Paper 14 DATA

Paediatric and Perinatal Epidemiology 1997, 11, 4456

Clinical approach to the analysis of causes of death

in the first two years of life of very-low-birthweight
infants in a multicentre setting

Carlo Corchia*, Amedeo Spagnolo{, Umberto de

Vonderweid{, Carlo Zorzi}, Valeria Chiandotto},
Stefano Chiappe**, Patrizia Colarizi{{, M. Angela
Didato{{ and Roberto Paludetto}}
*Department of Paediatrics and Neonatology, University of Sassari,
{Istituto Superiore di Sanita, Rome, {Department of Paediatrics,
University of Trieste, }Department of Paediatrics, University of Padua,
}Division of Neonatology, General Hospital, Udine, **Department of
Child Health and Neonatology, University of Cagliari, {{Neonatal
Intensive Care Unit, University `La Sapienza' Rome, {{Department of
Paediatrics, University of Palermo, and }}Department of Paediatrics,
University `Federico II', Naples, Italy

Summary. Mortality in the first 2 years of 634 very-low-birthweight

infants admitted to eight neonatal intensive care units in Italy, and the
factors associated with the net probability of death from each cause, were
studied by means of the Cox proportional hazard model. A clinical
classification of the causes of death was used. Overall mortality was
33.7% (intercentre range 12.652.9%). The highest cause-specific mortality
rates were observed for respiratory problems, intra-ventricular haemor-
rhage (IVH) and infections (14.5%, 6.3% and 5.7% respectively). The
leading causes of death were respiratory problems and IVH in the first
week of life, infections from the second week up to the end of the first
month, and bronchopulmonary dysplasia (BPD) afterwards. Birthweight
51000 g, gestational age 530 weeks, absence of spontaneous respiratory
activity, unknown body temperature and pH 57.20 at admission were
associated with death from respiratory problems and IVH. Male sex,
birthweight 51000 g and unknown body temperature at admission were
associated with death from BPD. Mortality from infections was higher in
one centre; no other differences emerged among the eight NICUs. The

Address for correspondence: Dr Carlo Corchia, Department of Neonatology, Annunziata

Hospital, Via F. Migliori, 87100 Corsenza, Italy.

44
# 1997 Blackwell Science Ltd.
 Paper 14 DATA
Causes of death in VLBW infants 45

classification of the causes of death employed and the use of the net
probabilities of death appear as practical and useful instruments to study
the relationship between specific aspects of medical care and mortality,
and to investigate the reasons for differences in performance between
neonatal units.

Introduction
Studies on outcomes of very-low-birthweight infants (VLBWIs) have focused
mainly on mortality and long-term morbidity.15 The causes of death have seldom
been investigated,6,7 although their analysis can be helpful for perinatal and
neonatal evaluation. Cause-specific mortality rates may reflect particular aspects
of care for VLBWIs, whose survival is strictly related to medical procedures.
Some classifications of perinatal, neonatal and infant deaths have been
proposed, in which causes of death are grouped into physiopathological
categories that can help to identify the factors associated with mortality in a
specific clinical setting.815 These classifications are predominantly based on
pathological and necropsy findings or on unfavourable events of pregnancy and
delivery, but often do not allow a clear distinction among causes when applied to
VLBWIs, whose deaths would be largely attributed to `conditions associated with
immaturity', without any further distinction. Moreover, previous studies of causes
of death in VLBWIs have not taken into account the net probability of death from
each cause that expresses the probability of dying from a specific cause adjusted
for competing causes.
In the present study, a classification of the causes of death of VLBWIs is
presented in which the single causes are grouped in such a way that each group
can reflect specific aspects of medical care. This classification has been used in
order (1) to study the cause-specific net probability of death at different age
intervals in the first two years; and (2) to disclose, among some infants'
characteristics and a set of variables describing the condition of the infants at
admission to the NICU, those associated with the cause-specific risks of death.
Data on VLBWIs prospectively collected in eight Italian NICUs over a 2-year
period were used.

Subjects and methods

Study subjects were 634 babies enrolled in the Italian multicentre study on very-
low-birthweight infants. Eight NICUs participated in this project, located in
different geographical areas of the country. They prospectively collected data on

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
46 C. Corchia et al.

all the newborns weighing 5001499 g at birth admitted during the calendar years
1987 and 1988. Surviving infants were followed up using a common predefined
protocol. The whole project, the characteristics of the study infants and the results
about in-hospital mortality and follow-up status at 2 years of age have been
described elsewhere.16
In the present study, mortality up to 2 years of age and causes of death were
investigated. For each infant who died, the physician responsible for the project in
each centre identified the main cause of death, according to clinical judgement,
laboratory data and necropsy finding. Subsequently, two of the authors (CC and
CZ) independently coded these causes according to the Ninth Revision of the
International Classification of Diseases and Causes of Death.17 In one case, the
coders disagreed and a consensus was necessary. Causes were then grouped into
the following categories: birth defects, asphyxia, respiratory problems, intra-
ventricular haemorrhage (IVH), infections, bronchopulmonary dysplasia (BPD)
and others. The group `asphyxia' comprised conditions arising during or shortly
before the onset of labour and delivery, leading to acute impairment of gaseous
exchanges. The category `respiratory problems' included, along with respiratory
distress syndrome (RDS) and other diseases of the lungs, also the deaths due to
`immaturity', since it was judged that the immaturity of the respiratory system
had given the largest contribution to death in these cases. The diagnosis of IVH
was accepted only when based on ultrasound and/or necropsy ascertainment.
BPD was diagnosed when respiratory signs persisted longer than 28 days of age
and were associated with characteristic radiographic findings.18
Analysis was performed using the life-table method by means of the BMDP
statistical package.19 The whole observation period was divided into the following
age intervals: 023 hours, 16 days, 713 days, 1427 days, 2859 days, 23
months, 45 months, 611 months, 1217 months and 1823 months. The
cumulative survival with 95% confidence intervals (95% CI) and the net
probability of death from each cause were calculated. The latter expresses the
probability of death adjusted for competing causes, and was computed for each
cause by regarding infants who died from all the other causes as if they had been
withdrawn alive, under the assumption that withdrawals had the same cause-
specific risk of death as those remaining under observation.20 Deaths as a result of
birth defects were not judged to compete with other causes and were excluded.
Separate stratification analyses were performed according to certain character-
istics of the study infants, namely sex, birthweight (500999 or 10001499 g),
gestational age (530 or 30+ weeks), place of birth (inborn or outborn) and NICU
at which the patients were cared for. Survival curves and the net probabilities of
death were compared using the Mantel test.20 Statistical significance was accepted
at P 50.05.
These variables were subsequently entered in a set of Cox proportional hazard
models, in order to explore the effect of each factor on the net probabilities of

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
Causes of death in VLBW infants 47

death adjusted for all the other factors of interest.21 The following variables,
recorded at admission to the NICUs and found to be associated with an increased
risk of in-hospital mortality in the same patients,16 were also added to the models:
full course of prenatal steroids (yes or no), Apgar score at 1 minute (54 or 4+),
spontaneous respiratory activity (yes or no), body temperature (5358C or 358+)
and pH (57.20 or 7.20+). The Cox model was used to calculate the cumula-
tive hazard function, where the dependent variable was the survival time and
the independent variables or covariates were the prognostic factors of interest.
All the variables were simultaneously entered into the model, and for each
one the OR and 95% CI of the cause-specific net probability of death
were calculated.

Results
Of the 634 babies enrolled, 191 weighed 500999 g at birth and 443 weighed 1000
1499 g. Gestational age ranged from 2139 completed weeks. After admission, no
infant was transferred to other units. A total of 210 infants (33.1%) died before
discharge from hospital; those who survived were discharged at a median age of
57 days (range 17365 days). Four more babies died after discharge and 30 were
lost to follow-up. Mortality by NICU of admission showed a wide range of values,
from a minimum of 12.6% to a maximum of 52.9%.
Survival was 92.3% at the end of the first day of life, 77.1% at 1 week, 70.9% at
1 month and 66.1% at 1 year (Fig. 1). Only one baby died beyond the first birthday,
and none died between 18 and 23 completed months.
Necropsy examination was performed in 62.6% of the infants who died (134
out of 214), with values ranging from 7.1% to 100% in the eight NICUs. No
relationship could be demonstrated between the frequency of necropsy examina-
tion and the distribution of the causes of death in each NICU.
The cause-specific death rates are shown in Table 1. Respiratory problems
were the leading causes, followed in order by IVH, infections, BPD, other causes,
asphyxia and, finally, birth defects. The respiratory problems group accounted for
nearly 44% of all deaths and included 62 cases of RDS, 12 cases of other lung
diseases and 18 deaths attributed to `immaturity'. Infections included 28 sepsis,
one meningitis and seven bronchopneumonia. In the birth defects group, there
were two anencephalies, two congenital heart disease, two multiple congenital
anomalies, one trisomy 13 and one trisomy 18. Of the 13 deaths in the group `other
causes', nine occurred before discharge and included three intravascular
disseminated coagulations, three necrotising enterocolitis, one hypovolaemic
shock, a case of cardiac arrest not otherwise specified and a death from pulmonary
oedema following surgery for oesophageal atresia; the other four deaths occurred
after discharge and included three cases of sudden infant death syndrome and one
endocarditis following surgery for infundibular aortic stenosis.

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
48 C. Corchia et al.

Figure 1. Cumulative survival of the infants enrolled in the study. Ninety-five per cent
confidence intervals are also shown.

The cause-specific net probabilities of death in each age interval for infants still
surviving at the beginning of the interval are shown in Table 2. The predominant
causes in the first week of life were respiratory problems and IVH; deaths from
asphyxia were mostly concentrated in the first 24 hours. From the second week up
to the end of the first month, infections were the predominant causes. All deaths
from BPD occurred beyond the first month and before 1 year of age. Finally, the
probability of dying from other causes was quite uniform in all the age intervals.

Table 1. Cause-specific death rates in the first two years of life

n Rates (%) 95% CI
Birth defects 8 1.26 0.392.13
Asphyxia 10 1.58 0.612.55
Respiratory problems 92 14.51 11.7717.25
IVH 40 6.31 4.428.20
Infections 36 5.68 3.887.48
BPD 15 2.37 1.193.55
Others 13 2.05 0.953.15

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
Causes of death in VLBW infants 49

Table 2. Cause-specific net probabilities of death (%) in each age interval for survivors at
the beginning of the interval
Causes of death

Age Respiratory
interval Asphyxia problems IVH Infections BPD Others
023 hours 1.32 5.01 0.82 0 0 0.16
16 days 0.19 9.30 4.76 2.23 0 0.37
713 days 0 1.68 0.84 2.51 0 0.21
1427 days 0 0.22 0.44 1.75 0 0.44
2859 days 0.23 0 0.23 0.69 1.38 0.46
23 months 0 0 0.49 0 0.98 0.49
45 months 0 0 0 0.25 0.76 0.25
611 months 0 0 0 0 0.52 0.26
1217 months 0 0 0 0 0 0.26
1823 months 0 0 0 0 0 0

The net probabilities of death from respiratory problems, IVH, infections and
BPD were higher for infants of birthweight 51000 g and gestational age 530
weeks (Table 3). Mortality from IVH was higher for the outborn than the inborn
babies. Statistically significant differences among NICUs were found in the net
probabilities of death from respiratory problems and IVH, which disappeared
when the two groups of causes were pooled together, and from infections, with
NICU H showing a higher value than all the other NICUs. No associations were
found between mortality from asphyxia or other causes and the stratification
variables taken into account, or between sex and any of the cause-specific net
probabilities of death.
The results of the multivariate analyses with the Cox model are presented in
Table 4. Since in 42 subjects data about body temperature at admission were
missed, a value of 358C+ was designated as reference and two indicator variables
were used, namely 5358C and unknown. Also the NICU where the infants were
cared for was taken as an indicator variable, with NICU D as reference.
Mortality from asphyxia was associated with a pH 57.20 at admission. An
increased risk of death from respiratory problems and IVH together was found for
birthweight 51000 g, gestational age 530 weeks, absence of respiratory activity,
unknown body temperature and pH 57.20 at admission, but some differences
emerged when the two groups of causes were analysed separately: mortality from
IVH was associated with gestational age 530 weeks and absence of respiratory
activity, while mortality from respiratory problems was associated with birth-
weight 51000 g, gestation 530 weeks, unknown body temperature and pH
57.20 at admission.
Mortality from infections was higher when birthweight was 51000 g and in
NICU H. An increased risk of dying from BPD was found for birthweight

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
Causes of death in VLBW infants 51

Table 4. Adjusted odds ratios of cause-specific net probabilities of death in the first 2 years
of life according to selected variables
Variables Odds ratio 95% CI
Asphyxia
pH at admission 57.20 5.51 1.0828.17
Respiratory problems or IVH
Birthweight 500999 g 2.11 1.403.18
Gestational age 530 weeks 5.09 2.829.18
No spontaneous respiratory activity 2.21 1.463.34
Body temperature at admission unknown 2.23 1.263.94
pH at admission 57.20 2.05 1.373.06
Respiratory problems
Birthweight 500999 g 3.06 1.805.20
Gestational age 530 weeks 3.41 1.716.77
Body temperature at admission unknown 2.15 1.134.10
pH at admission 57.20 2.35 1.443.84
IVH
Gestational age 530 weeks 11.34 3.3638.31
No spontaneous respiratory activity 4.64 2.199.82
Infections
Birthweight 500999 g 3.71 1.668.28
Centre H 4.61 1.3815.39
BPD
Sex (females) 0.31 0.100.95
Birthweight 500999 g 3.48 1.0911.06
Body temperature at admission unknown 7.58 1.3343.13
Only the statistically significant variables are shown. Variables entered into the Cox models:
sex, birthweight, gestational age, place of birth, NICU, antenatal steroids, Apgar score at 1
minute, spontaneous respiratory activity at admission, body temperature at admission and
pH at admission.

51000 g and unknown body temperature at admission, while the risk was lower
for female infants. Finally, no association could be demonstrated between the net
probability of death from `other causes' and any of the variables investigated.

Discussion
The overall mortality rate found in this study is higher than that reported
elsewhere for VLBWIs,1,4,5,7,22,23 although a direct comparison among studies is
difficult because of the different criteria used to calculate mortality and the
heterogeneous groups of subjects taken into account. A cumulative survival rate of
74% up to the end of the second year of life was given for VLBWIs in a report from
one centre1, whereas in the present study the overall value was 66%.

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
52 C. Corchia et al.

The validity of the study of the causes of death has been questioned, because it
was thought to be a simplistic approach to the relationships between causes and
effects, neglecting interactions,24 and because the best method of preventing deaths
ascribed to a particular cause (for example, BPD) may not be to address the cause
directly. However, it proves useful for clinical purposes, especially for evaluation
and spacetemporal comparisons of the performances of neonatal units.
In the present study, deaths caused by respiratory problems, IVH and BPD, in
spite of their strong relationships, were considered separately, in order to explore
their different temporal patterns and specific risk factors. All the infants whose
death was due to `immaturity' not otherwise specified weighed less than 1000 g,
and 13 weighed less than 750 g. These cases were probably the consequence of a
multisystem involvement, although they were all included in the group of res-
piratory problems. The three deaths attributed to necrotising enterocolitis were
classified in the group of `other causes' because of the multiple factors involved in
the origin and outcome of the disease.25
The cause-specific net probabilities of death were used for the analysis. With
this approach, the probability of death from a specific cause is adjusted for other
competing causes under the assumption that, had the subjects dying from other
causes remained alive, they would have had the same risk for the cause of interest
as the surviving population.20.
Two-thirds of the 214 deaths (68%) occurred in the first week of life and 86% in
the first 28 days. A characteristic temporal pattern emerged from the cause-specific
net probabilities of death in each age interval. Respiratory problems were the
leading causes in the first week, followed by IVH. Strict links exist between these
two groups of events, and the attribution of the main cause of death when both
conditions are simultaneously present is influenced by the physicians' adherence
to one or other of the aetio-pathogenetic hypotheses of the diseases and by the
judgement about the clinical, instrumental and necropsy data of any single
patient.26 In this respect, the possibility of performing an ultrasound scan soon
after birth appears to be of great value to reveal congenital and very early onset
IVHs, which might represent more than one-third of all IVHs.27 On the other
hand, even in the most experienced hands, neonatal cranial ultrasonography may
show a certain degree of interobserver variability and lack of precision.28,29 All
these reasons probably contributed to the variability found in mortality from
respiratory problems and IVH among the NICUs. This interpretation is supported
by the decrease in variability when both causes were considered together.
From the second to the fourth week of life the leading causes of death were
infections. This finding is in agreement with the observed increased importance of
infectious diseases as causes of infant mortality in low birthweight infants.30
Deaths observed after the first month were 14% of the total; this value is in the 7
17% range found in other studies.1,4,5,31,32 The most frequent cause of death in this
age interval was BPD. Infections and BPD are the likely consequence of the

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
Causes of death in VLBW infants 53

increasingly extensive use of intensive care techniques that prolong survival but
expose the small and immunologically deficient babies to a high risk of acquiring
severe nosocomial infections, BPD or both.30,33
An increased risk of dying from asphyxia was found when pH at admission
was 57.20. In this case, acidosis is only an indicator variable, reflecting one of the
biochemical consequences of asphyxia itself and the effectiveness of subsequent
medical care. Moreover, it is possible that in some cases the presence of a low pH
at admission was one of the criteria used to attribute the death to `asphyxia'. A
low pH was also related to an increased risk of death from respiratory problems,
together with an unknown body temperature at admission. It can be supposed
that the body temperature was not recorded when the babies were very ill, and
this can therefore be taken as an indicator of the severity of the clinical condition at
admission; it is likely that in such circumstances infants were also hypothermic.
Hypoxia, hypothermia and acidosis are known to reduce surfactant production
and this can explain the association with deaths from respiratory problems.34 The
risk of death from IVH showed a strong relationship with gestational age 530
weeks and with the absence of respiratory activity at admission. These
associations probably reflect the anatomical and physiological immaturity of the
subependymal germinal matrix in the very premature babies, along with further
impairment of the regulation of cerebral blood flow and blood pressure
subsequent to perinatal distress and resuscitation manoeuvre.35
The immaturity of the respiratory system and a decreased production of
surfactant can justify the associations between mortality from BPD and
birthweight 51000 g and unknown body temperature at admission.36 A
decreased risk of dying from BPD was found for female when compared with
male infants; this is in agreement with previous studies, which identified male sex
as a risk factor for the disease.37
Quite surprisingly, no relationship could be demonstrated between absence of
a full course of prenatal steroids and mortality from respiratory problems and/or
IVH. The point estimate of the odds ratio when the two groups of causes of death
were considered together was 2.80; the wide 95% CI (0.869.11) was probably
because of the low number of infants who received the treatment, only 10% of the
total, a frequency similar to that found in other multicentre studies.4,5
In spite of the wide range of values of mortality in the eight NICUs, no
differences in cause-specific death rates among them could be demonstrated, with
the exception of the higher mortality from infections in NICU H. This means
that most of the variability in crude death rates is accounted for by differences
in the populations served and in the clinical conditions of the infants at the
moment of admission.
In conclusion, the good agreement between the two coders and the clinical
correlates of the groups of the causes of death employed, make the classification
adopted in this study a practical instrument for gaining information for the

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
54 C. Corchia et al.

evaluation of care for VLBWIs. It could be a useful adjunct to clinical risk indexes
in assessing the performance of neonatal intensive care units,38,39 especially when
used for NICU certification. From this study, the improvement of perinatal care in
the delivery room, the prompt stabilisation of vital functions following delivery
and during transportation, and a more frequent use of antenatal steroids for the
induction of lung maturity appear as important points for decreasing mortality in
VLBWIs. Finally, the prevention and control of infectious diseases demand strict
surveillance and proper interventions, because of the increasing risk of a neonatal
intensive care patient acquiring nosocomial infections.

Acknowledgements
This study was partly supported by a MURST 40% grant from Sassari University.

References
1 Yu VYH, Watkins A, Bajuk B. Neonatal and postneonatal mortality in very low
birthweight infants. Archives of Disease in Childhood 1984; 59:987999.
2 Aylward GP, Pfeiffer SI, Wright A, Verhulst SJ. Outcome studies of low birth weight
infants published in the last decade: a metaanalysis. Journal of Pediatrics 1989; 111:515520.
3 Hack M. The outcome of neonatal intensive care. In: Care of the High Risk Neonates.
Editors: Klaus MH, Fanaroff AA. Philadelphia: Saunders, 1993, 456463.
4 Hack M, Horbar JD, Malloy MH, Tyson JE, Wright E, Wright L. Very low birth weight
outcomes of the National Institute of Child Health and Human Development neonatal
network. Pediatrics 1991; 87:587597.
5 The Investigators of the VermontOxford Trials Network Database Project. The
VermonOxford Trials Network: very low birth weight outcomes for 1990. Pediatrics 1993;
91:540545.
6 Autio-Harmainen H, Rapola J, Hoppu K, Osterlund K. Causes of neonatal deaths in a
pediatric hospital neonatal unit. Acta Paediatrica Scandinavica 1983; 72:333337.
7 Verloove-Vanhorick SP, Verwey RA. Project on preterm and small for gestational age
infants in the Netherlands 1983 (thesis). The Netherlands: State University Leiden, 1987.
8 Karlberg P, Priolisi A, Landstrom T, Selstam U. Clinical analysis of causes of death
with emphasis on perinatal mortality. Monographs in Paediatrics 1977; 9:86120.
9 Wigglesworth J. Monitoring perinatal mortality: a pathophysiological approach. Lancet
1980; ii:684686.
10 Keeling JW, MacGillivray I, Golding J, Wigglesworth J, Berry J, Dunn PM.
Classification of perinatal death. Archives of Disease in Childhood 1989; 64:13451351.
11 Cole S, Hartford RB, Bergsjio P, McCarthy B. International collaborative effort (ICE) on
birth weight, plurality, perinatal, and infant mortality. III: A method of grouping
underlying causes of infant death to aid international comparisons. Acta Obstetrica
Gynecologica Scandinavica 1989; 68:113117.
12 Dollfus C, Patetta M, Siegel E, Cross AW. Infant mortality: a practical approach to the
analysis of the leading causes of death and risk factors. Pediatrics 1990; 86:176183.
13 Golding J. Epidemiology of fetal and neonatal death. In: Fetal and neonatal pathology.
Editor: Keeling JW. London: Springer-Verlag, 1993; 165181.

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA

50
Table 3. Cumulative cause-specific net probabilities of death (%) at 2 years of age according to some characteristics of the study subjects
Causes of death

C. Corchia et al.
Respiratory
Respiratory problems
Asphyxia problems IVH + IVH Infections BPD Others
Sex
# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456

Females 1.6 14.3 6.6 19.9 6.0 2.6 2.4

Males 1.8 16.7 8.4 23.7 8.6 4.9 3.4
Birthweight
10001499 g 1.2 5.9 4.9 10.5 4.1 2.0 2.5
500999 g 3.0 38.8*** 15.5*** 48.5*** 19.2*** 11.1*** 4.2
Gestational age
30+ weeks 1.0 3.7 1.0 4.6 4.2 1.9 2.7
530 weeks 2.5 27.3*** 14.9*** 38.2*** 11.4** 6.2* 3.1
Place of birth
Inborn 2.2 15.0 4.8 19.1 9.3 3.1 1.7
Outborn 1.2 15.8 10.0* 24.3 5.4 4.1 4.1
NICU
A 0 3.2 3.3 6.4 3.2 3.6 0
B 2.5 5.4 6.6 11.6 9.0 4.6 1.3
C 2.3 16.6 2.3 18.6 2.8 0 11.8
D 0.8 14.6 10.6 23.7 7.0 4.6 1.2
E 3.2 21.5** 3.4** 24.1 6.3** 0 3.6
F 0 10.6 17.8 26.6 5.5 4.3 2.4
G 0 20.8 5.9 25.5 4.0 4.6 3.9
H 5.7 22.3 2.4 24.2 24.2 8.1 0
*P50.05; **P50.01; ***P50.001.
 Paper 14 DATA
Causes of death in VLBW infants 55

14 Pugliese A, D'Argenio P, Arsieri R, Russo R, Palumbo F. Cause di mortalita neonatale

in Italia. Rivista Italiana di Pediatria (Italian Journal of Pediatrics) 1993; 19:241249.
15 Alberman E, Botting B, Blatchley N, Twidell A. A new hierarchical classification of
causes of infant deaths in England and Wales. Archives of Disease in Childhood 1994; 70:403
409.
16 de Vonderweid U, Spagnolo A, Corchia C, Chiandotto V, Chiappe S, Chiappe F et al.
Italian multicentre study on very low-birth-weight babies. Neonatal mortality and two-year
outcome. Acta Paediatrica 1994; 83:391396.
17 World Health Organization. Manual of the International Statistical classification of Diseases,
Injuries, and Causes of Death. Geneva: WHO, 1977.
18 Northway WH. Bronchopulmonary dysplasia: twenty-five years later. Pediatrics 1992;
89:969973.
19 Dixon WJ, Brown MB. BMDP Statistical Software. Berkeley: University of California
Press, 1985.
20 Kahn HA, Sempos TC. Statistical Methods in Epidemiology. New York: Oxford
University Press, 1989.
21 Cox DR. Regression models and life tables. Journal of the Royal Statistical Society Series B
1972; 34:187220.
22 Horbar JD, McAuliffe TL, Adler SM, Albersheim S, Cassady G, Edwards W et al.
Variability in 28-day outcomes for very low birth weight infants: an analysis of 11 neonatal
intensive care units. Pediatrics 1988; 82:554559.
23 Pharoah POD, Alberman ED. Annual statistical review. Archives of Disease in Childhood
1990; 65:147151.
24 Rothman KJ. Modern Epidemiology. Boston: Little, Brown & Company, 1986.
25 Stoll BJ. Epidemiology of necrotizing enterocolitis. Clinics in Perinatology 1994; 21:205
218.
26 Leviton A, Van Marter L, Kuban KCK. Respiratory distress syndrome and intracranial
hemorrhage: cause of association? Inferences from surfactant clinical trials. Pediatrics 1989;
84:915922.
27 Paneth N, Pinto-Martin J, Gardiner J, Wallenstein S, Katsikiotis V, Hegyi T et al.
Incidence and timing of germinal matrix/intraventricular hemorrhage in low birth weight
infants. American Journal of Epidemiology 1993; 137:11671176.
28 Pinto J, Paneth N, Kazam E, Kairam R, Wallenstein S, Rose W et al. Interobserver
variability in neonatal cranial ultrasonography. Paediatric and Perinatal Epidemiology 1988;
2:4358.
29 Hope PL, Gould SJ, Howard S, Hamilton PA, de L Costello AM, Reynolds EOR.
Precision of ultrasound diagnosis of pathologically verified lesions in the brains of very
preterm infants. Developmental Medicine and Child Neurology 1988; 30:457471.
30 Jason JM. Infectious disease-related deaths of low birth weight infants, United States,
1968 to 1982. Pediatrics 1989; 84:296303.
31 Tangermann RH, McCarthy BJ, Schmidt E. Comparing cause-specific infant mortality
in a West German State and the United States of America. Acta Paediatrica Scandinavica 1990;
79:11431149.
32 Alo CJ, Howe HL, Nelson MR. Birth-weight-specific infant mortality risks and leading
causes of death. Illinois, 19801989. American Journal of Diseases of Children 1993; 147:10851089.
33 Beckwitt Turkel S, Sims ME, Guttenberg ME. Postponed neonatal death in the
premature infant. American Journal of Diseases of Children 1986; 140:576579.
34 Bohin S, Field DJ. The epidemiology of neonatal respiratory disease. Early Human
Development 1994; 37:7390.

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456
 Paper 14 DATA
50 C. Corchia et al.

35 Volpe JJ. Intraventricular hemorrhage and brain injury in the premature infant. Clinics
in Perinatology 1989; 16:361386.
36 Thurlbeck WM. Prematurity and developing lung. Clinics in Perinatology 1992; 19:497
519.
37 Kennedy KA. Epidemiology of acute and chronic lung injury. Seminars in Perinatology
1993; 17:247252.
38 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; 91:969975.
39 The International Neonatal Network. The CRIB (clinical risk index for babies) score: a
tool for assessing initial risk and comparing performance of neonatal intensive care units.
Lancet 1993; 342:193198.

Book review

Handbook of Biological Effects of Electromagnetic Fields. 2nd Edition. Edited

by Charles Polk and Elliot Postow. Boca Raton: CRC Press, 1996, pp. 618, 99

This is a most useful text for those of us who are trying to investigate epidemiolog-
ically the possible effects of electromagnetic fields (EMFs). It outlines the different
types of field, and the doses at varying distances from them, including electric and
magnetic fields, microwave and radio frequency and their known biological
effects. A useful chapter by Richard Stevens summarises the epidemi-ological
studies on cancer and reproduction, including those on video display terminals,
and concludes that `In spite of continued questions about the biological rationale
and plausibility of EMF harming human health, the suggestive epidemi-ological
data cannot and should not be dismissed.' Further high-quality studies are clearly
needed.

JEAN GOLDING

# 1997 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 11, 4456