Iyasu S, Randall LL, Welty TK, Hsia J, Kinney HC, Mandell F, McClain M, Randall B, Habbe D, Wilson H, Willinger M. Risk Factors for Sudden Infant Death Syndrome Among Northern Plains Indians. JAMA. 2002;288(21):2717-2723. doi:10.1001/jama.288.21.2717
Author Affiliations: Division of Reproductive Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Ga (Drs Iyasu and Hsia and Ms Randall); Aberdeen Area Indian Health Service, Rapid City, SD (Dr Welty); Children's Hospital Boston, Harvard Medical School, Boston, Mass (Dr Kinney); Harvard Medical School, Boston, Mass (Dr Mandell); Massachusetts SIDS Center, Boston Medical Center, Boston (Ms McClain); LCM Pathologists, PC, Sioux Falls, SD (Dr Randall); Clinical Laboratory of the Black Hills, Rapid City, SD (Dr Habbe); Department of Pathology, Providence Memorial Hospital, El Paso, Tex (Dr Wilson); and National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Md (Dr Willinger). Dr Iyasu is now with the Division of Pediatrics, Office of Counter Terrorism and Pediatric Drug Development, Center for Drug Evaluation and Research, Food and Drug Administration, Rockville, Md.
Context Sudden infant death syndrome (SIDS) is a leading cause of postneonatal
mortality among American Indians, a group whose infant death rate is consistently
above the US national average.
Objective To determine prenatal and postnatal risk factors for SIDS among American
Design, Setting, and Participants Population-based case-control study of 33 SIDS infants and 66 matched
living controls among American Indians in South Dakota, North Dakota, Nebraska,
and Iowa enrolled from December 1992 to November 1996 and investigated using
standardized parental interview, medical record abstraction, autopsy protocol,
and infant death review.
Main Outcome Measures Association of SIDS with maternal socioeconomic and behavioral factors,
health care utilization, and infant care practices.
Results The proportions of case and control infants who were usually placed
prone to sleep (15.2% and 13.6%, respectively), who shared a bed with parents
(59.4% and 55.4%), or whose mothers smoked during pregnancy (69.7% and 54.6%)
were similar. However, mothers of 72.7% of case infants and 45.5% of control
infants engaged in binge drinking during pregnancy. Conditional logistic regression
revealed significant associations between SIDS and 2 or more layers of clothing
on the infant (adjusted odds ratio [aOR], 6.2; 95% confidence interval [CI],
1.4-26.5), any visits by a public health nurse (aOR, 0.2; 95% CI, 0.1-0.8),
periconceptional maternal alcohol use (aOR, 6.2; 95% CI, 1.6-23.3), and maternal
first-trimester binge drinking (aOR, 8.2; 95% CI, 1.9-35.3).
Conclusions Public health nurse visits, maternal alcohol use during the periconceptional
period and first trimester, and layers of clothing are important risk factors
for SIDS among Northern Plains Indians. Strengthening public health nurse
visiting programs and programs to reduce alcohol consumption among women of
childbearing age could potentially reduce the high rate of SIDS.
The infant mortality rate among American Indians is consistently above
the national average, primarily due to a higher death rate during the postneonatal
period (28-364 days). Sudden infant death syndrome (SIDS) is the leading cause
of postneonatal mortality, and until 1997, was also the leading cause of infant
mortality among American Indians.1- 3 SIDS
is defined as the sudden death of an infant younger than 1 year that remains
unexplained after a thorough case investigation, including the performance
of a complete autopsy, an examination of the death scene, and a review of
the infant's clinical history.4 In 1999, the
SIDS rate was 1.5 per 1000 live births for American Indian infants and 0.7
per 1000 live births for all races combined.5
The Aberdeen Area Indian Health Service (AAIHS), which serves Indian
communities in North Dakota, South Dakota, Nebraska, and Iowa, has the highest
rate of infant mortality among the Indian Health Service (IHS) areas.6,7 SIDS is the leading cause of infant
deaths in the Aberdeen Area, accounting for more than one fourth of the infant
deaths and more than half of the postneonatal deaths. Although there was a
42% decline in the SIDS rates for all IHS regions, from 2.77 per 1000 live
births in 1992-19946 to 1.61 per 1000 in 1996-1998,7 the rate in the Aberdeen Area has remained relatively
constant: 3.66, 3.55, and 3.46 per 1000 live births for 1992-1994, 1994-1996,
and 1996-1998, respectively.6- 8
One of the objectives of the Aberdeen Area Infant Mortality Study (AAIMS),
which was conducted in collaboration with the Aberdeen Area Tribal Chairman's
Health Board, was to determine prenatal and postnatal risk and protective
factors for SIDS among Northern Plains American Indian infants.
From December 1, 1992, through November 30, 1996, a case-control study
was conducted in the AAIHS. Nine tribes and 1 urban American Indian community,
constituting two thirds of the service area population, participated. The
methodology has been previously described9 and
is summarized below.
Tribal resolutions of support for the study were obtained. After investigators
obtained input from tribal spiritual leaders, the AAIHS and the National IHS
institutional review boards approved the study protocol. Informed consent
was obtained from a parent or legal guardian to conduct the parental interviews
and review the medical records.
Eligible cases were American Indian infants residing on or near reservations
or participating communities who died before 1 year of age, excluding infants
who died during their delivery hospitalization. Case infants were classified
as American Indian if either of their parents was an enrolled tribal member
or if they were eligible for care at an IHS facility. Reports from public
health nurses (PHNs), medical record department staff, emergency staff, members
of the Perinatal Infant Mortality Review (PIMR) Committee, death certificates
on American Indian infants, and obituaries in local newspapers were reviewed
to identify cases.9
Two living control infants were matched to each case infant by postnatal
age and community or reservation of residence. For age matching, AAIMS investigators
used a list of eligible American Indian live-born infants by birth date and
selected those born just before and just after each case. There were 3 refusals
among the controls, and they were replaced with the next eligible infant with
the closest birth date to the case.
All data were collected retrospectively. Two American Indian nurse interviewers
conducted parental interviews using an eighth-grade level, culturally competent
questionnaire that solicited information about demographic and socioeconomic
factors; maternal medical and obstetric history; neonatal history; and a wide
range of potential risk factors including fetal and infant exposures.
Mothers were asked about use of cigarettes, alcohol, and illicit drugs
during the 3 months prior to pregnancy, during each trimester, and during
the postpartum period. Similarly, mothers were asked about binge drinking
(≥5 drinks in 1 sitting). The alcohol use questions had been used in the
clinical setting prior to the study and were subsequently formally validated
in this population.10
Standardized autopsy protocols were used. The majority of autopsies
were performed by 3 pathologists whose jurisdiction included the study area
and who participated in the study as members of the steering committee and
the PIMR. Standard death scene investigation protocols were developed and
tribal and county coroners were trained to use them. Medical records for cases
and controls were abstracted and reviewed.
The PIMR committee determined the cause of death for all infants after
reviewing all available information. The committee confirmed a diagnosis of
SIDS only if an autopsy was performed and sufficient information from the
autopsy and scene supported the diagnosis. In the absence of an autopsy, or
if the cause of death was uncertain, the committee assigned a diagnosis of
To assess risk factors for SIDS, we examined characteristics of case
and control infants using the χ2 test for categorical variables
and the 2-tailed t tests for continuous variables
and performed a matched conditional logistic regression using the proportional
hazards regression procedure in the Statistical Analysis System.11 We
performed multivariate analyses using likelihood methods to build models that
included significant factors obtained from univariate analyses, while taking
into account biological plausibility. We therefore modeled starting with the
4 most significant independent variables and removing or adding the other
variables to a current model based on −2 log likelihood. We modeled
maternal drinking (any drinking during the 3 months before pregnancy or first
trimester in model 1) and binge drinking (first trimester binge drinking in
model 2) separately. We included maternal smoking in the final models given
its significance in other studies. Potential confounders such as maternal
age, education, marital status, and birth weight were assessed. Interactions
between bed sharing or sleep position and prenatal maternal smoking or alcohol
use, postnatal alcohol use or smoking, and layers of clothing were examined
and considered significant at P = .10. Odds ratios
(ORs) were considered significant if their 95% confidence intervals (CIs)
excluded 1.0 or if the P values were <.05.
Seventy-two deceased American Indian infants younger than 1 year were
enrolled. Autopsy reports were obtained in 56 cases (5 infants whose deaths
were classified as "unexplained" were not autopsied and were assigned a diagnosis
of undetermined; 9 infants whose deaths were "explained" did not have autopsies
ordered; and 2 infants whose deaths were classifed as "infectious" had autopsies
but reports could not be located). Thirty-seven cases were SIDS, 27 cases
had explained causes (infections, injuries, congenital anomalies), and 8 cases
were undetermined. Among the 37 SIDS cases, 1 parent refused to be interviewed
and 3 parents had moved and could not be located. For the 33 cases with parental
interviews, the death scene protocol was completed by the coroner or from
police reports in 24 cases, and in 9 cases, summaries of scene investigations
were reviewed by the PIMR and determined to be compatible with a diagnosis
Data for 66 control infants matched to the 33 case infants were analyzed.
Five cases and their matching controls were from an urban Indian community
and the rest were from rural Indian reservations. The median age difference
between case and control infants was 2 days (range, 0-30 days). The median
interval from the date of death of the index case to the parental interview
was 30 days (range, 8-250 days) for case infants and 33 days (range, 2-330
days) for control infants.
The mean age at death of the SIDS infants was 109 days, 51.5% were male,
and 64.7% died during the autumn and winter months. We found no significant
differences between case and control infants in the mean values of selected
maternal and infant sociodemographic and health care utilization factors,
except in mean monthly household income (Table 1).
Sociodemographic and Health-Related Factors. Parents of SIDS infants were significantly more likely than control
parents to have 12 years of education or less and less likely to have a telephone
in the home (Table 2). Infants
born to mothers who reported fewer than 7 prenatal visits were at a significantly
increased risk for SIDS as were those whose mothers reported that inadequate
transportation was a barrier. Infants whose mothers reported being visited
by a PHN either before or after birth had a significantly lower risk for SIDS.
Adverse Maternal Behaviors. A higher percentage of case mothers reported smoking cigarettes during
the 3 months prior to pregnancy and during the 3 trimesters than control mothers,
but these differences were not statistically significant (Table 2). Smoking rates were high among both cases and controls
with the highest rates during the 3 months preceding pregnancy, decreasing
during each of the subsequent trimesters, and increasing after delivery to
almost prepregnancy levels. Among those who reported smoking during pregnancy,
the average number of cigarettes smoked per day did not vary significantly
by case or control status (5.8 vs 6.2 cigarettes per day).
A higher percentage of case mothers reported using alcohol during the
3 months prior to pregnancy and during each trimester than controls (Table 2). The difference in the percentage
using alcohol was statistically significant during the first trimester only.
Alcohol use for both groups was highest 3 months prior to pregnancy and lowest
during the second and third trimester and increased after delivery.
Binge drinking was more common among case than control mothers, but
the difference was significant only for the first trimester (Table 2). First trimester binge drinking was associated with a 6-fold
increased risk for SIDS. Rates of binge drinking decreased during pregnancy,
but remained higher among case mothers than control mothers.
Among drinkers, case mothers consumed an average of 4.5 drinks per day
vs 4.1 for control mothers on the days that they drank (P<.08); case mothers had an average of 1.9 drinking days per month
vs 1.1 for control mothers (P<.03); and case mothers
had an average of 4.8 binge drinking days per trimester vs 2.6 for control
To explore whether the association between SIDS and maternal binge drinking
reflects differences in maternal nutritional status, we examined maternal
pregravid body mass index and trimester-specific hematocrit and hemoglobin
levels. Maternal pregravid body mass index of less than 25 (69% of case and
55% of control mothers) was associated with a small nonsignificant increased
risk for SIDS (OR, 1.8; 95% CI, 0.6-4.7). Using standard trimester-specific
cut-offs, we compared low vs high levels of hematocrit (<33%, <32%,
and <33% for trimesters 1, 2, and 3, respectively) and hemoglobin (<11.0
g/dL, 10.5 g/dL, and 11.0 g/dL for trimesters 1, 2, and 3, respectively).
The OR for the association between low hematocrit values and SIDS progressively
decreased from the first through the third trimester but none were statistically
significant (OR, 4.45 [95% CI, 0.24-81.7]; OR, 1.41 [95% CI, 0.26-7.6]; OR,
0.85 [95% CI, 0.18-4.0]). We found similar associations between low hemoglobin
levels and SIDS.
About 10% of mothers reported using illicit drugs during pregnancy,
but differences between case and control mothers were not statistically significant.
Marijuana was the most frequently used drug.
Infant Sleep Care Factors. More than half of the infants usually shared a bed with their parent
at night in the 2 weeks preceding the case infant's death (Table 2), with similar percentages for case and control infants.
No significant interactions (P = .10) were observed
between usual bed sharing and maternal cigarette smoking or alcohol consumption.
The percentage of infants usually put to sleep on their stomachs in
the 2 weeks prior to death did not differ between case and control infants
(15.2% vs 13.6%). The percentages on their sides and backs were also similar
as were the percentages usually found on their stomachs during the night.
No significant interactions (P = .10) were observed
between usual sleep position and cigarette smoking or alcohol consumption
during pregnancy or the postpartum period.
Infants who had 2 or more layers of clothing or covers were at an increased
risk for SIDS, although the increase was only statistically significant for
infants with 2 or more layers of clothing (Table 2). No significant interactions (P =
.10) were observed between layers of clothing or covers and usual sleep position
or bed sharing.
Those who had fewer than 3 well-baby visits were at almost 14 times
greater risk for SIDS. Two thirds of case mothers and half of control mothers
reported that they ever breast-fed their infant, but the difference was not
We evaluated the following variables in a conditional logistic regression
model: maternal education (≤12 years vs >12 years), paternal education
(≤12 years vs >12 years), telephone in the home (present vs absent), number
of prenatal visits (<7 vs ≥7), maternal smoking during pregnancy (mothers
who reported smoking during pregnancy vs those who did not), layers of clothing
(0-1 vs ≥2), periconceptional maternal alcohol use (mothers who reported
using alcohol during the 3 months before or the first trimester of pregnancy
vs those that did not), binge drinking (mothers who reported binge drinking
during the first trimester vs those who did not), prenatal or postnatal PHN
visit (any vs none) and well-baby visits (<3 vs ≥3).
Periconceptional alcohol drinking (model 1) was associated with an increased
risk for SIDS (adjusted OR [aOR], 6.2; 95% CI, 1.6-23.3) (Table 3) as was first trimester drinking, but the model fit was
slightly better for periconceptional drinking. First trimester maternal binge
drinking (model 2) was associated with SIDS (aOR, 8.2; 95% CI, 1.9-35.3).
Neither binge drinking nor use of alcohol during the second or third trimester
Infants who usually had 2 or more layers of clothing had a greater risk
of dying of SIDS than those who had fewer layers (aOR, 6.2; 95% CI, 1.4-26.5).
When we excluded 4 of 33 matched triplets for which case mothers were interviewed
in the winter and control mothers in the summer, the OR fell from 6.2 to 5.2
but remained significant. Infants whose homes were visited by a PHN had a
significantly lower risk for SIDS than those who were never visited (aOR,
0.2; 95% CI, 0.1-0.8). Maternal smoking was associated with an increased risk
for SIDS, but the OR did not reach statistical significance. We found no significant
interactions among the risk factors included in the final model.
This study of SIDS among American Indians identified 3 factors that
are amenable to public health action and further research: (1) visits by PHNs,
(2) periconceptional maternal alcohol drinking and first trimester binge drinking,
and (3) infant layers of clothing.
Infants in homes that had any visit by a PHN before or after birth were
one-fifth less likely to die of SIDS than those in homes that were never visited.
Public health nursing is an integral component of the IHS programs and is
entirely community based. One possible explanation for the absence of a visit
is inaccessibility. However, nurse visits were not correlated with reports
of transport barriers to care, the number of well-baby visits, or the number
and timing of prenatal visits (data not shown). Further study is needed to
confirm the protective effect of PHN visits and to identify the effective
components of outreach activities.
A recent evaluation of home visiting programs concluded that variability
in results from one program to another indicates that the benefits of the
programs cannot be generalized.12 Two randomized
controlled trials of home visitation during the mother's pregnancy and her
child's first 2 years of life showed that such visits were associated with
positive pregnancy and childhood outcomes.13- 15 In
another study, the implementation of a universal postpartum nurse-visiting
program resulted in a significant reduction in acute care visits during the
infant's first 2 weeks of life.16
To our knowledge, this is the first study to report an association between
SIDS and periconceptional maternal alcohol consumption and binge drinking
during the first trimester. Few published studies have reported a relationship
between maternal alcohol use and SIDS, and none of them found an independent
correlation between maternal prenatal alcohol use and the risk of SIDS.17- 19 However, one study
did find a significant association between postnatal maternal alcohol use
and SIDS. One study examined maternal binge drinking during the month before
the infant's death but did not find it to be associated with SIDS.19
The proportion of pregnant control mothers reporting alcohol use during
the third trimester in the AAIMS (6.6%) is similar to the proportion of mothers
in the AAIHS (6.3%) who reported drinking during pregnancy on birth certificates
in 1994-1996.8 The proportion is 4.5% for all
IHS areas and 1.5% for all races in the United States.
In our study population, the pattern of drinking is predominantly binge
drinking, and the OR for maternal binge drinking and SIDS was highest during
the first trimester. Craniofacial anomalies, low birth weight, decreased head
circumference, and congenital anomalies have been correlated with alcohol
exposure in the first trimester in other studies.20- 22 However,
in this study, none of the infants who died had microcephaly, craniofacial
anomalies, or major brain malformations. We also found no evidence to suggest
that the association between alcohol and SIDS is mediated by poor prenatal
maternal nutritional status.
In addition, by definition, assignment of the SIDS diagnosis meant that
there was no evidence of abuse or neglect associated with excessive postnatal
alcohol consumption. More research is needed to confirm these findings and
elucidate the pathways leading to increased risk.
Excess thermal insulation for a given room temperature has been associated
with increased SIDS risk,23,24 and
the risk is further increased by viral illness25 and
prone sleep position.24 We found that usually
wearing 2 or more layers of clothing at night, not including the diaper, increased
an infant's risk for SIDS more than 6-fold. Neither the number of covers nor
the type (thin or thick blanket, sheet, quilt, or comforter) was significantly
associated with SIDS risk.
Bed sharing in combination with maternal smoking during pregnancy has
been shown to be associated with an increased risk for SIDS.19,26- 28 This
increased risk is also associated with other risk factors, ie, recent maternal
alcohol consumption, the infant being covered by a duvet, and parental tiredness.28 Bed sharing is routine among Northern Plains Indians.
While we did not observe significant interactions between bed sharing and
cigarette smoking or alcohol consumption, this study may lack the power to
adequately assess the relationship of bed sharing to other risk factors.
The primary limitation of the study is the small sample size. The study
had a 40% power to detect a 2-fold difference in smoking between case and
control mothers and a 42% power to detect a 2-fold difference in gestational
age between case and control infants. However, despite the small sample size,
positive associations of potentially modifiable contributors to SIDS were
found, although CIs were wide.
Another limitation is that the standard death scene form was not completed
on all unattended deaths in spite of the availability of formal coroner training
programs. Tribal police investigated half of the deaths because there was
no enabling tribal legislation for coroners. The PIMR reviewed coroner, police,
and emergency medical technician reports. When written reports of the scene
investigation were not provided, the personnel who investigated the deaths
were interviewed for information on possible homicide, overlying, or other
Recall bias regarding events during pregnancy or around the time of
infant death is another potential limitation. However, it is unlikely that
differential maternal recall bias between cases and controls occurred because
of the high self-reported rates of smoking and alcohol use and the similar
interval between death and interview of the cases and controls. Other case-control
SIDS studies that examined recall bias did not find an appreciable impact
on the important associations.29,30
Our results provide new evidence that factors in the periconceptional
period contribute to SIDS risk in addition to those identified in the prenatal
and postnatal periods. They suggest that public health outreach and programs
to reduce alcohol consumption among women of childbearing age could have an
impact on SIDS rates in this population.