ArticlePDF AvailableLiterature Review

Diagnosis of Neonatal Sepsis: A Clinical and Laboratory Challenge



The evaluation of tests for neonatal sepsis is important because the infection may present a very serious threat to the baby. There is an urgent need to know whether the baby has sepsis to institute treatment as quickly as possible. Confirmation of the diagnosis may take time, and diagnostic tests are used to obtain a rapid indication of the infection status. These tests are not perfect. Some real cases of infection will produce negative test results, whereas some babies without infection will test positive. The potential usefulness of the test will depend, above all, on the clinical condition of the baby. If the baby is really very sick, the test will not give very much additional information. Similarly, if the baby is evidently well, a clinical examination will be sufficient and a positive test result would not dramatically increase the probability that the baby is infected. It is in situations in which the clinical picture leaves the physician in doubt about the infection status that a diagnostic test is likely to be most useful. Thus, the result of a diagnostic test must be evaluated in the light of the clinical condition of the baby. Extensive literature exists on single laboratory tests or combinations of tests, as well as tests used together with risk factors and/or clinical signs, to diagnose neonatal sepsis. In many instances, the results of the evaluations have been conflicting. There are several possible explanations for the divergent results, and the purpose of this review is to update readers on the topic and raise issues that should be addressed in the future. Throughout pregnancy and until the membranes rupture, the fetus is relatively protected from the microbial flora of the mother by the chorioamniotic membranes, the placenta, and poorly understood antibacterial factors in amniotic fluid (1). However, there …
Diagnosis of Neonatal Sepsis: A Clinical and Laboratory Challenge
The evaluation of tests for neonatal sepsis is important
because the infection may present a very serious threat to
the baby. There is an urgent need to know whether the
baby has sepsis to institute treatment as quickly as possi-
ble. Confirmation of the diagnosis may take time, and
diagnostic tests are used to obtain a rapid indication of the
infection status. These tests are not perfect. Some real
cases of infection will produce negative test results,
whereas some babies without infection will test positive.
The potential usefulness of the test will depend, above all,
on the clinical condition of the baby. If the baby is really
very sick, the test will not give very much additional
information. Similarly, if the baby is evidently well, a
clinical examination will be sufficient and a positive test
result would not dramatically increase the probability
that the baby is infected. It is in situations in which the
clinical picture leaves the physician in doubt about the
infection status that a diagnostic test is likely to be most
useful. Thus, the result of a diagnostic test must be
evaluated in the light of the clinical condition of the baby.
Extensive literature exists on single laboratory tests or
combinations of tests, as well as tests used together with
risk factors and/or clinical signs, to diagnose neonatal
sepsis. In many instances, the results of the evaluations
have been conflicting. There are several possible explana-
tions for the divergent results, and the purpose of this
review is to update readers on the topic and raise issues
that should be addressed in the future.
Basic Physiology of Neonatal Infection
Throughout pregnancy and until the membranes rupture,
the fetus is relatively protected from the microbial flora of
the mother by the chorioamniotic membranes, the pla-
centa, and poorly understood antibacterial factors in
amniotic fluid (1). However, there are many ways that
infectious agents can reach the fetus or newborn to cause
infection. Procedures disturbing the integrity of the uter-
ine contents, such as amniocentesis (2 ), cervical cerclage
(3), transcervical chorionic villus sampling (4 ), or percu-
taneous blood sampling (2, 5), can permit entry of skin or
vaginal organisms, causing amnionitis and secondary
fetal infection. Certain bacteria, particularly Treponema
pallidum and Listeria monocytogenes, can reach the fetus
through the maternal bloodstream despite placental pro-
tective mechanisms, causing transplacental infection. This
process is uncommon, but it leads to either congenital
infection not unlike infections caused by certain viruses or
Toxoplasma or to stillbirth resulting from overwhelming
Initial colonization of the neonate usually takes place
after rupture of the maternal membranes (1 ). In most
cases, the infant is colonized with the microflora of the
birth canal during delivery. However, particularly if the
rupture of membranes lasts longer than 24 h, vaginal
bacteria may ascend and in some cases produce inflam-
mation of the fetal membranes, umbilical cord, and pla-
centa (6, 7). Fetal infection can result from aspiration of
infected amniotic fluid (8 ), leading to stillbirth, premature
delivery, or neonatal sepsis (2, 7, 9, 10). The organisms
most commonly isolated from infected amniotic fluid are
anaerobic bacteria, group B streptococci, Escherichia coli,
and genital mycoplasmas (2, 9). Infection of the mother at
the time of birth, particularly genital infection, is the
principal pathway of maternal transmission (1, 11) and
can play an important role in the development of infection
in the neonate. Transplacental hematogenous infection
during or shortly before delivery (including the period of
separation of the placenta) is possible, although it seems
more likely that the infant is infected during passage
through the birth canal. Finally, bacteria can be intro-
duced after birth from the environment surrounding the
baby, either in the nursery or at home.
Many pre- and intrapartum obstetric complications are
associated with an increased risk of infection in newborn
infants. Among these are premature onset of labor, pro-
longed rupture of the fetal membranes, uterine inertia
with high forceps extraction, and maternal pyrexia (12 ).
Sophisticated equipment for respiratory and nutritional
support combined with invasive techniques provide life
support to the ill infant. Arterial and venous umbilical
catheters, central venous catheters, peripheral arterial and
venous cannulas, urinary indwelling catheters, hyperali-
mentation infusions, and assisted ventilation provide
enormous opportunities for relatively nonvirulent patho-
gens to establish infection and to invade the host (12 ).
Transient bacteremia may accompany procedures that
traumatize the skin and mucosal membranes. Bacteremia
was identified in infants who received endotracheal suc-
tioning; the bacteremia was present immediately after the
procedure, but culture results were negative at 10 min
(13). Invasion of the bloodstream may follow multiplica-
tion of the organisms in the upper respiratory tract or
other foci. Once bacteria gain access to the blood stream,
mechanisms are activated by the host to eliminate the
microbial intruder. Usually the organism is efficiently
cleared by the monocyte-macrophage system after opso-
nization by antibody and complement. Thus, the bactere-
mia produces only short-lived illness. Sometimes, how-
ever, depending on the age of the patient, the virulence
and number of bacteria in the blood, the nutritional and
immunologic status of the host, and the timing and nature
of therapeutic intervention, a systemic inflammatory re-
sponse is established that can progress independently of
the original infection (14, 15).
Perinatally vs Postnatally Acquired Infections
Infections that are manifest early in the first week of life
are usually attributable to microorganisms transmitted
from mother to infant and have an epidemiology different
from those of infections acquired later in the neonatal
period (16 ). There is no time point that clearly distin-
guishes maternally transmitted infections from environ-
Clinical Chemistry 50, No. 2, 2004 279
mentally transmitted infections (17). However, epidemi-
ologic studies of neonatal infections usually divide so-
called early-onset infections from late-onset infections at
somewhere between days 3 and 7 of life with the assump-
tion that early-onset infections are presumably transmit-
ted perinatally from the mother and late-onset infections
are acquired postnatally from an environmental source
(16–26). Therefore, a report of a new test for the diagnosis
of neonatal sepsis must be critically evaluated taking into
account the arbitrary time points used for separating
maternally derived infections from postnatal infections.
The postnatal age provided as a basis for diagnosing
either early- or late-onset disease may have a striking
effect on the concentrations of analytes in cases. This is
equally important for controls. Only when the time points
for separating the two patterns of disease are uniform can
consistency in all studies be expected. We are far from
this. In the last decade, many studies have not even
differentiated between early- and late-onset infection, and
still more disappointing, they have examined the accu-
racy of modern laboratory tests in groups of newborn
infants with wide-ranging postnatal ages (27–29). These
differences may alter the diagnostic characteristics of a
particular test and confound the comparison of published
reports (30).
Systemic Response to Infection in Newborns
Neonatal sepsis, sepsis neonatorum, and neonatal septi-
cemia are terms that have been used to describe the
systemic response to infection in newborn infants. There
is little agreement on the proper use of the terms, i.e.,
whether their use should be restricted to bacterial infec-
tions, positive blood cultures, or severity of illness (31 ).
In 1991, the American College of Chest Physicians and
the Society of Critical Care Medicine convened a Consen-
sus Conference in an attempt to provide a conceptual and
practical framework to define the systemic inflammatory
response to infection, which is a progressive injurious
process that falls under the generalized term “sepsis” and
includes sepsis-associated organ dysfunction as well (32).
The term systemic inflammatory response syndrome
(SIRS) is used to describe a clinical syndrome character-
ized by two or more of the following: (a) fever or
hypothermia, (b) tachycardia, (c) tachypnea or hyperven-
tilation, and (d) abnormal white blood cells or increase in
immature forms. SIRS may be a result of a variety of
immunologic, endocrinologic, traumatic, surgical, chemo-
therapeutic, and infectious insults (32, 33 ). Sepsis is con-
sidered when there is a systemic response to a possible
infection. Evidence of bacteremia or an infectious focus is
not required (32, 33 ). When sepsis is accompanied by
organ dysfunction, hypoperfusion, or hypotension, the
sepsis is considered severe. Septic shock ensues when
hypotension develops despite adequate fluid replace-
ment. Finally, in the presence of altered organ function in
an acutely ill patient, so severe that homeostasis cannot be
maintained without intervention, multiple-organ dys-
function syndrome is diagnosed. Very recently, a group of
experts and opinion leaders revised the 1991 sepsis guide-
lines and found that apart from expanding the list of signs
and symptoms of sepsis to reflect clinical bedside experi-
ence, no evidence exists to support a change to the
definitions (34).
The application of the above terminology guidelines to
septic newborns, however, needs careful assessment (i.e.,
age-related reference values for blood pressure, heart rate,
respiratory rate, and leukocyte count). Furthermore, the
application of a staging system (including sepsis, severe
sepsis, septic shock, and multiple-organ dysfunction syn-
drome) may not be the best approach to disease or risk
stratification in the newborn. Early organ abnormalities
may not be manifest, so that earlier stages in the evolution
of the syndrome may not be identified. Furthermore, the
often fulminant or rapid course of the disease in the
newborn may limit the staging system outlined above to
just a snapshot in time of this dynamic process.
Currently, criteria for neonatal sepsis usually include
documentation of infection in a newborn infant with a
serious systemic illness in which noninfectious explana-
tions for the abnormal pathophysiologic state are ex-
cluded or unlikely. However, even culture is not free from
error because it can be falsely sterile, as suggested by
postmortem cultures (35 ), or because of the low yield
caused by insufficient sample volumes, intermittent or
low-density bacteremia, or suppression of bacterial
growth by earlier (i.e., intrapartum) antibiotic administra-
tion. Theoretically, this would lead to an underrepresen-
tation of truly infected newborn infants. On the other
hand, in only a few studies were the definitions of sepsis
stringent enough to distinguish culture contaminants
from true systemic infections. A decade ago, Pourcyrous
et al. (36) recognized that 83% of blood cultures yielding
organisms with low-grade or questionable virulence were
the result of contamination during collection. In all of
these cases, antibiotic therapy was not administered or
was inadequate by generally accepted standards, but
clinical courses were uneventful. If the term “culture-
proven” sepsis is used to mean infants whose clinical
signs of infection and/or abnormal laboratory results are
fully explained by the yield of typical skin or upper
respiratory flora from a single blood culture, it is neces-
sary to specify whether, in this situation, clinical signs and
abnormal laboratory results have been resolved with
specific antimicrobial therapy or worsened without it
(37). This approach, taking full account of the clinical
course, should be more fruitful, leading to an unequivocal
standard criterion.
Faced with these questions, and as found in studies on
sepsis in adult patients, part of the neonatology literature
has abandoned positive cultures as a critical point of the
“gold standard”. However, it must be emphasized that
there is no universal agreement on the definition of
neonatal clinical septicemia. In recent years, some authors
have suggested that presence of just one clinical sign
(compatible with infection) along with a C-reactive pro-
tein (CRP) value 10 mg/L, is sufficient to make the
diagnosis of early- and late-onset neonatal clinical septi-
cemia (38, 39 ). Others have assigned infants with only
280 Chiesa et al.: Diagnosis of Neonatal Sepsis
maternal risk factor(s) for infection to the sepsis group to
develop criteria that are relevant and simple enough for
clinical practice (40). In other studies, the presence of
two or three categories (by organ system) of clinical signs
of infection in the infant has been taken to strongly
support a diagnosis of septicemia (41 43 ). Absent from
neonatology publications, however, are data on just how
common a given clinical sign is in all infants ever evalu-
ated (rather than in infants with positive cultures or
infants with a specific type of infection). Thus, with the
exception of the clinical scenario of a newborn with
clear-cut signs of infection such as septic shock, the
possibility that infants with only clinical evidence of
infection may have been assigned an incorrect diagnosis is
intrinsic to all studies of this nature (37 ). In contrast to our
insistence on the need for an unequivocal and uniform
standard criterion for establishing culture-positive as well
as culture-negative sepsis, some authors have considered
a CRP value 10 mg/L combined with an immature:total
neutrophil ratio 0.25 as a criterion to start antibiotic
therapy even in babies with no symptoms of infection
In conclusion, it would seem that the absence of a
universally accepted standard definition has led to au-
thors creating their own definitions to suit the purposes of
their own particular studies. However, these definitions
should be explicit with regard to type of baby (defined by
gestational age and birth weight), babys age, culture
status, symptom status, and illness severity.
Illness Severity in Newborns
Although illness severity is a familiar medical concept, it
is sometimes difficult to assess. An important feature of
morbidity in newborn infants is its integrative and global
nature. Individual attributes (e.g., hypoglycemia, oral
feeding difficulties, and seizures) fail to capture the over-
all neonatal health and morbidity status. Until recently,
birth weight, gestational age, and Apgar score were often
considered sufficient proxy measures of morbidity at
birth. Although birth weight and gestational age stratifi-
cation may be adequate for some purposes, they do not
account for variations in illness severity completely. The
Apgar score was originally designed to identify neonates
in need of immediate cardiopulmonary intervention (45
48) and is determined predominantly by acute intra-
partum events. Low Apgar scores may actually reflect
gestational age, birth weight, sedation, or congenital mal-
formations (47).
In recent years, two major neonatal severity measures,
the Score for Neonatal Acute Physiology (SNAP) and the
Clinical Risk Index for Babies (CRIB), have been devel-
oped. Both SNAP and CRIB focus on measuring and
scoring physiologic derangements, following the example
of the Acute Physiology and Chronic Health Evaluation in
adult intensive care units (ICUs) (49) and the Pediatric
Risk of Mortality in pediatric ICUs (50 ). The rationale is
that, regardless of disease or diagnosis, derangements
from the physiologic norm increase the likelihood of
adverse outcome and that the greater the derangements,
the greater the risk. The composite severity can be repre-
sented by the weighted sum of derangements across all
organ systems (51).
CRIB was designed for ease of data collection. It uses a
12-h baseline period from birth (52). In the population
31 weeks of gestational age, most illness is adequately
captured by sampling only three items in the respiratory/
metabolic systems (worst base deficit and highest and
lowest appropriate oxygen requirements). SNAP, first
described in 1993 (53 ), uses the worst recorded values of
26 routinely measured physiologic variables (including 8
that can be scored for high or low values) during the first
24 h of stay in the neonatal ICU (NICU). It has been
prospectively validated in multiple NICUs covering very
different population groups and has been found to be
highly correlated with other indicators of illness severity
(5356). From this, the SNAP-Perinatal Extension (SNAP-
PE) captures SNAP physiology scores, combining them
with additional scoring for three potent perinatal mortal-
ity risks (i.e., birth weight, low Apgar score, and small for
gestational age status), all of which are independent of
physiologic derangement (57 ). Thus the SNAP-PE score is
a combined physiologic and perinatal measure of mortal-
ity risk.
As a first-generation newborn illness severity score,
SNAP is suitable for outcomes research, but it is cumber-
some for routine clinical use because of the number and
complexity of its items. On the other hand, CRIB is
difficult to apply to infants born outside the hospital. The
recently revised SNAP-II scores six items, covering six
systems (pH, temperature, blood pressure, oxygenation,
urine output, presence of multiple seizures), but has
performance equivalent to that of SNAP and CRIB for
both very low birth-weight infants and larger infants (58).
Laboratory Tests
normal standard: general considerations
A normal standard, in the context of a laboratory test,
refers to what is commonly called a normal range by
statisticians. It should not, however, be confused with the
normal (gaussian) distribution. The normal distribution is
a symmetric distribution in which the interval defined by
the mean 1.96 SD includes the central 95% of the values.
The distributions of most clinical indices do not approxi-
mate the normal distribution, and this simple interval
cannot be used to define an interval that includes the
central 95% of the values of a clinical index. A normal
range, or normal standard, is an interval that refers to the
distribution of the index in healthy individuals; the nor-
mal range includes the central 95% of these values. In
statistical terms, the normal range can be defined as the
interval between the 2.5th percentile and the 97.5th per-
centile of the distribution of the index in healthy individ-
uals. If this interval is used in the context of a diagnostic
test, it has, by definition, a specificity of 95% (the speci-
ficity is 97.5% only if one tail of the distribution is
considered a positive test result). The normal range pro-
vides no information about the sensitivity of the test.
Clinical Chemistry 50, No. 2, 2004 281
normal (and abnormal) ranges in the
newborn infant
At birth, the fetus makes an abrupt transition from the
protective environment of the uterus to the outside world;
the newly born baby must undergo extreme physiologic
changes to survive this transition (59 ). It is therefore not
surprising that many physiologic and metabolic processes
change constantly during the first few days of life. These
changes profoundly affect various kinds of laboratory
values (e.g., hormones, biochemical indices, immunologic
products, and cytokines), and the mean reference values
in the early neonatal period differ from those measured in
later periods (60, 61). Analysis of the reliability of labora-
tory tests in the diagnosis of neonatal sepsis must there-
fore take account of the fact that postnatal age may
dramatically affect the interpretation of what constitutes
the normal (and equally important the abnormal) value of
a laboratory test (60).
Virtually all published guidelines stress the need for
reference values. Unfortunately, the agreement ends here.
In this context, no discussion of the issue would be
complete without consideration of the interpretation of
two laboratory aids whose performance has acquired an
almost ritual quality: CRP and complete blood cell count
In the majority of published reports, upper limits for
CRP during the neonatal period have been obtained from
symptomatic uninfected patients (6272 ). Thus there are
few studies of upper limits for CRP in the healthy
newborn (7376 ). Furthermore, most of these studies of
healthy neonates were cross-sectional and based on small
samples with wide-ranging postnatal ages. Gutteberg et
al. (73), who determined (by radial immunodiffusion)
CRP concentrations in 16 apparently healthy newborns at
unspecified sampling times during the first month of life,
obtained normal 97.5th percentile upper limits for CRP of
5 mg/L. In the study by Forest et al. (74 ), 69 newborns
labeled as apparently healthy newborns for their nor-
mal postnatal course, despite their initial admission to the
NICU, were sampled at unspecified times from birth up
to the 6th week of life. Of these, 68 had a CRP concentra-
tion (as measured by enzyme immunoassay) 10 mg/L.
In the study by Schouten-Van Meeteren et al. (75 ), 95% of
38 apparently healthy newborns who were sampled be-
tween 12 and 24 h after birth had CRP concentrations 10
mg/L (as determined by a fluorescence polarization im-
munoassay). Nonetheless, it is less well known that the
false-positive rate for these methods in the apparently
healthy neonate throughout the first 30 days of life is 8%
for CRP (30). In addition, a lack of follow-up (or outcome)
data has been a notable weakness in the literature on the
squeaky clean well infants.
We have followed these precepts in a recent longitudi-
nal study investigating the pattern of CRP response in the
healthy neonate at three fixed neonatal ages: 0, 24, and
48 h (76). We were not interested in establishing CRP
ranges for an ideal population of healthy infants (i.e., term
infants with no risk factors), but in setting up ranges that
may be important for clinicianseveryday experience. The
CRP reference intervals that we established at birth (95th
percentile, 5.0 mg/L), at 24 h (95th percentile, 14.0 mg/L),
and at 48 h of life (95th percentile, 9.7 mg/L) included
neonates who were not necessarily free of history of
maternal and intrapartum complications but whose post-
natal clinical course from birth to the 4-week follow-up
visit was unremarkable, implying therefore, no need of
management (including antimicrobial treatment) through-
out this study period.
The most commonly cited study of neonatal CBC is that
of Manroe et al. (77 ). In the period from 0 to 24 h of age,
the critical decision time for most neonatal sepsis work-
ups(78 ), Manroe et al. based their graphs on 108 infants
sampled in a cross-sectional way. The 10th and 90th
percentile envelopes were defined by visual inspection. It
is not comforting to think that so many of us have long
accepted these norms without question. From the study
by Schelonka et al. (78), the normal ranges for leukocyte
indexes in 193 healthy term infants with no identifiable
perinatal risk factors for infection were at4hoflife
considerably broader than those described previously by
Manroe et al. (77) for the first 24 h of life. Thus, if one
applies the reference intervals of Manroe et al. (77 ) to
healthy term infants, one would label huge numbers of
them as being at extremely high risk for sepsis. Leaving
aside the methodologic aspects, it is important to remem-
ber that the actual physical sampling can lead to dramatic
changes in the CBC results. It has been well documented
that the CBC depends on the infants age (77 80 ),on
whether the sample is arterial or venous (81), and on
whether the infant is crying vigorously (81 ). This means
that for a given infant, a test value is not a static value;
there is considerable intraindividual variability. In this
context, it is worrying to think that so many of us have
accepted a very popular index, the immature:total neu-
trophil ratio, without question, although the test is subject
to considerable interobserver variation. Because seg-
mented and band neutrophils exist on a continuum of
cellular maturation, the use of discrete boundaries is
artificial and subject to observer bias. In 1993, the College
of American Pathologists surveyed 6600 hematology tech-
nicians in a band neutrophil identification exercise (82 ).
Because of poor reproducibility of band neutrophil iden-
tification in this large sample, the College no longer tests
laboratory proficiency in the differentiation of segmented
and band neutrophils (82).
In view of this dynamic behavior, do we also need
age-specific cutoff values for differentiating symptomatic
newborns with infection from those with no infection? In
1980, Philip and Hewitt (83 ) suggested that, for simplic-
ity, the same laboratory cutoffs for markers used in
diagnosing or ruling out neonatal sepsis should be ap-
plied during the first postnatal week. Two decades later,
there are even updated systematic reviews on the accu-
racy of modern laboratory tests for the diagnosis of sepsis
in the newborn that support this contention (29). Perhaps
the most surprising thing is that such methodologic flaws
have also plagued the very latest applications of CRP.
Philip and Mills (84 ) recommended that at any neonatal
282 Chiesa et al.: Diagnosis of Neonatal Sepsis
age a CRP value 10 mg/L in the presence of one (or
more) clinical sign(s) or one (or more) risk factor(s) for
infection should be the clinical pathway for transferring a
neonate from the well-baby nursery to the NICU and
starting antimicrobial therapy. Franz et al. (38, 39 ),onthe
other hand, considered a conventional CRP value 10
mg/L, in the presence of one (or more) clinical sign(s)
compatible with infection, as a criterion to make a diag-
nosis of clinical septicemia at any neonatal age in NICU
babies. Against this background, we have recently shown
(85) that failure to recognize specific cutoff values by a
given test for each time point of evaluation over the first
48 h of life may confound the interpretation of what
constitutes a true negative and a true positive value
in the diagnosis of neonatal infection.
role of new markers
In recent years, the search for diagnostic tests for sepsis in
newborn infants has turned to cytokines as well as to
other substances associated with the inflammatory re-
sponse, in some cases induced by cytokines, as possible
indicators of infection. A few new markers remain prom-
ising, of which interleukin (IL)-6 is the most intensively
studied. In addition, procalcitonin (PCT) appears to show
considerable promise as a diagnostic test for neonatal
Data pertaining to reference intervals for the new
markers are very limited. We have recently shown that
both IL-6 and PCT present a natural fluctuation in the
immediate postnatal period (76, 86), necessitating very
careful adjustments in the normal ranges. This may ex-
plain the conflicting cutoff points for abnormal values that
have been reported for these two markers (28, 38, 8792).
We have also shown that some confounding factors, per
se, should be taken into account to define physiologic
concentrations of both IL-6 and PCT (76, 86). Of note, the
kinetics of IL-6 during the first 48 h of life in healthy
infants are different in the near-term infant compared
with kinetics in the term neonate, suggesting a gestational
age-dependent effect on IL-6 values over the first 48 h of
life (76).
Reports in the literature about the usefulness of new
indicators of infections have been conflicting. In a recent
study we (85 ), like others (28, 88 ), found that the sensi-
tivity of IL-6 as well as that of PCT is low, in the range of
7080% at birth, by far the most critical decision point
when evaluating a newborn to rule out sepsis. In the
immediate postnatal period, however, although the sen-
sitivity of IL-6 decreases over time, the sensitivity of PCT
(as well as that of CRP) increases, making serial measure-
ments useful for those situations in which one needs to
decide how long to treat. Because the sensitivity of these
markers varies over time, their use requires specific cutoff
values for each time point of evaluation over the first 48 h
of life. We also showed that IL-6 and PCT concentrations
during the early neonatal period may relate differentially
to clinical complications in the perinatal period (85 ).
effect of illness severity and risk status on
The reliability of most laboratory tests for the differential
diagnosis of infectious vs noninfectious systemic inflam-
matory response has been assessed in highly diverse
groups of ill neonates with a mixture of diagnoses and
conditions and has yielded discrepant results (30). In this
situation, because infectious as well as noninfectious
diagnoses and the conditions themselves may vary in
severity, some of the variation among published reports
might reflect differences in baseline severity and risk
status, independently of the presence of infection. There is
wide variation in clinical severity among NICUs: admis-
sion ranging from critically ill infants with multiple-
organ-system failure to mildly ill term infants with tran-
sient problems related to the birth process or healthy
premature infants who require technologic support until
mature. Unfortunately, a major problem with the litera-
ture on the diagnostic value of laboratory aids for diag-
nosis of neonatal infection is its failure to evaluate the
effect of overall illness severity on a given marker.
We therefore recently evaluated the influence of illness
severity on CRP, IL-6, and PCT for the diagnosis of sepsis
during the first 48 h of life in critically ill newborns
admitted to the NICU (85 ). To this end, we used both the
SNAP and SNAP-PE scores. We found that CRP, IL-6, and
PCT increase in the presence of bacterial infection and
that their increases are independent of illness severity.
However, we also found that illness severity has the
potential to confound IL-6 concentrations in that, among
babies without infection, the higher the illness severity,
the higher the IL-6 concentration after birth. It is clear
from the above that the diagnostic value of certain labo-
ratory aids, such as IL-6, may be altered by physiologic
severity and risk indexes.
Methodologic Issues
general considerations
The value of a clinical index observed in a patient may be
a useful aid for diagnosis, if the distribution of values
among healthy individuals is substantially different from
the distribution in true cases of the disease. At one
extreme, if the two distributions are identical, clearly the
value observed for a single patient can provide no infor-
mation about the diagnosis, and the test is useless. At the
other extreme, if the two distributions do not overlap, a
cutoff value between the two distributions can provide a
certain diagnosis.
The most common situation is that the two distribu-
tions overlap and the test cannot give a sure indication of
the disease state of the patient. In fact, a single value may
be used as a cutoff between a positive and a negative test
result, and the quantitative value of the index is reduced
to a dichotomy. The test is usually evaluated by calculat-
ing its sensitivity and specificity. The sensitivity of the test
is the probability that a patient with the disease or
condition is identified as positive by the test. The speci-
ficity is the probability that a healthy individual has a
negative test result. These two probabilities may be esti-
Clinical Chemistry 50, No. 2, 2004 283
mated by applying the test to a sample of known cases
and another sample of known healthy individuals. How-
ever, the sensitivity and specificity are not of direct
interest to clinicians when they are faced with the clinical
problem of making a diagnosis for an individual patient.
What clinicians would like to know is whether their
patients have the disease, but unfortunately they cannot
be certain about this. This uncertainty is also measured by
a probability. Epidemiologic studies can be used to esti-
mate the prevalence of the disease in a particular clinical
setting. This is the probability that the individual has the
disease before the diagnostic test is done, and is often
called the prior probability of the disease. Suppose that
the test is applied to a patient and the test result is
positive. The clinician should now believe, even more
certainly, that the patient has the disease. This degree of
belief is measured by a probability called the predictive
value of a positive test, which can be estimated from the
sensitivity, specificity, and prevalence of the disease in the
clinical setting.
This is a gross simplification of the reasoning used by a
clinician faced with the problem of making a diagnosis for
a single patient, however. For example, suppose that the
test has been evaluated, and it is found that the predictive
value of the positive test is 70%. That is, a person with a
positive test has a 70% probability of having the disease.
What can this tell a clinician, whose patient has a positive
test result, about the disease status? It certainly does not
mean that the clinician believes that the patient has a 70%
chance of having the disease. This value, 70%, is based on
information obtained from a sample of patients of this
type. Much more information is available for the evalu-
ation of an individual patient, and this will have been
collected by the physician in the case history. The positive
test result will be just one additional piece of information.
If the patient has important signs and symptoms of the
disease before the test, his or her prior probability will be
high, and a positive test result may almost confirm the
diagnosis. On the other hand, if the patient shows few or
no signs of the disease, a positive test will not imply a
high probability that the patient has the disease. A diag-
nostic test will potentially be most useful for patients
whose conditions and case histories suggest that they may
have the disease.
On the basis of a review of diagnostic tests of bacterial
infections in infancy (27), it has been concluded that these
diagnostic tests may be of limited use in the absence of
other relevant clinical information about a patient. In-
deed, the authors of the review suggest that when a
clinician is faced with an infant with a possible serious
infection, treatment may be started with antibiotics imme-
diately. Clearly the test result will not affect the decision
to start treatment if there is already strong evidence of the
infection. Similarly, in the absence of indications of infec-
tion, the clinician may not start treatment even if the test
were positive. The test is most likely to be useful when the
case history and the condition of the patient leave the
clinician in serious doubt about the presence of infection,
in which case the diagnostic test may be used as a decision
rule: start treatment if the test is positive; delay treatment
if the test result is negative.
accuracy of laboratory methods
Given the increased mobility of patients, comparable
(true) test results are essential for a rational and cost-
effective diagnostic approach in laboratory medicine (93).
From the standpoint of laboratory practice, however,
when examining all of the data published to date on the
laboratory aids for diagnosis of neonatal sepsis, it is
obvious that differences in laboratory techniques have
also been in part responsible for the conflicting opinions
about the reliability of a given test during the neonatal
period. Again, no discussion of the issue would be com-
plete without consideration of the different methods used
to measure CRP, the most commonly used marker for
identifying neonates with sepsis.
The original test was a simple precipitin test, usually in
a microcapillary tube, in which the height of the precipi-
tant defined the amount of CRP present. A comparison of
reactions obtained by different investigators using the
capillary tube method revealed widely disparate results,
depending on the sensitivity of the commercial antiserum
used in the assay (94). Not until the early 1980s did rapid
and reliable quantitative immunoassays become commer-
cially available in which monoclonal CRP-specific anti-
body was used. Some investigators used immunoassays
that permitted direct visualization of a CRPantibody
complex through particle agglutination (i.e., latex agglu-
tination) or through precipitation (i.e., radial immunodif-
fusion, immunoturbidimetry, nephelometry). Other in-
vestigators used immunoassays that enlisted a marker for
detection (i.e., RIA, enzyme-multiplied immunoassay
technique). Although the slide agglutination test is rapid
and convenient and still used by some investigators (95),
it is only semiquantitative and subject to reagent variabil-
ity (96).
Studies comparing fully automated turbidimetric and
nephelometric methods for CRP with older assays have
shown superior precision, with far greater speed, sensi-
tivity, and reproducibility. However, these assay methods
still have limited sensitivity, and until recently, CRP
concentrations below 10 mg/L could not be measured
precisely, leading to widespread adoption of this value, or
even higher, as the upper limit of the healthy reference
interval. This is satisfactory for some purposes in general
medicine because the marked acute-phase responses that
characterize bacterial infections, ischemic necrosis of tis-
sue, and most active inflammatory conditions usually
lead to much higher CRP values. However, for neonates,
health-associated reference values are below conventional
threshold values. Furthermore, newborns may be unable
to produce high amounts of acute-phase proteins and
respond to infection with a smaller increase in CRP than
The clinical need for highly sensitive CRP assays was
first recognized in neonatal pediatric practice (71 ), and
their recent development holds promise for a further
increase in the diagnostic accuracy of neonatal infection.
284 Chiesa et al.: Diagnosis of Neonatal Sepsis
However, before introduction of the new tests into routine
neonatal practice, the CRP reference intervals to be estab-
lished by the newly developed assays need to be com-
pared with the traditional ones. In this context, further
standardization efforts are required to ensure that high-
sensitivity CRP assays have the requisite precision at low
CRP concentrations (97 ), particularly at birth when CRP
upper reference limits fall below the conventional thresh-
old values of 35 mg/L. On the other hand, another issue
that merits discussion is the possibility of underestimat-
ing with the high-sensitivity assays the true CRP concen-
tration because of a prozone effect (97). Thus, before
introduction of the high-sensitivity CRP assay as routine
indicator of more severe inflammation, further investiga-
tion is required to improve agreement in the higher range
among the different methods.
Further potential caveats arise from the type of assay
used to quantify the biomarker. For example, use of
different assay types and different antibodies in immuno-
assays for tumor necrosis factor-
) may be re-
sponsible for the discrepancies among TNF-
values (98 ).
The various immunoassays available for TNF-
free (bioactive) or bound plus free (total) TNF-
. Thus,
immunoassays may detect not only bioactive TNF-
, but
also inactive fragments, monomers, and polymers; bind-
ing of TNF-
to its soluble receptors (p55 or p75) may also
interfere with its determination by immunoassay (98 ).
Other investigators have chosen to measure TNF-
bioassay (e.g., mouse fibrosarcoma cell line WEHI 164)
because it may reflect the activity of cytokine (i.e., the
amount of bioactive TNF-
) in vivo better than the
immunoassay (99).
Another crucial point is that methods must be repro-
ducible in time and space, with a turnover time sufficient
to meet the demands of clinical practice. Some methods
have automated dilution, whereas others require a man-
ual dilution prepared off-line. In the majority of studies
on the diagnostic value of cytokines in identifying or
excluding the septic neonate, a manual immunoassay was
used (41, 100, 101). This method usually requires 2 4h.
Despite being faster, the manual method for cytokine
determination is not clinically practical because techni-
cians competent in assaying cytokines with ELISAs would
need to be available 24 h a day. Even if this were possible,
interobserver error would be a problem. The development
of readily available automatic assays, with fast turn-
around times, is essential for potential widespread clinical
application of cytokines.
One seldom discussed problem is that only in recent
reports are methods for calculating measures of diagnos-
tic accuracy or making comparisons accompanied by
statistical measures of their precision (i.e., 95% confidence
intervals) (40, 85). The issue is critical because of the
relatively small number of patients in each report, with
consequent possibly wide confidence intervals.
Can we learn anything from certain potential method-
ologic problems alluded to above? It is clear from the
above discussion that the wide variations among studies
on the methods (and the results) preclude any meaningful
synthesis, such as metaanalysis, of all reported studies. As
recently suggested by Escobar (102), it is time that we
begin to debate the methods we use to measure test
performance, rather than just how a given test performs.
This issue has been the subject of a recent report on
Standards for Reporting of Diagnostic Accuracy (STARD)
(103). An important part of the STARD plan is to evaluate
its effect on the reporting of studies. Whether the STARD
report is a step in the right direction toward complete and
accurate reporting of studies of diagnostic accuracy in
newborn infants presents a challenging new research
1. Klein JO. Bacterial sepsis and meningitis. In: Remington JS, Klein JO, eds.
Infectious diseases of the fetus and newborn infant. Philadelphia: WB
Saunders, 2001:943–98.
2. Gibbs RS, Duff P. Progress in pathogenesis and management of clinical
intraamniotic infection. Am J Obstet Gynecol 1991;164:1317–26.
3. Charles D, Edwards WR. Infectious complications of cervical cerclage. Am J
Obstet Gynecol 1981;141:1065–71.
4. Fejgin M, Amiel A, Kaneti H, Ben-Nun I, Beyth Y. Fulminant sepsis due to
group B
-hemolytic streptococci following transcervical chorionic villi
sampling [Letter]. Clin Infect Dis 1993;17:142–3.
5. Wilkins I, Mezrow G, Lynch L, Bottone EJ, Berkowitz RL. Amnionitis and
life-threatening respiratory distress after percutaneous umbilical blood
sampling. Am J Obstet Gynecol 1989;160:427– 8.
6. Bernirschke K. Routes and types of infection in the fetus and newborn.
Am J Dis Child 1960;99:714 –21.
7. St. Geme JW Jr, Murray DL, Carter J, Hobel CJ, Leake RD, Anthon BF, et al.
Perinatal bacterial infection after prolonged rupture of amniotic mem-
branes: an analysis of risk and management. J Pediatr 1984;104:608–13.
8. Blanc WA. Pathways of fetal and early neonatal infection: viral placentitis,
bacterial and fungal chorioamnionitis. J Pediatr 1961;59:473–96.
9. Hillier SL, Krohn MA, Kiviat NB, Watts DH, Eschenback DA. Microbiologic
causes and neonatal outcomes associated with chorioamnion infection.
Am J Obstet Gynecol 1991;165:955– 61.
10. Yoder PR, Gibbs RS, Blanco JD, Castaneda YS, St. Clair PJ. A prospective,
controlled study of maternal and perinatal outcome after intra-amniotic
infection at term. Am J Obstet Gynecol 1983;145:695–701.
11. Prober CG. Clinical approach to the infected neonate. In: Long SS,
Pickering LK, Prober CG, eds. Principles and practice of pediatric infectious
diseases. New York: Churchill Livingstone, 1997:603– 6.
12. Sa´ez-Llorens X, McCracken GH Jr. Perinatal bacterial diseases. In: Feigin
RD, Cherry JD, eds. Textbook of pediatric infectious diseases. Philadelphia:
WB Saunders, 1998:892–926.
13. Freedman RM, Baltimore R. Fatal Streptococcus viridans septicemia and
meningitis: a relationship to fetal scalp electrode monitoring. J Perinatol
1990;10:272– 4.
14. Sa´ez-Llorens X, McCracken GH Jr. Sepsis syndrome and septic shock in
pediatrics: current concepts of terminology, pathophysiology, and manage-
ment. J Pediatr 1993;123:497–508.
15. Sa´ez-Llorens X, Lagrutta F. The acute-phase host reaction during bacterial
infections in children. Pediatr Infect Dis J 1993;12:83–7.
16. Baltimore RS. Neonatal nosocomial infections. Semin Perinatol 1998;22:
17. Harris JAS, Goldmann DA. Infections acquired in the nursery: epidemiology
and control. In: Remington JS, Klein JO, eds. Infectious diseases of the
fetus and newborn infant. Philadelphia: WB Saunders, 2001:1371– 418.
18. Stoll BJ, Gordon T, Korones SB, Shankaran S, Tyson JE, Bauer CR, et al.
Early-onset sepsis in very low birth weight neonates: a report from the
National Institute of Child Health and Human Development Neonatal
Research Network. J Pediatr 1996;129:72– 80.
19. Stoll BJ, Gordon T, Korones SB, Shankaran S, Tyson JE, Bauer CR, et al.
Late-onset sepsis in very low birth weight neonates: a report from the
National Institute of Child Health and Human Development Neonatal
Research Network. J Pediatr 1996;129:63–71.
20. Hauth JC, Merenstein GB. Infection control. In: Hauth JC, Merenstein GB,
eds. Guidelines for perinatal care. Elk Grove Village, IL: American Academy
of Pediatrics and American College of Obstetricians and Gynecologists,
21. Baker CJ. Group B streptococcal infections. Clin Perinatol 1997;24:59
22. Schrag SJ, Zywicki S, Farley MM, Reingold AL, Harrison LH, Lefkowitz LB, et
al. Group B streptococcal disease in the era of intrapartum antibiotic
prophylaxis. N Engl J Med 2000;342:15–20.
Clinical Chemistry 50, No. 2, 2004 285
23. Wendel GD Jr, Leveno KJ, Sa´nchez PJ, Jackson GL, McIntire DD, Siegel JD.
Prevention of neonatal group B streptococcal disease: a combined intra-
partum and neonatal protocol. Am J Obstet Gynecol 2002;186:618 26.
24. Schrag SJ, Zell ER, Lynfield R, Roome A, Arnold KE, Craig AS, et al. A
population-based comparison of strategies to prevent early-onset group B
streptococcal disease in neonates. N Engl J Med 2002;347:2339.
25. Towers CV, Briggs GG. Antepartum use of antibiotics and early-onset
neonatal sepsis: the next 4 years. Am J Obstet Gynecol 2002;187:495
26. Stoll BJ, Hansen N, Fanaroff AA, Wright LL, Carlo WA, Ehrenkranz RA, et al.
Changes in pathogens causing early-onset sepsis in very-low-birth-weight
infants. N Engl J Med 2002;347:240 7.
27. Fowlie PW, Schmidt B. Diagnostic tests for bacterial infection from birth to
90 daysa systematic review. Arch Dis Child Fetal Neonatal Ed 1998;78:
F92 8.
28. Mehr S, Doyle LW. Cytokines as markers of bacterial sepsis in newborn
infants: a review. Pediatr Infect Dis J 2000;19:879 87.
29. Malik A, Hui CPS, Pennie RA, Kirpalani H. Beyond the complete blood cell
count and C-reactive protein. A systematic review of modern diagnostic
tests for neonatal sepsis. Arch Pediatr Adolesc Med 2003;157:511 6.
30. Weinberg GA, Powel KR. Laboratory aids for diagnosis of neonatal sepsis.
In: Remington JS, Klein JO, eds. Infectious diseases of the fetus and
newborn infant. Philadelphia: WB Saunders, 2001:1327 44.
31. Gotoff SP. Infections of the neonatal infant. In: Behrman RE, Kliegman RM,
Jenson HB, eds. Nelson textbook of pediatrics. Philadelphia: WB Saun-
ders, 2000:538 52.
32. Members of the American College of Chest Physicians/Society of Critical
Care Medicine Consensus Conference Committee. American College of
Chest Physicians/Society of Critical Care Medicine Consensus Confer-
ence: definitions for sepsis and organ failure and guidelines for the use of
innovative therapies in sepsis. Crit Care Med 1992;20:864 74.
33. Angus DC, Wax RS. Epidemiology of sepsis: an update. Crit Care Med
2001;29(Suppl):S109 16.
34. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, et al. 2001
SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Confer-
ence. Crit Care Med 2003;31:1250 6.
35. Squire E, Favara B, Todd J. Diagnosis of neonatal bacterial infection:
hematologic and pathologic findings in fatal and nonfatal cases. Pediatrics
1979;64:60 4.
36. Pourcyrous M, Bada HS, Korones SB, Baselski V, Wong SP. Significance of
serial C-reactive protein responses in neonatal infection and other disor-
ders. Pediatrics 1993;92:4315.
37. Benitz WE, Han MY, Madan A, Ramachandra P. Serial serum C-reactive
protein levels in the diagnosis of neonatal infection. Pediatrics 1998;102:
38. Franz AR, Kron M, Pohlandt F, Steinbach G. Comparison of procalcitonin
with interleukin 8, C-reactive protein and differential white blood cell count
for the early diagnosis of bacterial infections in newborn infants. Pediatr
Infect Dis J 1999;18:666 71.
39. Franz AR, Steinbach G, Kron M, Pohlandt F. Reduction of unnecessary
antibiotic therapy in newborn infants using interleukin-8 and C-reactive
protein as markers of bacterial infections. Pediatrics 1999;104:44753.
40. Nupponen I, Andersson S, Ja¨rvenpa¨a¨ A-L, Kautiainen H, Repo H. Neutrophil
CD11b expression and circulating interleukin-8 as diagnostic markers for
early-onset neonatal sepsis. Pediatrics 2001;108:e12.
41. Buck C, Bundschu J, Gallati H, Bartmann P, Pohlandt F. Interleukin-6: a
sensitive parameter for the early diagnosis of neonatal bacterial infection.
Pediatrics 1994;93:54 8.
42. Messer J, Eyer D, Donato L, Gallati H, Matis J, Simeoni U. Evaluation of
interleukin-6 and soluble receptors of tumor necrosis factor for early
diagnosis of neonatal infection. J Pediatr 1996;129:574 80.
43. Panero A, Pacifico L, Rossi N, Mancuso G, Stegagno M, Chiesa C.
Interleukin 6 in neonates with early and late onset infection. Pediatr Infect
Dis J 1997;16:370 5.
44. Grella MJ. C-Reactive protein to determine the duration of antibiotic therapy
in infants with suspected sepsis [Letter]. Pediatrics 1997;100:900.
45. Apgar V. A proposal for a new method of evaluation of the newborn infant.
Curr Res Anesth Analg 1953;32:260 7.
46. Schmidt B, Kirpalani H, Rosenbaum P, Cadman D. Strengths and limita-
tions of the Apgar score: a critical appraisal. J Clin Epidemiol 1988;41:
47. Jepson HA, Talashek ML, Tichy AM. The Apgar score: evolution, limitations,
and scoring guidelines. Birth 1991;18(Suppl 2):S8392.
48. Marrin M, Paes BA. Birth asphyxia: does the Apgar score have diagnostic
value? Obstet Gynecol 1988;72(Suppl 1):S120 3.
49. Knaus WA, Zimmerman JE, Wagner DP, Draper EA, Lawrence DE. APACHE-
acute physiology and chronic health evaluation: a physiologically based
classification system. Crit Care Med 1981;9:5917.
50. Pollack MM, Ruttimann UE, Getson RR. Pediatric risk of mortality (PRISM)
score. Crit Care Med 1988;16:1110 6.
51. Richardson D, Tarnow-Mordi WO, Lee SK. Risk adjustment for quality
improvement. Pediatrics 1999;103(Suppl 1):S255 65.
52. International Neonatal Network. The CRIB (clinical risk index for babies
score)a tool for assessing initial neonatal risk and comparing perfor-
mance of neonatal intensive care units. Lancet 1993;342:193 8.
53. Richardson DK, Gray JE, McCormick MC, Workman K, Goldmann DA. Score
for neonatal acute physiology: a physiologic severity index for neonatal
intensive care. Pediatrics 1993;91:61723.
54. Escobar GJ, Fischer A, Li DK, Kremers R, Armstrong MA. Score for neonatal
acute physiology: validation in three Kaiser Permanente neonatal intensive
care units. Pediatrics 1995;96:918 22.
55. Petridou E, Richardson DK, Dessypris N, Malamitsi-Puchner A, Mantagos
S, Nicolopoulos D, et al. Outcome prediction in Greek neonatal intensive
care units using a score for neonatal acute physiology (SNAP). Pediatrics
1998;101:1037 44.
56. Pollack MM, Koch MA, Bartel DA, Rapoport I, Dhanireddy R, El-Mohandes
AA, et al. A comparison of neonatal mortality risk prediction models in very
low birth weight infants. Pediatrics 2000;105:10517.
57. Richardson DK, Phibbs CS, Gray JE, McCormick MC, Workman-Daniels K,
Goldmann DA. Birth weight and illness severity: independent predictors of
neonatal mortality. Pediatrics 1993;91:969 75.
58. Richardson DK, Corcoran JD, Escobar GJ, Lee SK. SNAP-II and SNAPPE-II:
simplified newborn illness severity and mortality risk scores. J Pediatr
59. Behrman RE, Shiono PH. Neonatal risk factors. In: Fanaroff AA, Martin RJ,
eds. Neonatal-perinatal medicine. St. Louis, MO: Mosby, 2002:1726.
60. Van Lente F, Pippenger CE. The pediatric acute care laboratory. Pediatr Clin
North Am 1987;34:231 46.
61. Mizuno S. Granulocyte colony-stimulating factor and neonatal infection
[Letter]. J Pediatr 1997;130:844 5.
62. Mathers NJ, Pohlandt F. Diagnostic audit of C-reactive protein in neonatal
infection. Eur J Pediatr 1987;146:14751.
63. Ainbender E, Cabatu EE, Guzman DM, Sweet AY. Serum C-reactive protein
and problems of newborn infants. J Pediatr 1982;101:438 40.
64. Speer CH, Bruns A, Gahr M. Sequential determination of CRP,
antitrypsin and haptoglobin in neonatal septicemia. Acta Paediatr Scand
1983;72:679 83.
65. Pourcyrous M, Bada HS, Korones SB, Barrett FF, Jennings W, Lockey T.
Acute phase reactants in neonatal bacterial infection. J Perinatol 1991;
11:319 25.
66. Berger C, Uehlinger J, Ghelfi D, Blau N, Fanconi S. Comparison of C-reactive
protein and white blood cell count with differential in neonates at risk for
septicemia. Eur J Pediatr 1995;154:138 44.
67. Sabel K-G, Wadsworth C. C-reactive protein (CRP) in early diagnosis of
neonatal septicemia. Acta Paediatr Scand 1979;68:82531.
68. Shine B, Gould J, Campbell C, Hindocha P, Wilmot RP, Wood CBS. Serum
C-reactive protein in normal and infected neonates. Clin Chim Acta
69. Schmidt BK, Kirpalani HM, Corey M, Low DE, Philip AGS, Ford-Jones EL.
Coagulase-negative staphylococci as true pathogens in newborn infants: a
cohort study. Pediatr Infect Dis J 1987;6:1026 31.
70. Racine A, Abribat D, Ensergueix G, Lucas Y, Poux JB. Value of C-reactive
protein assay for the early diagnosis of neonatal infection in a general
hospital maternity ward and pediatric department. Ann Pediatr (Paris)
71. Wasunna A, Whitelaw A, Gallimore R, Hawkins PN, Pepys MB. C-Reactive
protein and bacterial infection in preterm infants. Eur J Pediatr 1990;149:
424 7.
72. Ishibashi M, Takemura Y, Ishida H, Watanabe K, Kawai T. C-Reactive
protein kinetics in newborns: application of a high-sensitivity analytic
method in its determination [Technical Brief]. Clin Chem
2002;48:1103 6.
73. Gutteberg TJ, Haneberg B, Jorgensen T. Lactoferrin in relation to acute
phase proteins in sera from newborn infants with severe infections. Eur
J Pediatr 1984;142:379.
74. Forest JC, Larivie`re F, Dolce´ P, Masson M, Nadeau L. C-Reactive protein as
biochemical indicator of bacterial infection in neonates. Clin Biochem
1986;19:192 4.
75. Schouten-Van Meeteren NYN, Rietveld A, Moolenaar AJ, Van Bel F.
Influence of perinatal conditions on C-reactive protein production. J Pediatr
1992;120:621 4.
76. Chiesa C, Signore F, Assumma M, Buffone E, Tramontozzi P, Osborn JF, et
al. Serial measurements of C-reactive protein and interleukin-6 in the
immediate postnatal period: reference intervals and analysis of maternal
and perinatal confounders. Clin Chem 2001;47:1016 22.
77. Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood
count in health and disease. I. Reference values for neutrophilic cells.
J Pediatr 1979;95:89 98.
78. Escobar GJ. The neonatal sepsis work-up: personal reflections on the
development of an evidence-based approach toward newborn infections in
a managed care organization. Pediatrics 1999;103(Suppl 1):S360 73.
286 Chiesa et al.: Diagnosis of Neonatal Sepsis
79. Schelonka RL, Yoder BA, desJardins SE, Hall RB, Butler TJ. Peripheral
leukocyte count and leukocyte indexes in healthy newborn term infants.
J Pediatr 1994;125:603 6.
80. Schelonka RL, Yoder BA. The WBC count and differential: its uses and
misuses. Contemp Pediatr 1996;13:124 41.
81. Christensen RD, Rothstein G. Pitfalls in the interpretation of leukocyte
counts of newborn infants. Am J Clin Pathol 1979;72:608 11.
82. Schelonka RL, Yoder BA, Hall RB. Band neutrophil counts in neonates
[Letter: Reply]. J Pediatr 1995;126:505 6.
83. Philip AGS, Hewitt JR. Early diagnosis of neonatal sepsis. Pediatrics
1980;65:1036 41.
84. Philip AGS, Mills PC. Use of C-reactive protein in minimizing antibiotic
exposure: experience with infants initially admitted to a well-baby nursery.
Pediatrics 2000;106:e4.
85. Chiesa C, Pellegrini G, Panero A, Osborn JF, Signore F, Assumma M, et al.
C-Reactive protein, interleukin-6, and procalcitonin in the immediate post-
natal period: influence of illness severity, risk status, antenatal and
perinatal complications, and infection. Clin Chem 2003;49:60 8.
86. Assumma M, Signore F, Pacifico L, Rossi N, Osborn JF, Chiesa C. Serum
procalcitonin concentrations in term delivering mothers and their healthy
offspring: a longitudinal study. Clin Chem 2000;46:15837.
87. Gendrel D, Assicot M, Raymond J, Moulin F, Francoual C, Badoual J, et al.
Procalcitonin as a marker for the early diagnosis of neonatal infection.
J Pediatr 1996;128:570 3.
88. Monneret G, Labaune JM, Isaac C, Bienvenu F, Putet G, Bienvenu J.
Procalcitonin and C-reactive protein levels in neonatal infections. Acta
Paediatr 1997;86:209 12.
89. Lapillonne A, Basson E, Monneret G, Bienvenu J, Salle BL. Lack of
specificity of procalcitonin for sepsis diagnosis in premature infants
[Letter]. Lancet 1998;351:12112.
90. Sachse C, Dressler F, Henkel E. Increased serum procalcitonin in newborn
infants without infection [Technical Brief]. Clin Chem 1998;44:1343 4.
91. Petzold L, Guibourdenche J, Boissinot C, Benoist JF, Luton D, Demelier JF,
et al. Apport de la procalcitonine dans le diagnostic des infections
materno-foetales. Ann Biol Clin (Paris) 1998;56:599 602.
92. Maire F, He´raud MC, Loriette Y, Normand B, Be`gue RJ, Labbe´ A. Inte´reˆtde
la procalcitonine dans les infections ne´onatales. Arch Pe´diatr 1999;6:
93. Mu¨ller MM. Implementation of reference systems in laboratory medicine.
Clin Chem 2000;46:19079.
94. Felix NS, Nakajima H, Kagan BM. Serum C-reactive protein in infections
during the first six months of life. Pediatrics 1966;37:270 7.
95. Manucha V, Rusia U, Sikka M, Faridi MMA, Madan N. Utility of haemato-
logical parameters and C-reactive protein in the detection of neonatal
sepsis. J Paediatr Child Health 2002;38:459 64.
96. Hanson LA, Jodal U, Sabel K-G, Wadsworth C. The diagnostic value of
C-reactive protein. Pediatr Infect Dis 1983;2:879.
97. Roberts WL, Moulton L, Law TC, Farrow G, Cooper-Anderson M, Savory J, et
al. Evaluation of nine automated high-sensitivity C-reactive protein meth-
ods: implications for clinical and epidemiological applications. Part 2. Clin
Chem 2001;47:418 25.
98. Whiteside TL. Cytokine measurements and interpretation of cytokine
assays in human disease. J Clin Immunol 1994;14:32739.
99. Beckmann I, Visser W, Struijk PC, van Dooren M, Glavimans J, Wallenburg
HC. Circulating bioactive tumor necrosis factor-
, tumor necrosis factor-
receptors, fibronectin, and tumor necrosis factor-
inducible cell adhesion
molecule VCAM-1 in uncomplicated pregnancy. Am J Obstet Gynecol
100. de Bont ES, Martens A, van Raan J, Samson G, Fetter WP, Okken A, et al.
Diagnostic value of plasma levels of tumor necrosis factor
) and
interleukin-6 (IL-6) in newborns with sepsis. Acta Paediatr 1994;83:
696 9.
101. Ku¨ster H, Weiss M, Willeitner AE, Detlefsen S, Jeremias I, Zbojan J, et al.
Interleukin-1 receptor antagonist and interleukin-6 for early diagnosis of
neonatal sepsis 2 days before clinical manifestations. Lancet 1998;352:
102. Escobar GJ. Effect of the systemic inflammatory response on biochemical
markers of neonatal bacterial infection: a fresh look at old confounders
[Editorial]. Clin Chem 2003;49:212.
103. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, et
al. Towards complete and accurate reporting of studies of diagnostic
accuracy: the STARD initiative. Clin Chem 2003;49:1 6.
Claudio Chiesa
Alessandra Panero
John F. Osborn
Antonella F. Simonetti
Lucia Pacifico
National Research Council
Rome, Italy
Departments of
Public Health Science
La Sapienza
University of Rome
Rome, Italy
*Address correspondence to this author at: Department of
Pediatrics, La Sapienza, University of Rome, Viale R. Elena
324, 00161 Rome, Italy. Fax 39-06-4997-9215; e-mail
DOI: 10.1373/clinchem.2003.025171
Clinical Chemistry 50, No. 2, 2004 287
... The cut-off value is an appropriate threshold above which results can be considered abnormal. These values reported in the literature range from 1.5 to 20 mg/L with wide-ranging sensitivities and specificities (Chiesa et al., 2004). ...
Full-text available
Background: Neonatal sepsis (NS) is a major health problem throughout the world. The diagnosis of sepsis is challenging due to the non-specific nature of the clinical presentation, the variety of other neonatal disorders with the differential diagnostic workup, lack of sensitivity and specificity of available diagnostic procedures, and the delay in the results of blood cultures in addition to high negative results reported. The diagnosis of suspected sepsis has to be based on clinical symptoms together with biochemical parameters. A diagnostic marker with high diagnostic sensitivity and specificity would be a valuable tool for decreasing the burden of neonatal sepsis Purpose: Evaluate the validity of interleukin-6 (IL-6) in the early diagnosis of neonatal sepsis or the use of a combination of diagnostic markers, C-reactive protein (CRP), and IL-6. Methods: The study included 30 patients with NS (Group I) and 30 healthy newborns as control (Group II) were admitted to Neonatal Intensive Care Unit (NICU) from January 2017 to June 2017. All neonates were subjected to history taking, clinical examination, and laboratory investigations including complete blood count (CBC), blood culture and sensitivity testing, CRP, and IL-6. Results: the most causative organism of neonatal sepsis in NICU was Klebsiella spp. followed by CONS. IL-6 results with cut-off value 50pg/ml, the sensitivity was 100%, the specificity was 90.32%, the positive predictive value of 90.63%, negative predictive value, and the diagnostic accuracy was 95.16. Moreover, IL-6 levels are significantly higher statistically in NS patients than controls. Conclusion: This study validated the diagnostic capability of IL-6 and showed that the combination of CRP and IL-6 as a panel for the early diagnosis of NS could enhance the sensitivity in the diagnosis of NS and may provide a new diagnostic strategy for NS patients Objective: Evaluate the validity of interleukin-6 (IL-6) in early diagnosis of neonatal sepsis or the use of combination of diagnostic markers, C-reactive protein (CRP) and IL-6. Patients and methods: The study included 30 patients with NS (Group I) and 30 apparently healthy newborns as control (Group II) were admitted to Neonatal Intensive Care Unit (NICU) from January 2017 to June 2017. All neonates were subjected to history taking, clinical examination, and laboratory investigations including: complete blood count (CBC), blood culture and sensitivity testing, CRP, and IL-6. Results: the most causative organism of neonatal sepsis in NICU was klebsiella spp. followed by CONS. IL-6 results with cut-off value 50pg/ml, the sensitivity was 100%, the specificity was 90.32%, positive predictive value of 90.63%, negative predictive value, and the diagnostic accuracy was 95.16. Moreover, IL-6 levels are significant higher statistically in NS patients than controls. Conclusion: This study validated the diagnostic capability of IL-6 and showed that the combination of CRP and IL-6 as a panel for the early diagnosis of NS could enhance the sensitivity in the diagnosis of NS and may provide a new diagnostic strategy for NS patients.
... A similar study by Sakha et al. [23] showed that the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of PCT were 66.7%, 50%, 28.6%, and 83.3%, which is lower than that of the current study. Chiesa et al. [24] studied the dependability of PCT concentration in 28 infants having severe early-onset of neonatal sepsis. They found that the sensitivity, specificity, NPV, and PPV were 92.6%, 97.5%, 96.8%, and 94.3%, respectively. ...
Full-text available
Aim: The aim of this study was to evaluate the early indicators of sepsis (sepsis screening) and their statistical correlation with sepsis in neonatal abdominal surgery. Materials and Methods: A prospective observational study was performed on thirty consecutive neonate cases aged between 0 and 28 days with surgical abdomen at the Paediatric Surgery Department, ABVIMS and Dr. Ram Manohar Lohia Hospital, New Delhi. The study duration was 18 months. Septic screening was done in all neonates on day 0, 1, 3, 7, and 14 days of surgery with serum procalcitonin, C-reactive protein, total leukocyte count, immature/total neutrophil ratio, and microerythrocyte sedimentation rate. A septic screening-positive patient (three or more positive parameters out of five) was correlated with sepsis and analysis was done. Results: A total of 30 neonates of abdominal surgical cases were included consequently, out of which 56.7% (n = 17) were male and 43.3% (n = 13) were female. Maximum cases were of congenital diaphragmatic hernia 20% (n = 6) and then anorectal malformation 16.7% (n = 5). About 70% of neonates were sepsis screening positive. Fifty percentage of neonates were diagnosed to have sepsis on the clinical or laboratory findings, so sensitivity and specificity of sepsis screening were 93.33% and 40%, respectively. There was total 30% mortality in this study. Conclusion: Sepsis screening is an early marker of sepsis, which can be used to help in early detection of neonatal surgical sepsis and timely intervention that can lead to decrease mortality and morbidity in neonatal surgery.
... N eonatal sepsis is a potentially fatal disease that occurs when a serious infection affects infants within the 1 st month of life; it is a systemic disease defined by the systemic inflammatory response syndrome. [1] The presence of at least two clinical signs and/or two or more symptoms such as temperature waving, hypotension, reduced perfusion with paleness and spotted skin, acidic breath, arrhythmia, apnea, breathing difficulties, grunting, hypoxemia, restlessness, stupor, epilepsy, feeding intolerance, abdominal discomfort, jaundice, hemorrhage, and variability in white blood cells [WBCs] count). [2] Despite using an effective antimicrobial therapy with advanced services provided in the intensive care unit (ICU), morbidity and mortality of neonatal sepsis remain high in immunocompromised premature neonates. ...
... This requires a multilevel approach because only one drug may not be that effective in severe conditions. The remnants of toxins produced by bacteria can be encountered by antitoxin drugs to neutralize any immune response [40]. ...
Full-text available
COVID-19 acquired symptoms have affected the worldwide population and increased the load of Intensive care unit (ICU) patient admissions. A large number of patients admitted to ICU end with a deadly fate of mortality. A high mortality rate of patients was reported with hospital-acquired septic shock that leads to multiple organ failures and ultimately ends with death. The patients who overcome this septic shock suffer from morbidity that also affects their caretakers. To overcome these situations, scientists are exploring progressive theragnostic techniques with advanced techniques based on biosensors, biomarkers, biozymes, vesicles, and others. These advanced techniques pave the novel way for early detection of sepsis-associated symptoms and timely treatment with appropriate antibiotics and immunomodulators and prevent the undue effect on other parts of the body. There are other techniques like externally modulated electric-based devices working on the principle of piezoelectric mechanism that not only sense the endotoxin levels but also target them with a loaded antibiotic to neutralize the onset of inflammatory response. Recently researchers have developed a lipopolysaccharide (LPS) neutralizing cartridge that not only senses the LPS but also appropriately neutralizes with dual mechanistic insights of antibiotic and anti-inflammatory effects. This review will highlight recent developments in the new nanotechnology-based approaches for the diagnosis and therapeutics of sepsis that is responsible for the high number of deaths of patients suffering from this critical disease.
Introduction: Enterovirus (EV) infections are the most frequent infections in the neonatal period and in many cases lead to hospital admission of the newborn (NB). The aim of this study was to determine the incidence of EV in the etiology of neonatal meningitis and to define the clinical characteristics of newborns with EV meningitis. Material and method: Retrospective observational cohort study. Including 91 NBs with meningitis and gestational age greater than 34 weeks gestational age (GA) attended in our center over a period of 16 years. Results: The percentage of NBs with EV meningitis was higher than that of NBs with bacterial meningitis (BM) and accounted for 78% (n=71). Half of the NBs with EV infection had a history of epidemic environment among their caregivers. Fever was present in 96% of cases as a clinical sign and, in general, sensory disturbances represented the main neurological alterations. Antibiotics (ATB) were given to 71.4% of patients with EV infection. Detection of EV in CSF samples showed a high sensitivity for the diagnosis of EV meningitis. The most frequently implicated EV types were echovirus 11, coxsackievirus B5, echovirus 18, 25 and 7. Conclusions: The results of this series show that enterovirus infection is a common cause of neonatal meningitis. These data underline the importance of rapid EV testing of infants with suspected meningitis. This allows early diagnosis and reduces antibiotic treatment, hospitalization time and related costs.
Objectives Biomarkers in sepsis are an arena of avid research as they can facilitate timely diagnosis and help reduce mortality. Presepsin is a promising candidate with good diagnostic performance reported in adult and neonatal studies. However, there is no clear consensus about its utility in the pediatric age group. This study aimed to synthesize scientific evidence regarding the diagnostic and prognostic performance of presepsin in pediatric sepsis. Data Sources A systematic literature search was conducted in MEDLINE/PubMed, Cochrane Central Register of Clinical Trials, Google Scholar, and Scopus to identify relevant studies reporting the diagnostic and prognostic accuracy of presepsin. Study Selection Using Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, we retrieved all controlled trials and observational studies on presepsin as a biomarker in children aged <19 years with sepsis. Data Extraction Two authors independently performed study screening, data extraction, and quality assessment of the included studies. Data Synthesis Among the 267 citations identified, a total of nine relevant studies were included in the final analysis. The pooled diagnostic sensitivity and specificity of presepsin were 0.99 (95% confidence interval [CI]; 0.97–1.00) and 0.88 (95% CI; 0.83–0.92), respectively, with a diagnostic odds ratio (DOR) of 28.15 (95% CI; 0.74–1065.67) and area under the curve (AUC) in summary receiver operating curve of 0.89. Prognostic accuracy for presepsin had a sensitivity of 0.64 (95% CI; 0.35–1.0), specificity of 0.62 (95% CI; 0.44–0.87), and DOR of 3.3 (95% CI; 0.20–53.43). For procalcitonin, the pooled sensitivity for diagnostic accuracy was 0.97 (95% CI; 0.94–1.00), specificity was 0.76 (95% CI; 0.69–0.82), DOR was 10.53 (95% CI; 5.31–20.88), and AUC was 0.81. Conclusion Presepsin has good diagnostic accuracy with high sensitivity and specificity. Its prognostic accuracy in predicting mortality is low.
Full-text available
Background: Neonatal sepsis is a major cause of morbidity and mortality in newborn.Early diagnosis of neonatal sepsis is the top priority to all clinicians for early initiation of appropriate treatment which will reduce the morbidity and mortality in neonates and minimizes the unnecessary antibiotic therapy which will lead to overwhelming threat of multidrug-resistant organism. Aims and Objectives: The recent study was undertaken to study the reliability and effectiveness of hematological parameters and Rodwell’s hematological score for early diagnosis of sepsis in neonates. Materials and Methods: In this study,212 neonates were studied out of which 88 were blood culture proven sepsis and 39 were culture negative but hematological parameters positive probable sepsis.The rest 85 were both culture negative and hematological parameter negative and were considered normal baby and taken as control group in this study. Hematological parameters were studied in all cases included in the study which include total leukocyte count, absolute neutrophil count, immature/total neutrophil ratio, platelet count, C-reactive protein (CRP), and other components of Rodwell’s scoring criteria. Recent statistical analysis of all the components and hematological scoring system (HSS) done to find out the reliability and effectiveness. Results: Analysis of hematological parameters showed that absolute neutrophil count, immature to total neutrophil ratio and platelet count have a better sensitivity and specificity in relation to other parameters.CRP and total leukocyte count have better specificity but sensitivity was very low.In the HSS, it has been shown that this scoring system was highly sensitive and specific with a high positive predictive value and negative predictive value.Among the different parameters studied, association of total leukocyte count(P<0.0001)and reduced platelet count(P<0.0001) is the most consistent indicator found in this study but association with CRP was found to be statistically insignificant(P=0.0703). Conclusion: HSS system, total leukocyte count, absolute neutrophil count and platelet count are very reliable indicator for early diagnosis of neonatal sepsis which is cheap, simple, less time consuming, and cost effective.
Resumen Introducción Las infecciones por enterovirus (EV) constituyen las infecciones más frecuentes en el periodo neonatal y provocan en muchos casos el ingreso hospitalario del recién nacido (RN). El objetivo del estudio es conocer la incidencia de los EV en la etiología de las meningitis neonatales y definir qué características clínicas presentan los RN con meningitis por EV. Material y método Estudio observacional retrospectivo de cohortes. Incluye 91 RN con meningitis y edad gestacional mayor de 34 semanas (SG) atendidos en nuestro centro durante un periodo de 16 años. Resultados El porcentaje de RN con meningitis por EV fue superior al de RN con meningitis bacteriana y representó el 78% (n = 71). La mitad de los RN con infección por EV presentó antecedentes de ambiente epidémico entre sus cuidadores. La fiebre apareció en el 96% de los casos como signo clínico y, en general, las alteraciones del sensorio representaron las principales alteraciones neurológicas. Un 71,4% de los pacientes con infección por EV recibió antibióticos. La detección de EV en muestras de LCR mostró una elevada sensibilidad para el diagnóstico de meningitis por EV. Los tipos de EV más frecuentemente implicados fueron echovirus 11, coxsackievirus B5, echovirus 18, 25 y 7. Conclusiones Los resultados de esta serie muestran que la infección por enterovirus es una causa común de meningitis neonatal. Estos datos subrayan la importancia de realizar pruebas de detección rápida de EV en lactantes con sospecha de meningitis. Ello permite obtener un diagnóstico precoz y reducir el tratamiento antibiótico, el tiempo de hospitalización y los costes relacionados.
In 1996 the American College of Obstetricians and Gynecologists, along with the American Academy of Pediatrics and the Centers for Disease Control and Prevention, recommended one of two courses for healthcare providers toidentify women who should receive intrapartum antibiotic prophylaxis to prevent group B streptococcal (GBS) disease in infants aged 1 week or less: screening by culture or assessment of clinical risk factors. These two approaches have now been compared in a retrospective cohort study of approximately 630,000 live births. Study cases were accessed in the years 1998-1999 in eight geographic areas of the United States where active surveillance for GBS infection was under way. A random sample of these areas yielded 5144 births. The 312 cases of early onset GBS disease translated into an overall incidence of 0.5 case per 1000 live births. No clinical risk factors for transmitting GBS disease were identified in 62% of cases. More than half of the women in the overall population (52%) were screened before delivery. On univariate analysis, prenatal culture screening correlated with a lower risk of early onset GBS disease than was the clinical risk-based approach; the relative risk (RR) was 0.48. The most prominent risk factors were a previous infant having GBS disease and intrapartum fever. GBS bacteriuria during pregnancy was not a factor, but more than 80% of all women with bacteriuria received intrapartum antibiotics. Screening continued to be associated with a reduced risk on multivariate analysis (RR, 0.46). Screening protected full-term infants when analyzed separately. Intrapartum antibiotic treatment was 89% effective in preventing early onset GBS disease in culture-positive women. It was estimated that, assuming perfect implementation of either strategy, intrapartum antibiotics would be used in 31% of screened women and 29% of those having risk-based assessment. These findings suggest that it makes sense to further consider a recommendation favoring universal screening to prevent GBS disease in newborn infants.
The increasingly widespread use of antibiotics during labor and delivery raises the question of whether the incidence or causes of early onset sepsis (within 72 hours of birth) in very-low-birth-weight (VLBW) infants havechanged. The study population included 5447 VLBW infants, weighing from 401 to 1500 g at birth in the years 1998-2000, who had at least one blood culture in the first 3 days of life. They were compared with 7606 VLBW infants born during 1991-1993. Early onset sepsis was diagnosed in 84 study infants for a rate of 15.4 per 1000 live births. The rate in 1991-1993 had been 19.3 per 1000 births. Approximately 60% of infections in the recent cohort were due to Gram-negative organisms, and 44% were attributed to Escherichia coli. In contrast, Gram-negative organisms caused a majority of the earlier infections. The rate of infection by group B streptococci fell from 5.9 to 1.7 per 1000 live births, whereas that of E. coli infection rose from 3.2 to 6.8 per 1000. All but 2% of cases occurred in infants weighing more than 500 g, and the gestational age at birth was similar in the two periods. In the later period, 85% of E. coli isolates were resistant to ampicillin, whereas all 26 isolates tested were sensitive to third-generation cephalosporins. Maternal antibiotic use during hospitalization could not be related to either the risk of E. coli or group B streptococcal sepsis or the risk of death from either of these infections. After adjusting for gestational age, septic infants were at significantly increased risk of respiratory distress syndrome, severe intraventricular hemorrhage/periventricular leukomalacia, and bronchopulmonary dysplasia. Septic infants also were likelier than others to die within 4 months of birth, again adjusting for gestational age. Infants with Gram-negative sepsis were likelier to die than those infected by a Gram-positive organism, but the difference was significant only within 3 days after birth. Early onset sepsis, especially that caused by Gram-negative organisms, remains an important cause of neonatal death.
Assessment of different hospitals' performance in neonatal intensive care has tended to rely on risk of mortality adjusted only for birthweight. We have developed a neonatal scoring system, CRIB (clinical risk index for babies), to take account of other factors. Scores are given for birthweight, gestational age, maximum and minimum fraction of inspired oxygen and maximum base excess during the first 12 h, and presence of congenital malformations. CRIB was developed retrospectively in a cohort of 812 infants of birthweight 1500 g or less or gestational age less than 31 weeks treated in four UK tertiary hospitals between 1988 and 1990. The scoring system was then validated by comparing its value as a predictor of hospital death with that of birthweight in a separate cohort of 488 similar infants. The area under the receiver operating characteristic (ROC) curve for predicting death in this validation cohort (a measure of the predictor's accuracy) was significantly greater for CRIB than for birthweight alone (0·90 [SE 0·05] vs 0·78 [0·03], p=0·03). Both indices of initial risk were used to assess the performance of nine tertiary and four non-tertiary hospitals in the UK. We looked at hospital mortality among the 1300 infants in the development and validation cohorts plus 248 other infants of birthweight 1500 g or less or gestational age less than 31 weeks. Without adjustment for initial risk, there was no clear difference in group mortality between the tertiary and non-tertiary hospitals. However, calculation of risk- adjusted mortality by means of CRIB showed that babies were twice as likely to die in the non-tertiary hospitals as in the tertiary hospitals (odds ratio 2·12 [95% Cl 1·39-3·24]). Adjustment for birthweight alone also showed higher odds of death in non-tertiary hospitals (1·45 [1·01-2·11]). CRIB is a robust index of initial neonatal risk that is more accurate than birthweight and simple enough for routine use. International comparisons of staffing policy and organisation of neonatal intensive care units are needed to investigate the reasons for variations in performance.