  |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Journal of the New Zealand Medical Association,
16-December-2005, Vol 118 No 1227
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Hong Kong case-control study of sudden unexpected infant
death
Tony Nelson, Ka-Fai To, Yuk-Ying Wong, Jim Dickinson,
Kai-chow Choi, Ly-Mee Yu, Yvonne Ou, Chun-Bong Chow, Eric Wong, Nelson Tang,
Magnus Hjelm, Lawrence Chen
Sudden infant death syndrome (SIDS) was originally defined
as the sudden death of an infant or young child, which is unexpected by history,
and in which a thorough postmortem examination fails to demonstrate an adequate
cause of death.1 Other definitions have been suggested,2 and recently a new
definitional and diagnostic approach has been proposed.3
At a time when New Zealand was reporting very high rates of
SIDS (3.6-7.4/1000 live births in 1986), Hong Kong had very low rates (0.3/1000
lives births in 1987).4,5 In the United States and Europe during the same
period, reported rates were in the region of 1 to 2/1000 live births.6
Subsequent research studies in a range of countries have
shown that certain childcare practices are associated with SIDS. Placing the
baby to sleep on the front (prone) has been identified to be the most important
risk factor; but other factors include smoking by the mother or father, the baby
sleeping in the same bed as someone else (bedsharing), baby sleeping in a room
separate from the parents, and baby not using a pacifier (dummy).
A striking reduction of SIDS incidence of more than 50% has
been observed in many countries after parents were advised not to place their
babies on their front to sleep.7 However the prone sleep position alone does not
explain widely disparate SIDS rates, and interaction with other childcare
practices appears to be important.4 This study aimed to identify factors
associated with an increased risk of SIDS in Hong Kong and to document the
causes of all unexpected child deaths under 2 years of age during a 4-year
period.
MethodsThis study covered the period 1 February 1999 to 31
January 2003 and was in two main parts: a case-control study and a descriptive
study.
Case-control
study—This compared information of those children who died (cases)
with a group of children who did not die (controls). Due to data privacy
concerns, cases could only be identified and contacted with the help of the
mortuary staff.
All families of unexpected child deaths were required
to attend the mortuary for identification purposes, and at this stage, the
mortuary staff sought agreement from families of suspected SIDS deaths for the
research nurse to contact them. At the same time, arrangements were made (when
possible) for a paediatric pathologist to perform the postmortem examination
using a modified international standardised protocol.8 Potential cases were
missed when the mortuary staff failed to contact the research nurse. The
Beckwith definition was used to decide whether deaths were due to SIDS.1
To ensure that the controls represented all children
born in Hong Kong, they were selected from both government and private hospitals
in proportion to the anticipated births at these hospitals. The majority of
hospitals participated in this process (representing 90% of anticipated births).
To obtain controls that were matched to the expected ages of the deaths, dates
of interview (“nominated date”) were randomly selected using
computer-generated numbers for all 1460 days of the study.
The age at interview and the “nominated
time” were randomly selected according to anticipated age distribution and
estimated time of death distribution of the deaths (based on previous New
Zealand and Hong Kong data). A date of birth was then calculated from this data.
The randomly allocated “nominated time” was during one of four time
periods: morning routine (6am–12 midday); afternoon routine (12
midday–6pm); evening routine (6pm–12 midnight); night-time routine
(12 midnight–6am). The reference sleep was the longest sleep during this
nominated time period.
Both cases and controls
were interviewed at home (or at another mutually agreed venue) using a
questionnaire similar to the one used by European Concerted Action on SIDS
study. The Hong Kong Observatory provided meteorological data related to the
estimated time of death or reference sleep.
The study was approved by
the Clinical Research Ethics Committee and Survey Ethics Committee of the
Chinese University of Hong Kong.
Descriptive
study—This involved obtaining the details of all deaths of children
under the age of 2 years that were reported to the coroner during the study
period. The new Hong Kong Coroners Ordinance was published in April 1997 and
implemented in 1998. The coroner decides whether or not a Death Report is
required. If a Death Report is requested, a detailed investigation is undertaken
by the Police. If the coroner considers that there is no likelihood of any
criminal or negligent acts, then a Death Report will not usually be requested
and an autopsy may not be required.
Based on the details available, deaths were
provisionally classified into four groups: A = not SIDS; B = probably not SIDS;
C = possibly SIDS; D = SIDS. Death reports and/or related information were then
viewed in detail by the correspondence author (T Nelson) for the B, C, and
D groups.9
Death certificates of all deaths not reported to the
coroner for the period February 1999 to December 1999 were also viewed to assess
the possibility of any SIDS deaths bypassing the coroner reporting system.
Statistical
analysis—For the case-control study, EpiInfo data entry and
data-checking programmes based on those used in the ECAS study were used.
Univariate analyses were performed using Chi-squared tests, Fisher’s exact
tests or Mann-Whitney tests as appropriate. Logistic regressions (using exact
method as appropriate) were employed to control for sex and socioeconomic
status. The confidence intervals for the incidence of SIDS were calculated using
Clopper-Pearson method. All statistical analyses, odds ratios, and their
associated 95% confidence intervals calculation were done using SPSS 11.5,
StatXact-4 (version 4.0.1 Cytel Software Corporation) and LogXact-4 for Windows
(version 4.1, Cytel Software Corporation). All tests were two-sided with
significance level at 0.05.
ResultsCase-control
study—Families of 18 children who had unexpected deaths were
interviewed, but two of the deaths were subsequently shown not be due to SIDS
(these 2 deaths were due to ornithine carboxylase deficiency and acute
bronchiolitis respectively). One paediatric pathologist (KFT) performed the
postmortem examination of 16 of these 18 deaths; 11 of these 16 deaths were
diagnosed as being consistent with SIDS, 2 as consistent with SIDS but with
marked fatty change of the liver (suggesting potential underlying inborn error
of metabolism), 1 as consistent with SIDS with an underlying obstructive
uropathy, 1 as interstitial pneumonia, and 1 as a urea cycle defect (ornithine
carboxylase deficiency). This latter case was excluded from the case-control
analysis. Detailed discussion of this case diagnosed as interstitial pneumonia
concluded that SIDS could not be excluded and hence this case was retained in
the case-control analysis.
Two of the 16 autopsies were undertaken by a forensic
pathologist of the Department of Health and neither death was diagnosed as SIDS.
However one case, diagnosed as pneumonia, was considered to be compatible with
SIDS after careful review of the history and autopsy report. The other case, a
19-month-old child diagnosed with acute bronchiolitis, was excluded. The
diagnoses for the 16 children included in the case-control analysis were thus
SIDS (14), interstitial pneumonia (1), and pneumonia (1).
Including the latter two cases, 7 of the 16 SIDS deaths
were considered to have some atypical features on history (e.g. age over 1 year)
or presence of other significant findings on autopsy (e.g. interstitial
pneumonitis) that may not be considered as SIDS using stricter definitions;2;3
thus they were classified as group C.9
The majority of case interviews were undertaken at the
case-family’s home; some interviews took place in hospital or at another
neutral venue. Although we attempted to interview case-families within 3 weeks
of the death, this was not always possible and the longest delay was 3 months.
Of 268 planned controls, 223 were interviewed (147 were the first control
selected and 76 were one of the 3 potential backup controls approached during
the postpartum period).
The majority of controls (197/223) were initially approached
during the postnatal period and 26 were identified retrospectively from the
labour ward delivery records. Most of the interviews were carried out within 3
days of the nominated date with one or both parents, or with the caregiver. The
majority 90% (201/223) of control interviews took place in the mother’s
home and the remainder at another venue such as the mother’s office, park,
or restaurant. There was no significant difference between the mean age of
infants who died (20.7 weeks at death) and controls (19.4 weeks at nominated
date), thus showing that the method of matching for age when selecting controls,
worked well.
Seventy-three percent of the controls were selected from
government hospitals and 27% from private hospitals, which was very similar to
the anticipated distribution of total births at these hospitals. Significant
differences were noted for child’s gender, parents’ marital and
socioeconomic status, mother’s age, and parents’ occupation (Table
1). Although not statistically significant, there was a trend to lower mean
birth weight for infants who died (2980 gms) than controls (3219 gms), and for
lower gestational age of infants who died (37.4) than controls (39.3).
Table 1. Demographic variables for cases and
controls
*p<0.05;
**p<0.01.
Univariate and logistic regression analyses controlling for
sex and socioeconomic status showed that recognised SIDS risk factors such as
sleep position, smoking by mother and father, and bedsharing were significant in
this study (Table 2). However bedsharing was only important when the infants
shared a bed with someone other than the parents. Smoking was only a consistent
risk factor for those mothers who were smoking during the 2nd trimester.
Amount of the last feed and illness in the last week were
also significant risk factors. The latter variable combined responses from a
number of variables related to the presence of symptoms, medical attention and
medications being taken between 7 and 1 day before death/reference sleep or in
the last 24 hours.
These included diagnosis with gastroenteritis; bronchiolitis
or pneumonia; visited doctor because of illness; admitted to hospital because of
illness; symptoms of fever; cough, cold, earache or vomiting; change in bowel
motion or change in colour or smell of urine; or given antipyretic, cough
mixture, or antibiotic. Smoking by the partner, use of an adult pillow, and
outside temperature were only significant risk factors on univariate
analysis.
The analyses were repeated on the subgroup of “typical
SIDS” (under 1 year of age with no atypical features n=9, and controls
under 1 year of age n=204), and (despite the small numbers) mother smokes,
bedsharing, head and body covered, and amount of last feed remained significant
after adjustment for sex and socioeconomic status.
Sleep position was not significant (only 1 of the 9 infants
was in this position when left and when found). Bedsharing was only significant
for the combined group (parents and other people). Illness in the last week was
significant on univariate analysis but not after adjustment. Smoking by the
partner, use of an adult pillow, and outside temperature were not significant on
univariate analysis (details available from the author).
Descriptive
study—183 unexpected infant deaths were reported to the coroner
during the 4-year study period and the coroner requested death reports for 52%
(95/183) of these deaths. There were 94 deaths (78 with death reports and 16
without)—provisionally classified as group B, C, or D—that were
reviewed in detail by the correspondence author.
The causes of death were then sub-categorised from two
perspectives: from the diagnosis recorded on the death report or other coroner
records; and from the diagnosis considered to be most appropriate after a
detailed review (Table 3). This showed that based on the coroner diagnosis,
there were 18 SIDS—but (based on the review) there were 33 potential SIDS
deaths. One death could not be classified due to inadequate information. The 17
death reports that were not checked had clear diagnoses such as congenital heart
disease or other illness (11), or the death report could not be located (1), or
was still pending (5). There was sufficient information on 3 of the latter to
confirm that the diagnosis was not SIDS.
Other coroner information was checked for the 18% (16/88) of
deaths with no Death Reports but with a diagnosis that could be a potential SIDS
(e.g. pneumonia, asphyxia). The 15 deaths that had not been classified by the
coroner as SIDS (but were nevertheless considered to be potential SIDS upon
review) had been classified as pneumonia, bronchopneumonia, aspiration pneumonia
or interstitial pneumonitis (8); asphyxia due to wedging (1); and unknown,
unascertained, or pending (6).
Review of 119 death certificates of deaths not reported to
the coroner during the period 1 February 1999 to 31 December 1999 showed no
evidence that potential SIDS deaths were bypassing the Hong Kong coronial
system.
Calculation of SIDS
incidence—During the 4-year period (1999–2002) there were
206,611 live births in Hong Kong (48,219 in 1999; 54,134 in 2000; 51,281 in
2001; 52,977 in 2002). Therefore, based on the maximum of 33 potential SIDS
rather than the 18 deaths officially classified as SIDS, the incidence of SIDS
in Hong Kong for the period 1999–2002 was estimated to be 0.16/1000 (95%
CI: 0.11–0.22/1000). It was noted that 7 of the 33 potential SIDS infants
(21%) were non-Chinese.
The 2001 Census showed that 6.2% of children under 5 years
and 5.1% of the total population was non-Chinese thus indicating that the rate
of SIDS in the Chinese population may be even lower. The official mortality
rates (per 1000 live births) for 1999, 2000, 2001, 2002 were 3.1, 3.0, 2.6, 2.3
for infant mortality and 1.4, 1.2, 1.0, 1.0 for postneonatal mortality
respectively (data from Hong Kong Census Department).
Table 2. Variables showing univariate- and sex and
socicoeconomic-adjusted relationship
*p<0.05; **p<0.01;
†ORu,:
univariate odds ratio;
‡ORA:
sex and socieconomic status-adjusted odds ratio;
§19.5
and 27°C were respectively the 25th and 75th percentiles of the temperature
readings provided by the Hong Kong Observatory for the estimated time of death
(cases) or nominated time (controls).
Table 3. Causes of 183 child deaths under 2 years of
age reported to the Hong Kong coroner during 1 February 1999 to 31 January
2003 classified according to (1) primary diagnosis from the coroner’s
death report, autopsy report or other information and (2) after detailed review
of these reports by one of us (T Nelson)
*4/11 multiple injury; **6/7
unexplained cerebral or subdural haematomas.
DiscussionThe incidence of SIDS in Hong Kong has fallen even further
from the low value of 0.3/1000 live births (95% CI: 0.18–0.46/1000) in
1987,4 to 0.16/1000 (95% CI: 0.11–0.22/1000) in 1999–2002. Review of
the coroner records showed that (when the study pathologist did not undertake
the postmortem examination) forensic or hospital pathologists were much more
likely to use the terms ‘unknown’ and ‘unascertained’.
This reflects an inherent problem of SIDS studies, when both under-diagnosis and
over-diagnosis can be potential problems. We erred on the side of over-diagnosis
when estimating SIDS incidence and if stricter definitions had been used, the
incidence would have been even lower.3
The case-control comparison shows that, as elsewhere, prone
sleeping and parental smoking are key risk factors, with high odds ratios.
However, 12 of 16 babies who died had been placed on their back, emphasising
that other factors must also be important. Bedsharing is a common behaviour in
Hong Kong: 24% of controls did so. Bedsharing with someone other than parent had
a sex and socioeconomic status-adjusted odds ratio of 4.9 (1.2–20.2),
hence suggesting that this practice appeared to be hazardous.
In Hong Kong, where SIDS incidence is low but bedsharing is
common, it can be assumed that methods of bedsharing are generally
“safe”. However our data suggests that even within this culture,
bedsharing is not always a safe practice and that parents and other caregivers
need to be aware of potential risks. Finding babies with their heads covered was
also common, as noted in other studies. The issue of diagnosis of SIDS was
addressed by repeating the analysis on a subgroup of “typical SIDS”
and controls less than 1 year old. This showed generally similar results.
Because of the lower rates than anticipated, our study had
to be extended from the originally planned 3 years to 4 years. We still obtained
small numbers of cases, which reduces the power to demonstrate differences, and
to conduct more extensive studies of the relationship of SIDS to risk factors.
We could only do follow-up interviews for those deaths that
were initially suspected to be SIDS and reported to us by the mortuary staff.
The response rate for controls was about 55% for the first control, but 90%
overall. Loss of controls occurred mainly because of moving to unknown addresses
or subsequent reluctance to be interviewed. It is not known to what extent
failure to interview the first selected control might have influenced the
results.
The time lag between the deaths and our interview was often
long, much longer than for the controls. This lag usually occurred because of
delays in arranging a time for interview that was acceptable to the families.
However, while the “reference sleep” would be only one among many
for controls, and therefore easy to forget, it was much more salient for the
family of the dead infants, so details would be anticipated to more readily
recalled. This is anticipated to reduce the potential for recall bias, although
certain variables, such as amount of last feed and illness of the baby during
the last week, could be more prone to such bias.
The rates of SIDS in Chinese populations are lower than in
Western countries,4,10,11 but this is one of the lowest reported. Of interest
was the greater proportion of non-Chinese in our cases (21%) relative to the
Census data, which showed that 5.1% of the population are non-Chinese.
Comparison of our attribution of cause with the official data shows that without
special care it is easy to miss potential cases, and the difference may be about
50%. Especially in populations with low rates of SIDS such as Hong Kong Chinese,
it is important to check for metabolic diseases (such as the ornithine
carboxylase deficiency we found), since even a few such cases could distort the
apparent rates.
Special attention must also be paid to those deaths
diagnosed as respiratory illness, unascertained or unknown.12 While the latter
two labels may reassure the pathologist that an unnatural cause remains an
option, these terms may carry negative connotations for bereaved parents, and
leave a lack of closure. By contrast, though the cause of SIDS is unknown, it
leaves the understanding that the cause is natural, and not a reason for
guilt.
The Hong Kong coronial system does not routinely have access
to medical professionals, yet coroners decide whether a death report and
investigation is required. We found several discrepancies in cause of death, and
these included some with important legal and medical implications, such as
suspected “shaken baby” syndrome. Therefore, a more formal process
is possibly needed to review all child deaths, whether reported to the coroner
or not. Such systems have been established elsewhere, such as the Confidential
Enquiry into Stillbirths and Deaths in Infancy (CESDI) established in 1992 in
the UK,13 or the approach of the Victorian Institute of Preventive Medicine
(VIFM) in Melbourne, Australia where a Scientific Officer investigates deaths
suspected to be SIDS, and the death scene information assists the pathologists
in determining the cause of death.14
A recent report of the Royal College of Pathologists and
Royal College of Paediatrics and Child Health on Sudden Unexpected Death in
Infancy (SUDI) highlights that parents who have suffered this terrible tragedy
have a right to a diagnosis, and that all SUDI postmortems should follow
recommended protocols to optimise the chances that the cause of death can be
identified.15
Parents also need sensitive support and follow-up. The
present study has shown that many of the flaws in the investigation and
management of SUDI, identified by the United Kingdom report, are also occurring
in Hong Kong. In the light of the findings of our study and publication of the
working group recommendations, it will be important for Hong Kong to take steps
to standardise the investigation and management of these deaths. Indeed, there
is clearly a need to establish a child mortality review system in Hong Kong to
systematically monitor all child deaths and to provide regular feedback to the
public, health authorities, and professionals.
In the absence of good understanding of the pathology of
SIDS, studies such as ours provide evidence for recommending changes in
child-rearing practices. The consistent relationship between parental smoking
and SIDS adds another reason to quit smoking. Pregnancy and the birth of a child
may be important events that could encourage parents to stop or reduce
smoking.16 While bedsharing is relatively common practice in Hong Kong, we must
inform parents that this is potentially unsafe under certain circumstances. In
addition, parents should be aware of the risks of bedclothing covering the
baby’s head.
This estimate of the even lower rate of SIDS in Hong Kong
after 10 years shows a need to monitor such rates regularly, with the hope of
better understanding this disease. We cannot at this point be sure whether there
has been substantial change in Hong Kong child-rearing practices, or whether
this may simply be the result of improved living conditions in the interim. Our
data has shown that when SIDS rates are very low, these rates may be distorted
with the use of labels such as unknown or unascertained.12
Authors:
E Anthony S Nelson, Professor in Paediatrics1; Ka -Fai To, Professor in
Anatomical and Cellular Pathology2; Yuk-Ying Wong, Research Nurse1; James A
Dickinson, formerly Professor of Family Medicine3; Kai-chow Choi, Research
Associate4; Ly-Mee Yu, Computer Officer4; Yvonne Ou, Private Practitioner5;
Chun-Bong Chow, Consultant Paediatrician6; Eric Wong, Computer Officer4; Nelson
L S Tang, Professor in Chemical Pathology7; Magnus Hjelm, formerly Professor of
Chemical Pathology7; Lawrence Chen, Psychologist5
Affiliations:
1Department of Paediatrics, 2Department of Anatomical and Cellular
Pathology, 3Department of Community & Family Medicine, 4Centre for
Epidemiology and Biostatistics, and 7Department of Chemical Pathology, The
Chinese University of Hong Kong; 5Private Practice;6Department of Paediatrics
and Adolescent Medicine, Princess Margaret Hospital; Hong Kong SAR, China
Acknowledgements:
This research project was funded by grant number HSRC821009 from the Health
Services Research Committee (Hong Kong). Laboratory work to investigate possible
inborn errors of metabolism was supported by a grant from the SK Yee Medical
Foundation.
We also thank nursing staff at the Hospital Authority and
private hospitals who helped with the selection and recruitment of controls. In
addition, Dr HK Mong and his staff of the Forensic Pathology Service of the
Department of Health provided support throughout the study; coroners Ian
Thomson, Paul Kelly, Andrew Chan, and Peter White provided support and advice;
Ms YYL Leung and Ms LF Wong of the Coroners Court facilitated review of the
coroner records; Peter Hunt, Angela Lau, and MR Dermaid-Groves of the Hong Kong
Police advised on introducing a standardised protocol for all unexpected infant
deaths undergoing Police investigation; Bob Carpenter and the European Concerted
Action against SIDS (ECAS) study group provided questionnaires and details of
the ECAS study design; and Ed Mitchell provided details of the New Zealand
Multicentre Cot Death Study.
Correspondence: Prof
EAS Nelson, Department of Paediatrics, 6/F Clinical Sciences Building, Prince of
Wales Hospital, Shatin, Hong Kong SAR. Fax: +852 26360020; email: tony-nelson@cuhk.edu.hk.
References:
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|