototherapy.
DATA ANALYSIS
The nomogram database includes all measured hour-specific
TSB values except for that relatively small number of values
obtained before age 18 hours. Data were recorded in epochs of 4
hours (or, age 2 hours) for the first 48 hours and in epochs of 12
hours (or age 6 hours) until 96 hours age and at epochs of 24
hours (or age 12 hours) for age 5 to 7 days. For each epoch at
least 300 data points and demonstration of a Gaussian distribution
were required for inclusion in the nomogram. From these data,
hour-specific TSB percentiles for each of the epochal periods were
calculated.
Statistical Analysis
The predictive ability of a predischarge TSB value, character-
ized by postnatal age in hours and measured between 18 to 72
hours, was assessed based on the frequency of any subsequent
significant hyperbilirubinemia (a subsequent TSB value in the
high-risk zone, 95th percentile, regardless of age). The relative
effectiveness of this as a vector was defined by a binary outcome
of either developing or not developing this specified level of
hyperbilirubinemia (designated as disease or no disease for the
purpose of explaining the statistical approach). The 5th, 25th, 40th,
50th, 75th, 90th, and 95th percentiles of TSB values were deter-
mined from the Gaussian distribution for each epoch and con-
nected as percentile tracks.
Predictive ability of the hour-specific TSB vector was assessed
for values above and below the percentile tracks that were used as
ARTICLES 7
by on March 21, 2006 www.pediatrics.org Downloaded from
risk demarcators. We calculated the probability, sensitivity, and
specificity for values above and below the 40th, 50th, 75th, and
95th percentiles. Intervals above, between, and below the demar-
cators were defined as zones. The zone above the 95th percentile
was labeled as high-risk and that below the 40th percentile as
minimal-risk or low-risk. TSB values between 40th and 95th per-
centiles were designated as being in the intermediate-risk zone;
this zone was further subdivided by the 75th percentile into up-
per- and lower-intermediate risk zones.
With the aid of software technology, it would have been pos-
sible to compute by logistic regression the risks associated with
individual TSB (based on a regression model dependent on TSB
level, age and the interaction of age, and TSB level). Previous
studies have shown that individual TSB values (defined by age in
days) might not predict outcome with a safe false-negative and
false-positive rate.
2830
We chose a nonparametric assessment of
the likelihood ratio (LR) for a positive result for each of the
age-specific percentile ranges or risk zones.
31
LR for a positive
result was determined for multiple levels of TSB results as located
by the density distribution in one of the risk zones.
3133
We then
developed a receiver operating characteristic (ROC) curve that
plotted the false-positive rate versus the true positive rate corre-
sponding to the LR for each risk zone. As an ROC curve moves
toward the upper left corner, its usefulness as a diagnostic tool is
seen to increase as sensitivity and specificity are maximized.
34
Visual Graphic Analysis Using the Zone-based
Predictive Nomogram
The zone-based predictive nomogram graphic helps summa-
rize serial TSB values in an individual infant. This allows for
immediate visual recognition of true positive and false-negative
predictions by plotting the successive risk zone positions of the
serial TSB values of each infant. In addition, the possible contri-
bution of an individual clinical or demographic risk factor can
now be more easily related to the change in risk zone positions.
Computer-based Data Collection and Graphic
Evaluation
Hour-specific bilirubin values, clinical and demographic risk
factors, relevant clinical interventions and follow-up information
were stored in a computer database program that displayed the
data for each individual newborn. The software program showed
a background graphic display of the nomograms percentile tracks
and risk zones. The postnatal age in hours corresponding to birth
time and the time of TSB sampling was automatically calculated
for each sample. Thus, serial TSB samples, accurately character-
ized by postnatal age (in hours), were available for automatic
plotting and delineation of risk status. In this way, if TSB values in
an individual newborn moved from one risk zone to another, the
rate of increase or decrease and related change in risk status was
immediately apparent. The progress of newborns whose TSB val-
ues remained consistently in their original (predischarge) percen-
tile zone was also visually apparent. Each of these graphically
displayed, computerized follow-up charts was scrutinized for its
accuracy, risk designation, and outcome both manually and by
computerized assessment.
RESULTS
During the entire study period, a total of 17 854
live births occurred at Pennsylvania Hospital, of
which 14 793 were directly admitted to the well-baby
nursery. Of these, 13 003 near-term and term new-
borns matched all inclusion criteria. From this pop-
ulation base, 2976 completed all requirements for the
hospital-supervised follow-up and had at least one
postdischarge TSB level measured in the hospital
laboratory. From these, 118 were excluded because of
transfer to the intensive care nursery for such diag-
noses as sepsis proven or presumed but treated,
intractable hypoglycemia or respiratory distress. In
addition, 18 newborns who required phototherapy
before age 60 hours because of rapidly rising predis-
charge bilirubin levels were excluded. The remaining
2840 healthy term and near-term newborns eligible
for discharge at age 24 to 72 hours constituted the
study population on which the predictive nomogram
was developed. The mean SD value for birth
weight was 3318 457 g (median: 3230 g) and for
gestational age 38.7 1.3 weeks (median: 39 weeks).
The racial distribution, mode of delivery, and mode
of feeding are tabulated in Table 1. These parameters
do not differ significantly from those of the entire
study pool of 13 003 well newborns.
The mean age for predischarge TSB sampling for
the 2840 newborns was 33.7 14.6 SD hours. Hy-
perbilirubinemia severe enough to cause visible
jaundice was often present at the time of the first
predischarge sample: 13.4% had a TSB 10 mg/dL,
4.3% of the newborns had values 12 mg/dL, .4%
(12/2840) had values 15 mg/dL and 2 newborns
had values 18 mg/dL. The Gaussian distribution of
all TSB data points (range: 300673 per epoch) for
each of the epochal periods from 18 to 132 hours age
is shown in Fig 1. The 5th, 10th, 25th, 40th, 50th, 75th,
90th, and 95th TSB percentiles were constructed and
the 40th, 75th, and 95th were used as risk zone de-
marcators (Fig 2).
A total of 230/2840 (8.1%) newborns were identi-
fied as having values above the 95th percentile track
at some time (predischarge or postdischarge) during
the first postnatal week. However, as shown in Table
2, only 126/2840 (4.4% or 1 of 23 newborns) had a
postdischarge TSB value in the high-risk zone (sig-
nificant hyperbilirubinemia). Based on these data,
the ratio of newborns with and without subsequent
significant hyperbilirubinemia is 126:2714 (or 1:22)
for our study population. Table 2 lists the predictive
ability of the 40th, 75th, and 95th percentile tracks as
risk demarcators. The ROC curve, as shown in Fig 3,
illustrates each risk zones significant ability to pre-
dict subsequent significant hyperbilirubinemia.
The LR that determines the risk assessment for
subsequent significant hyperbilirubinemia for each
predischarge risk zone is tabulated in Table 3. For the
2 to 4 days after discharge, TSB levels of most new-
borns remained in the predischarge TSB percentile-
based risk zone and subsequently decreased to lower
TABLE 1. Demographic Characteristics of the Study Popula-
tion
Demographics (n 2840) Percentage
Gender (female:male) 49.9:50.1%
Race:
White 43.4%
African-American 41.2%
Hispanic 3.6%
Asian 4.1%
Other 7.7%
Mode of delivery:
Spontaneous vaginal 77.9%
Assisted vaginal 12.5%
Elective cesarean 3.2%
Emergency cesarean 6.5%
Mode of feeding:
Breastfeeding only 49.5%
Formula feeding 40.8%
Breastfeeding plus supplement 9.9%
Phototherapy 4.1%
8 PREDICTIVE ABILITY OF A PREDISCHARGE HOUR-SPECIFIC SERUM BILIRUBIN
by on March 21, 2006 www.pediatrics.org Downloaded from
risk zones either spontaneously or, with individual-
ized nutritional counseling. Among the newborns
with a TSB in the high-risk zone predischarge (172/
2840 or 6.1% of the study population), 68 continued
to have subsequent significant hyperbilirubinemia.
On the other hand, in 104 the subsequent TSB de-
creased below the 95th percentile (Fig 4A) and the
ratio of newborns with or without disease was 2:3
and LR 14.08. TSB levels of a small but significant
number from the intermediate-zone newborns (58/
912, 6.4%) moved upwards to the high-risk zone after
discharge. Of 356 newborns in the upper intermedi-
ate-risk zone, 46 jumped to the high-risk zone on
follow-up and 310 did not (Fig 4B; ratio of newborns
with and without disease was 1:7 and LR 3.20).
This compared with the 556 newborns in the lower
intermediate-risk zone. Of these, 12 jumped tracks
into the high-risk zone on follow-up and 544 did not
(Fig 4C; ratio of newborns with and without disease
was 1: 45 and LR .48). Another 29 of these 556
newborns (5.2%) changed their risk status by moving
upwards but only into the upper intermediate-risk
zone.
Follow-up of newborns placed in the low-risk zone
Fig 1. Gaussian distribution of TSB values at se-
lected study epochal periods up to 72 hours.
Fig 2. Risk designation of term and near-term well newborns based on their hour-specific serum bilirubin values. The high-risk zone is
designated by the 95th percentile track. The intermediate-risk zone is subdivided to upper- and lower-risk zones by the 75th percentile
track. The low-risk zone has been electively and statistically defined by the 40th percentile track. (Dotted extensions are based on 300
TSB values/epoch).
ARTICLES 9
by on March 21, 2006 www.pediatrics.org Downloaded from
at discharge (1756/2840; 61.8%) showed them to be
the most predictable. Nearly 93.6% remained in the
40th percentile-risk zone; while, only 6.4% moved up
to the intermediate-risk zone. None (LR 0) jumped
up to the high-risk zone (Fig 4D). The only presum-
able exception could have been a near-term neonate
(newborn D.F., BW3335 g, GA 36 weeks, breast-
fed by a primiparous mother) who had a predictive
TSB value at the 40th percentile (4.3 mg/dL at 18
hours age) and was discharged early at maternal
request. On follow-up, at age 47 hours the TSB was
11.3 mg/dL (upper-intermediate risk zone) and
reached a value of 19.3 mg/dL at age 125 hours
(high-risk zone). Her clinical evaluation included
demonstration of lymphocytosis that was suggestive
of an acute viral infection; no other clinical risk fac-
tors were identified. Both phototherapy and nutri-
tional support were provided. Hyperbilirubinemia
resolved over the next several days.
Phototherapy was usually initiated when at least
two consecutive TSB values were in the high-risk
zone. This occurred in a total of 117/2840 newborns
or 4.1% of the study population. In 15 newborns,
phototherapy was commenced between 61 to 71
hours for a mean TSB value 16.6 1.6 SD mg/dL.
For the remaining 102 newborns, phototherapy was
started after age 72 hours for a mean TSB value
17.8 2 SD mg/dL at a mean age of 89.4 25.9 SD
hours. None of the newborns in the low-risk zone
received phototherapy. No newborn in the study
population required an exchange transfusion or de-
veloped a TSB value 25 mg/dL. None developed
acute signs of bilirubin encephalopathy. None are
known to have sequelae at about 1 year of age as
determined by telephone interviews of parents, pe-
diatric offices or feedback from area hospitals. To the
best of our knowledge, none of the healthy newborns
born from 1993 to 1997, who were not included in the
nomogram because of follow-up elsewhere, devel-
oped kernicterus.
The frequency of exclusive breastfeeding in our
study population was 49.3%. An additional 9.9%
who were provided with individualized lactation
counseling and support received expressed breast
milk and formula supplementation. Percentile tracks
for exclusively breastfed newborns appeared to be
about 1 to 1.5 mg/dL higher than for the exclusively
formula fed newborns at the 40th and 95th percentile
tracks. The sample size is not adequate to draw any
meaningful conclusions from these data.
DISCUSSION
Clinical experience and recent reports in the
United States suggest an increased occurrence of ker-
nicterus in otherwise healthy newborns during the
late 1980s and 1990s.
16
This has been attributed to
decreased clinical concern about the toxic potential
of bilirubin
117
and increased administrative and eco-
nomic pressures to limit length of hospitalization
and laboratory investigations that has resulted in an
era of lessened direct medical observation.
3538
Statis-
tically, newborns with TSB of 17 mg/dL during the
first week after birth represent a small segment of the
population; our study shows that this value is above
the 95th percentile at about 84 hours age and beyond.
It is this group of newborns who are at potential risk
for bilirubin-induced neurologic dysfunction (BIND)
including kernicterus.
9,3946
Strategies to prevent
BIND need to be practical, safe, effective, and based
on risk assessment. Recognizing this as a matter of
public health concern, the AAP developed a detailed,
consensus-based, practice parameter for the manage-
Fig 3. The ROC curve for the predictive abilities of the 40th, 75th,
and the 95th percentile-based risk zones and their false-positive
and false-negative rates (%). (Its location in the upper left corner is
indicative of the usefulness of the risk zones).
TABLE 2. Predictive Characteristics of Percentile Tracks as Risk Demarcators for Subsequent Significant Hyperbilirubinemia*
Location of
Predictive Predischarge TSB Vector
Outcome: Subsequent
Significant
Hyperbilirubinemia
Predictive Characteristics
Percentile Track as
Risk Demarcator
Number of Newborns
(Total 2840)
Present
(Total 126)
Absent
(Total 2714)
Positive
Predictive
Value
Negative
Predictive
Value
Sensitivity Specificity
Above 95th percentile 172 68 104 39.5% 97.8% 54.0% 96.2%
Below 95th percentile 2668 58 2610
Above 75th percentile 528 114 414 21.6% 99.5% 90.5% 84.7%
Below 75th percentile 2312 12 2300
Above 40th percentile 1084 126 958 11.6% 100% 100% 64.7%
Below 40th percentile 1756 0 1756
* Defined as when a subsequent TSB value reaches into the high-risk zone (95th percentile track).
10 PREDICTIVE ABILITY OF A PREDISCHARGE HOUR-SPECIFIC SERUM BILIRUBIN
by on March 21, 2006 www.pediatrics.org Downloaded from
ment of hyperbilirubinemia in healthy term new-
borns and the institution of preventive photothera-
py.
19
Although effective when implemented as
intended, clinical use of the guidelines has been lim-
ited by the absence of a prospective risk assessment
and by dependence on visual assessment of jaundice.
Predischarge Jaundice as a Predictive Vector for
Subsequent Hyperbilirubinemia
Early visual recognition of jaundice and accurate
estimation of its severity is crucial for effective im-
plementation of the AAP guidelines.
19
Unfortu-
nately, the presence of excessive jaundice for age is
often missed clinically, which means that the trigger
for measuring the first serum bilirubin level and
electing subsequent AAP algorithm recommenda-
tions is not set. This is a potentially serious problem.
Variability in the time of appearance of jaundice
from newborn to newborn and in the ability of the
professionals to see jaundice and estimate its sever-
ity, coupled with the considerable range of TSB val-
ues associated with its cephalo-caudal progression,
have been the subject of articles spanning nearly 60
years.
4749
Even in the present study, with health care
providers sensitized to the significance of clinical
jaundice, there were several instances when its early
appearance was missed (often attributable to con-
founding skin coloring) or a measured TSB was not
considered excessive for the newborns age in hours.
Additionally, in most of the recently reported
healthy term newborns who developed kernicterus,
significant jaundice was almost certainly present be-
fore the first hospital discharge, judging from the
height of TSB for age in hours at readmission.
16
Either the early icterus had not been noted or its
pathologic intensity for postnatal age was not appre-
ciated. TSB values must be excessive for age in hours
in otherwise healthy term and near-term newborns
to be at potential risk for BIND. It is in the context of
identifying such newborns before dangerous levels
are reached that a universal TSB screen, before dis-
charge, is recommended as a more specific predictive
vector than clinically recognized jaundice.
Statistical Approach
The statistical steps we have used to guide clinical
decision-making are based on a nonparametric as-
sessment and make no distributional assumptions.
The nomogram has been specifically devised to be
intuitively appealing to the clinician in daily practice.
The clinician is interested in the probability that the
newborn will or will not develop significant hyper-
bilirubinemia. In this scenario, data on sensitivity
and specificity, ROC curves, and negative or positive
predictive values do not always suffice. Also, the
application of sensitivity and specificity to test re-
sults that can have a continuous distribution of val-
ues is not clinically practical.
31
In contrast, the LR has
the ability to revise the prior probability of disease
upwards or downwards based on the subsequent
test result. For continuous variables, the LR (or like-
lihood quotient) is the ratio of two probabilities, the
probability that a given predischarge TSB value pre-
dicts the risk of disease (the true positive fraction)
divided by the probability of the same test result
when there is no disease (the false-positive fraction).
The use of LR for predicting an outcome of clinically
significant hyperbilirubinemia or disease (outcome
of a TSB level in the high-risk zone) encompasses the
impact of clinical and demographic risk factors.
These contributory factors affect the density distri-
bution of the TSB values and therefore the LR sum-
marizes all these factors into a single clinically rele-
vant index.
3133
Because TSB values reflect the
combined effects of bilirubin production and hepatic
excretion, the reasons why successive TSB values
that jump track require further evaluation.
Universal Screening
Based on our data (Table 3), the probability of
subsequent hyperbilirubinemia in our study popula-
tion as a whole may be expressed as a ratio of 1:22 for
disease and no disease (126 with and 2714 without
disease). The overall risk is none in the low-risk zone,
nearly halved (1:45) in the lower-intermediate zone,
tripled in the upper-intermediate zone (1:7) and in-
creased 14-fold in the high-risk zone (2:3) newborns.
Considering the threshold used to rationalize predis-
TABLE 3. Predictive Ability of Predischarge TSB in the Percentile-Based Risk Zones for Subsequent Significant Hyperbilirubinemia*
Location of Predictive Predischarge TSB Vector Outcome: Subsequent
Significant
Hyperbilirubinemia
Predictive Ability
Predischarge Hour-specific
TSB Risk Zone
Percentiles Total Present (P)* Absent (A) P:A Ratio Probability of
Disease
Likelihood Ratio
of Disease
High-risk zone 95th 172 68 104 2:3 2/5 14.08
Upper-intermediate 76th95th 356 46 310 1:7 1/8 3.20
Lower-intermediate 40th75th 556 12 544 1:45 1/46 0.48
Low-risk zone 40th 1756 0 1756 0 0 0
Total 2840 126 2714 1:22 1/23
* Defined as a subsequent TSB value in the high-risk zone (95th percentile track)
P:A Ratio: those with significant hyperbilirubinemia (present) to those without (absent).
Based on the prevalence of disease in a given population (present/(present absent); 1/23 in our population). The likelihood ratio may
be used to calculate a revised or posttest probability of disease:posttest odds pretest odds LR where odds probability (1
probability). Because the LR is prevalence independent, it can be used to calculate the posttest probability at any site. Alternatively, from
a clinical perspective, the revised probability may be obtained at any site by dividing the absent component of the present:absent ratio
for the total population by the LR to calculate the revised present:absent ratio for each risk zone.
ARTICLES 11
by on March 21, 2006 www.pediatrics.org Downloaded from
Fig 4. Outcome of newborns as defined by the percentage that remains or moves up to the high-risk zone after their risk assessment with the predischarge bilirubin value
(represented by the shaded area). A, Outcome for newborns designated in the high-risk zone (n 172); B, outcome of newborns in upper intermediate-risk zone (n 356); C,
outcome of newborns in the lower intermediate-risk zone (n 556); D, outcome of newborns in the low-risk zone (n 1756).
1
2
P
R
E
D
I
C
T
I
V
E
A
B
I
L
I
T
Y
O
F
A
P
R
E
D
I
S
C
H
A
R
G
E
H
O
U
R
-
S
P
E
C
I
F
I
C
S
E
R
U
M
B
I
L
I
R
U
B
I
N

b
y

o
n

M
a
r
c
h

2
1
,

2
0
0
6

w
w
w
.
p
e
d
i
a
t
r
i
c
s
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

charge screening for a variety of congenital inherited
disorders, these statistics provide justification for in-
stitution of a policy of universal TSB screening. Our
recommendation for universal TSB screening per-
formed at the time of the universal metabolic screen-
ing would be cost-effective because of the targeted
bilirubin follow-up facilitated and encouraged by the
nomogram as opposed to medical decisions not
based on risk assessment.
51,52
Low-risk zone new-
borns who were screened between 18 to 72 hours age
comprised 61.8% of our study population. They rep-
resent newborns who either were in the low-risk
zone or moved into the low-risk zone postdischarge.
None of these newborns moved into the high-risk
zone, required phototherapy or other than minimal
intervention, such as counseling on feeding tech-
niques. Bilirubin follow-up of this large component
of the population, once screened, could be safely
limited to a visual assessment of jaundice by an
experienced observer. In the future, significantly
more accurate noninvasive technologies now under
development to measure bilirubin staining of the
skin and its relation to serum bilirubin level may
suffice for both predischarge and postdischarge TSB
measurements and may theoretically offer an addi-
tional estimate of toxic risk.
30,52
This nomogram, developed from a diverse popu-
lation, attempts to anticipate risk of an errant expe-
rience for a newborn who unexpectedly develops
significant hyperbilirubinemia. We initially labeled
the low-risk zone by the 50th percentile. Its predic-
tive ability was nearly similar to that of the 40th
percentile. However, in view of the unavoidable
problems
2527
in measurement of TSB (at least 0.5
mg/dL), and the fact that many near-term newborns
(35/36 weeks GA) are discharged as if they were at
term (as in the clinical case described in the Results
section), the 40th percentile was considered a safer
demarcator. We made painstaking efforts to main-
tain accuracy and precision in bilirubin measure-
ments by using a single laboratory. A consistent li-
aison and communication between a local laboratory
and the pediatrician would also be effective in mon-
itoring for variation in bilirubin accuracy. The con-
servative definition of the 40th percentile low-risk
zone demarcator minimizes the prospect of an errant
experience that might occur in a near-term newborn
with multiple risk factors, or when a universal screen
is performed at 18 rather than at age 24 hours or
more. The low-risk zone designated by our data, as
below the 40th percentile track, may not be applica-
ble worldwide. However, the LR for each of the
bilirubin defined risk zones in our nomogram can be
used at geographic sites with different prior proba-
bilities of significant hyperbilirubinemia to calculate
site-specific risk assessment (see Table 3 legend).
Newborns Who Are Placed in the High-risk Zone
Before Discharge
As described earlier, we observed 18 newborns
who for no identified reason were in the high-risk
zone and needed phototherapy before age 60 hours.
They were kept in the hospital for further evaluation
and interventions. Because their predischarge serial
TSB values had already placed them in the disease
category, they did not belong in the nomogram to
predict disease. Similarly, newborns with known in-
creased risk secondary to ABO and Rh incompatibil-
ity with a positive Coombs test were excluded. How-
ever, when included in the nomogram, the
predischarge and prephototherapy TSB values did
not appear to alter the 95th percentile demarcator.
Instead, they appeared to effect the low-risk demar-
cator at age 18 to 30 hours. As expected, the TSB
values were located disproportionately in the high-
risk or upper- or intermediate-risk zones depending
on their rate of bilirubin production and capacity for
hepatic excretion. Subsequent tracking of these new-
borns was markedly aided by the use of the nomo-
gram. Further diagnostic evaluation of any newborn
placed in the high- and intermediate-risk zone new-
borns would help characterize their hyperbiliru-
binemia as related to increased production (measure-
ment of carbon monoxide production) versus a range
of hepatic excretion defects compounded by an in-
creased entero-hepatic circulation, or to a combina-
tion of all three.
54
Furthermore, any newborn who
reaches and remains in the high-risk zone would
benefit from long-term follow-up for sequelae of
both acute and chronic BIND.
CONCLUSION
In summary, we have demonstrated the predictive
usefulness and clinical value of universal bilirubin
sampling in all term and near-term newborns per-
formed at the time of the routine metabolic screen-
ing. When related to its percentile distribution for
postnatal age, this hour-specific TSB can define the
risk of subsequent clinically significant hyperbiliru-
binemia as high-risk (95th percentile), intermedi-
ate-risk (40th95th percentiles), and low-risk (40th
percentiles). We provide risk-based guidelines that
can target evaluation, intervention, and follow-up
according to the AAPs bilirubin practice parameter.
Application of this predictive bilirubin nomogram
would help in the prevention of BIND in the United
States, including kernicterus.
ACKNOWLEDGMENTS
We thank Drs Soraya Abbasi, John G. DeMaio, Jeffrey S. Ger-
des, and Carla Weis of the Section on Newborn Pediatrics for their
clinical and administrative support; the pediatric nurse practitio-
ners: Theresa Cotton, Patricia Hewson, Diane Manning, and Bar-
bara Medoff-Cooper; the lactation consultants: Tammy Arbeter
and Susan Carson; the hospital-based bilirubin check team: Chris-
tine Dalin, Rosemary Dworanczyk, Mary Grous, Donna Spitz; and
the home visiting nurses in Philadelphia and the South New
Jersey area for their meticulous and caring attention to the jaun-
diced newborns. This research was supported by the Newborn
Pediatrics Research Fund at Pennsylvania Hospital.
REFERENCES
1. Brown AK, Johnson L. Loss of concern about jaundice and the reemer-
gence of kernicterus in full-term infants in the era of managed care. In:
Fanaroff AA, Klaus MH, eds. The Year Book of Neonatal and Perinatal
Medicine. Philadelphia, PA: Mosby Yearbook; 1996:1728
2. Penn AA, Enzmann DR, Hahn JS, et al. Kernicterus in a full term infant.
Pediatrics. 1993;93:10031006
3. Stevenson DK. Kernicterus in a full-term infant: the need for increased
vigilance. Pediatrics. 1995;95:799. Letter
4. Maisels MJ, Newman TB. Kernicterus in otherwise healthy, breastfed
ARTICLES 13
by on March 21, 2006 www.pediatrics.org Downloaded from
term newborns. Pediatrics. 1995;96:730733
5. Sola A. Changes in clinical practice and bilirubin encephalopathy in
healthy term newborns. Pediatr Res. 1995;37:145. Abstract
6. MacDonald MG. Hidden risks: early discharge and bilirubin toxicity
due to glucose-6-phosphate dehydrogenase deficiency. Pediatrics. 1995;
96:734738
7. Washington C, Ector W, Abbound K, et al. Hemolytic jaundice due to
G6PD deficiency causing kernicterus in newborn female. South Med J.
1995;88:776779
8. Newman TB, Maisels MJ. Does hyperbilirubinemia damage the brain of
healthy full-term infants? Clin Perinatol. 1990;17:3311335
9. Johnson L. Hyperbilirubinemia in the term infant: when to worry, when
to treat. N Y State J Med. 1991;91:483489
10. Newman TB, Maisels MJ. Evaluation and treatment of jaundice in the
term newborn. A kinder, gentler approach. Pediatrics. 1992;89:809818
11. Brown AK, Seidman DS, Stevenson DK. Jaundice in healthy, term
neonates: do we need new action levels or new approaches? Pediatrics.
1992;89:827829
12. Cashore WJ. Hyperbilirubinemia. Should we adopt a new standard of
care? Pediatrics. 1992;89:824826
13. Johnson L. Yet another expert opinion on bilirubin toxicity! Pediatrics.
1992;89:829831
14. Wennberg RP. Bilirubin recommendations present problems: new
guidelines simplistic and untested. Pediatrics. 1992;89:821822
15. Poland RL. In search of a gold standard for bilirubin toxicity. Pediatrics.
1992;89:823824
16. Gartner LM. Management of jaundice in the well baby. Pediatrics. 1992;
89:826828
17. Valaes T. Bilirubin toxicity: the problem was solved a generation ago.
Pediatrics. 1992;89:819821
18. Maisels MJ, Newman TB. Jaundice in full-term and near-term babies
who leave the hospital within 36 hours: the pediatricians nemesis. Clin
Perinatol. 1998;25:295302
19. American Academy of Pediatrics, Provisional Committee for Quality
Improvement and Subcommittee on Hyperbilirubinemia. Practice
parameter: management of hyperbilirubinemia in the healthy term
newborn. Pediatrics. 1994;94:558565
20. American Academy of Pediatrics, Committee on Fetus and Newborn,
and ACOG Committee on Obstetrics. Maternal and Fetal Medicine: Guide-
lines for Perinatal Care. 3rd ed. 1992:108109
21. Martinez JC, Maisels J, Otheguy L, et al. Hyperbilirubinemia in the
breastfed newborn: a controlled trial of four interventions. Pediatrics.
1993;91:470473
22. Smith DW, Inguilo D, Martin D, et al. Use of noninvasive tests to predict
significant jaundice in full-term infants: preliminary studies. Pediatrics.
1985;75:278280
23. Stevenson DK, Vreman HJ. Carbon monoxide and bilirubin production
in neonates. Pediatrics. 1997;100:252254
24. Bhutani VK, Johnson LH. Probability of subsequent hyperbilirubinemia
in term healthy newborns with no ABO/Rh disease. Pediatr Res. 1996;
39:197. Abstract
25. Schreiner RL, Glick MR. Interlaboratory bilirubin variability. Pediatrics.
1982;69:277281
26. Rosenthal P. The laboratory method as a variable in the diagnosis of
hyperbilirubinemia. Am J Dis Child. 1987;141:1066
27. Vreman HJ, Verter J, Stevenson DK, et al. Interlaboratory variability of
bilirubin measurements. Clin Chem. 1996;42:869873
28. Gale R, Seidman DS, Dollberg S, Stevenson DK. Epidemiology of neo-
natal jaundice in the Jerusalem population. J Pediatr Gastroenterol Nutr.
1990;10:8286
29. Seidman DS, Ergaz Z, Paz I, et al. Can bilirubin levels at age 24 hours
predict the risk of neonatal jaundice? Pediatr Res. 1996;39:244. Abstract
30. Knudsen A. Prediction and non-invasive assessment of neonatal jaun-
dice in the term healthy newborn infant. Acta Paediatr. 1996;85:393397
31. Radcack KL, Rouan G, Hedges J. The likelihood ratio: an improved
measure for reporting and evaluating diagnostic test results. Arch Pathol
Lab Med. 1986;110:689693
32. Ransohoff DF, Feinstein AR. Problems of spectrum and bias in evalu-
ating the efficacy of diagnostic tests. N Engl J Med. 1978;299:926930
33. Albert A. On the use and computation of likelihood ratios in clinical
chemistry. Clin Chem. 1982;5:11131119
34. McNeil BJ, Hanley JA. Statistical approaches to the analysis of receiver
operating characteristics (ROC) curves. Med Decis Making. 1984;4:
137150
35. American Academy of Pediatrics, Committee on Fetus and Newborn.
Hospital stay for healthy term newborns. Pediatrics. 1995;96:788790
36. Braveman P. Problems associated with early discharge of newborn
infants: early discharge of newborn and mother. A critical review of the
literature. Pediatrics. 1995;96:716726
37. Braveman P, Kessel W, Egerter S, Richmond J. Early discharge and
evidence-based practice: good science and good judgment. JAMA. 1997;
278:334336
38. Kessel W, Kiely M, Nora AH, Sumaya CV. Early discharge: in the end,
it is judgment. Pediatrics. 1995;96:739742
39. Johnson LH, Boggs TR. Bilirubin-Dependent Brain Damage: Incidence and
Indications for Treatment. Phototherapy in the Newborn: An Overview.
Washington, DC: National Academy of Sciences; 1974:122149
40. Perlman M, Fainmesser P, Sohmer H, Tamari H, Wax Y, Pevsmer B.
Auditory nerve-brainstem evoked responses in hyperbilirubinemia ne-
onates. Pediatrics. 1983;72:658664
41. DeVries LS, Whitelaw AG, Dubowitz LMS. Relationship of serum bili-
rubin levels and hearing impairment in newborn infants. Early Hum
Dev. 1987;15:269277
42. Nakamura H, Takada S, Shimabuku R, et al. Auditory nerve and
brainstem responses in newborn infants with hyperbilirubinemia. Pedi-
atrics. 1986;75:703708
43. Maisels MJ, Newman TB. Kernicterus in otherwise healthy, breastfed
term newborns. Pediatrics. 1995;96:730733
44. van der Bor M, Ens-Dokkum M, Schreuder AM, et al. Hyperbiliru-
binemia in low birth weight infants and outcome at 5 years of age.
Pediatrics. 1992;89:359364
45. Newman TB, Klebanoff MA. Neonatal hyperbilirubinemia and long-
term outcome: another look at the Collaborative Perinatal Project. Pe-
diatrics. 1993;92:651657
46. Funato M, Tamai H, Shimada S, et al. Vigintiphobia, unbound bilirubin,
and auditory brainstem responses. Pediatrics. 1994;93:5053
47. Davidson LT, Merritt KK, Weech AA. Hyperbilirubinemia in the new-
born. Am J Dis Child. 1941;61:958980
48. Kramer LI. Advancement of dermal icterus in the jaundiced newborn.
Am J Dis Child. 1969;118:454458
49. Ebbesen F, Brodersen R. Risk of bilirubin acid precipitation in preterm
infants with respiratory distress syndrome: considerations of blood/
brain bilirubin transfer equilibrium. Early Hum Dev. 1982;6:341355
50. Johnson L, Bhutani VK. Guidelines for the management of jaundice in
the term and near-term infant. Clin Perinatol. 1998;25:555574
51. Newman TB, Easterling MJ, Goldman ES, Stevenson DK. Laboratory
evaluation of jaundice in newborns: frequency, cost, and yield. Am J Dis
Child. 1990;144:364368
52. Bratlid D. Bilirubin toxicity: pathophysiology and assessment of risk
factors. N Y State J Med 1991;91:489492
53. Bhutani VK, Johnson L, Gourley G, Dworanczyk R, Grous M. Non-
invasive measurement of total serum bilirubin by multi-wavelength
spectral reflectance by bili check in newborn patients. Pediatr Res. 1998;
43:167. Abstract
54. Gourley GR. Bilirubin metabolism and kernicterus. Adv Pediatr. 1997;
44:173229
14 PREDICTIVE ABILITY OF A PREDISCHARGE HOUR-SPECIFIC SERUM BILIRUBIN
by on March 21, 2006 www.pediatrics.org Downloaded from
DOI:10.1542/peds.103.1.6
1999;103;6-14 Pediatrics
Vinod K. Bhutani, Lois Johnson and Emidio M. Sivieri
Newborns
Subsequent Significant Hyperbilirubinemia in Healthy Term and Near-term
Predictive Ability of a Predischarge Hour-specific Serum Bilirubin for
This information is current as of March 21, 2006
& Services
Updated Information
http://www.pediatrics.org/cgi/content/full/103/1/6
including high-resolution figures, can be found at:
References
http://www.pediatrics.org/cgi/content/full/103/1/6#BIBL
at:
This article cites 51 articles, 19 of which you can access for free
Citations
http://www.pediatrics.org/cgi/content/full/103/1/6#otherarticles
This article has been cited by 45 HighWire-hosted articles:
Subspecialty Collections
http://www.pediatrics.org/cgi/collection/premature_and_newborn
Premature & Newborn
following collection(s):
This article, along with others on similar topics, appears in the
Permissions & Licensing
http://www.pediatrics.org/misc/Permissions.shtml
tables) or in its entirety can be found online at:
Information about reproducing this article in parts (figures,
Reprints
http://www.pediatrics.org/misc/reprints.shtml
Information about ordering reprints can be found online:
by on March 21, 2006 www.pediatrics.org Downloaded from