May 2003, VOLUME111 /ISSUE Supplement 1

Sleep Environment and the Risk of Sudden Infant Death Syndrome in an Urban Population: The Chicago Infant Mortality Study

  1. Fern R. Hauck, MD, MS*,
  2. Stanislaw M. Herman, PhD*,
  3. Mark Donovan, MS*,
  4. Solomon Iyasu, MBBS, MPH,
  5. Cathryn Merrick Moore, CDA*,
  6. Edmund Donoghue, MD§,
  7. Robert H. Kirschner, MD§,
  8. Marian Willinger, PhD
  1. *Loyola University Chicago Stritch School of Medicine, Maywood, Illinois
  2. Centers for Disease Control and Prevention, Atlanta, Georgia
  3. §Office of the Medical Examiner of Cook County, Chicago, Illinois
  4. National Institute of Child Health and Human Development, Bethesda, Maryland


Objective. To examine risk factors for sudden infant death syndrome (SIDS) with the goal of reducing SIDS mortality among blacks, which continues to affect this group at twice the rate of whites.

Methods. We analyzed data from a population-based case-control study of 260 SIDS deaths that occurred in Chicago between 1993 and 1996 and an equal number of matched living controls to determine the association between SIDS and factors in the sleep environment and other variables related to infant care.

Results. The racial/ethnic composition of the study groups was 75.0% black; 13.1% Hispanic white; and 11.9% non-Hispanic white. Several factors related to the sleep environment during last sleep were associated with higher risk of SIDS: placement in the prone position (unadjusted odds ratio [OR]: 2.4; 95% confidence interval [CI]: 1.7–3.4), soft surface (OR: 5.1; 95% CI: 3.1–8.3), pillow use (OR: 2.5; 95% CI: 1.5–4.2), face and/or head covered with bedding (OR: 2.5; 95% CI: 1.3–4.6), bed sharing overall (OR: 2.7; 95% CI: 1.8–4.2), bed sharing with parent(s) alone (OR: 1.9; 95% CI: 1.2–3.1), and bed sharing in other combinations (OR: 5.4; 95% CI: 2.8–10.2). Pacifier use was associated with decreased risk (unadjusted OR: 0.3; 95% CI: 0.2–0.5), as was breastfeeding either ever (OR: 0.2; 95% CI: 0.1–0.3) or currently (OR: 0.2; 95% CI: 0.1–0.4). In a multivariate model, several factors remained significant: prone sleep position, soft surface, pillow use, bed sharing other than with parent(s) alone, and not using a pacifier.

Conclusions. To lower further the SIDS rate among black and other racial/ethnic groups, prone sleeping, the use of soft bedding and pillows, and some types of bed sharing should be reduced.

  • sudden infant death
  • infant care
  • blacks
  • sleep
  • risk factors

Sudden infant death syndrome (SIDS), the leading cause of postneonatal mortality in the United States, currently accounts for approximately 3000 deaths per year.1 In the past 2 decades, SIDS rates among blacks consistently have been more than twice that of whites. On the basis primarily of research conducted in other countries,2 national interventions were developed to reduce prone sleeping and other factors associated with SIDS.3 Despite the success of these interventions that resulted in a decline of prone sleeping and SIDS rates among all racial/ethnic groups,4 the black-to-white ratio for SIDS still exceeds 2.0.1

The Chicago Infant Mortality Study was designed to examine risk factors for SIDS and other sudden infant death with the principal goal of gathering information to aid in eliminating disparities in postneonatal mortality between blacks and whites. This population-based case-control study, which took place in Chicago between 1993 and 1996, investigated sociodemographic, behavioral, and medical characteristics of the family and infant; characteristics of the home environment; and the circumstances of death. This article presents a comprehensive picture of SIDS risk in a primarily black urban population, giving particular attention to hazards in the sleep environment.


Case Selection and Data Collection

This study included all 260 Chicago resident infants whose death between November 1993 and April 1996 was determined by the Office of the Medical Examiner of Cook County, Illinois, to be caused by SIDS, resulting in 100% case ascertainment. A comprehensive death scene investigation included approximately 400 questions detailing the circumstances before death; the sleep environment of the child when last put down and found; the infant’s and family’s medical history; the mother’s prenatal alcohol, tobacco, and drug use history; and other factors pertinent to determining the cause of death.517 Photographs were taken to indicate the location and position of the infant when found. A scene investigation, autopsy, and review of the medical history were conducted for all 260 case infants, as specified by the SIDS definition used in this study: “The sudden death of an infant under 1 year of age, which remains unexplained after a thorough case investigation, including performance of a complete autopsy, examination of the death scene, and review of the clinical history.”18

Two weeks after the death, a standardized follow-up interview with the primary caregiver was conducted. This interview consisted of 235 questions addressing issues not included in the scene investigation, such as routine sleep habits of the infant, social stressors and supports, and access to and satisfaction with health care. The follow-up interview was conducted for 198 (76%) of the case infants; this was with the mother for 95% of the infants, 3% with the foster parent or legal guardian, and 2% with other relatives. Participants and nonparticipants in the follow-up interview were similar in race/ethnicity, marital status, parity, education, adequacy of prenatal care, and infant’s age at death. Reasons for nonparticipation were also similar by race/ethnicity.

Selection of Controls

One living control infant was matched to each case infant on (in order of priority) maternal race/ethnicity (self-reported), age at death/interview, and birth weight (<2500 g, 2500–<4000 g, and ≥4000 g; ±250 g if in the middle category).4 Potential control infants who met the matching criteria were identified through ongoing review of birth certificates at the Chicago Department of Public Health. They were randomly selected in groups of 20 for white infants and in groups of 40 for Hispanic and black infants, based on experience gained during the pilot phase of the study. The mothers of each group of infants were contacted simultaneously by mail and invited to participate. Mothers who responded to our invitation were interviewed on a “first come” basis, and once a control mother for a given case infant was interviewed, additional interested control responders for that infant were notified that they would not be needed.

The home interview consisted of nearly 500 questions taken from the death scene investigation and follow-up case interview, which were reworded to apply to a living infant. A reference sleep period was identified for the control infant to coincide with the time of day when the respective case infant was found unresponsive; all questions about “last sleep,” including positioning, referred to this reference sleep.

The control response rate was 4.7% (260 enrolled + 231 interested but nonenrolled of 10 464). There were no differences between enrolled control mothers and those who were interested but not needed, including race/ethnicity, education, marital status, smoking or drinking during pregnancy, high-risk pregnancy, parity, abnormal birth outcome, Kessner index of prenatal care,19 infant gender, and birth weight. Differences between enrolled responders and those who did not reply to our letter were found in maternal education (responders were more highly educated; P = .001), parity (responders had slightly lower parity; P = .027), and adequacy of prenatal care (65.2% of responders and 53.6% of nonresponders had adequate care; P = .001).

The study was approved by the institutional review board of the Loyola University Medical Center, Maywood, Illinois. A detailed description of the study methods has been reported elsewhere.20

Statistical Analysis

Analysis was conducted using data for the 260 SIDS infants and their 260 matched control infants. To determine differences in sociodemographic and sleep environment factors between cases and controls, we used the Cochran-Mantel-Haenszel statistic to estimate χ2 in comparisons of binary and nominal-scaled variables and the independent sample t test to compare interval data. For both the univariate and multivariate analyses, conditional logistic regression was used to take the matching into account. Unadjusted and adjusted odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were calculated. Bed sharing was defined conservatively as an infant sleeping with 1 or more people on the same sleep surface, such as a mattress or a sofa.

To determine the independent contribution of risk factors found to be significantly associated with SIDS on univariate analysis, we constructed a final multivariate model, using backward step-down variable selection. We also included maternal smoking during pregnancy in this model because it has been identified in previous research as a strong risk factor for SIDS21 and as an effect modifier for some sleep-related risk factors.2224 The population attributable risk refers to the theoretical proportion of cases that might have been prevented if a certain risk factor were eliminated. It was calculated for each of the risk factors that remained in the final multivariate model separately and for all of the risk factors combined, using the method of Bruzzi et al.25 Interactions were also tested between each of the sleep environment variables found significant on univariate analysis; as is often done for such analyses, the significance level for the interaction terms based on these comparisons was raised to .10.26 All analyses were conducted using SAS/STAT, release 6.12. (SAS Institute, Inc, Cary, NC).


As per the study design, cases and controls were similar on the matched factors. The racial/ethnic composition of both groups was 75.0% black, 13.1% Hispanic white, and 11.9% non-Hispanic white (2 Hispanic black case infants were classified as black, and black control infants were matched to them). The mean age (89 days and 85 days, respectively) and birth weight (2813 g and 2915 g, respectively) were not significantly different between cases and controls (P > .05). There were differences between cases and controls on nonmatched demographic factors. Case mothers were slightly younger than the control mothers (23.2, standard deviation 5.4 vs 24.8, standard deviation 6.4 years; P < .002). They also had lower educational attainment, had less adequate prenatal care as measured by the Kessner index, and were more likely to be single and have higher parity (all significant at P < .001). There were no differences in employment status.

Because several sociodemographic factors were found to be associated with increased risk for SIDS, analyses were done to examine the potential confounding effects of these and other related variables on the association between the exposure variables, ie, infant sleep habits and maternal behaviors, and SIDS. Maternal age, marital status, education, and adequacy of prenatal care were found to represent closely all of the factors and were therefore used in subsequent analyses for adjustment purposes.20

There were small differences between case and control infants in their sleeping surface and the location (Table 1), but none of these differences was statistically significant. More than half of the infants in both groups were sleeping on an adult bed mattress for the last sleep, and fewer than one quarter were sleeping in cribs.


The Chicago Infant Mortality Study, 1993-1996: Unadjusted and Adjusted Univariate ORs for Variables in the Sleep Environment*

Several factors related to the sleep environment were associated with a higher risk of SIDS (Table 1). Being placed in the prone position at last sleep was associated with having more than twice the risk of SIDS (OR: 2.4; 95% CI: 1.7–3.4), compared with nonprone positions. A soft sleep surface (defined through self-report as the infant’s head sinking 1 inch or more into the surface) was associated with having 5 times the risk of SIDS (OR: 5.1; 95% CI: 3.1–8.3), and pillow use was associated with almost 3 times the risk (OR: 2.5; 95% CI: 1.5–4.2). Covering of the head or face with bedding was associated with a similar risk level (OR: 2.5; 95% CI: 1.3–4.6), whereas other potential thermal factors, including swaddling during last sleep and the infant sweating in the past 2 days, were not.

SIDS infants were more likely than controls to have been ill with a runny nose or upper respiratory infection in the 2 days before death/interview (OR: 2.5; 95% CI: 1.7–3.8). Cases and controls did not differ in the frequency of cough, wheezing, diarrhea, or vomiting. Decreased risk of SIDS was found with pacifier use during last sleep (OR: 0.3; 95% CI: 0.2–0.5) and with breastfeeding for any length of time (ever; OR: 0.2; 95% CI: 0.1–0.3) or currently (OR: 0.2; 95% CI: 0.1–0.4).

Infant bed sharing with 1 or more people was associated with increased risk (OR: 2.7; 95% CI: 1.8–4.2). The OR associated with the mother alone (n = 49; 72% of the parental bed sharers) or with the mother and father together (n = 19; 28% of the parental bed sharers) was 1.9 (95% CI: 1.2–3.1). It was much higher for the other combinations of bed sharing, including other children alone or other children with 1 or both parents (OR: 5.4; 95% CI: 2.8–10.2). Fifteen SIDS infants shared a sofa during sleep, whereas no control infants did. After removal from analysis of these 15 SIDS cases and their matched controls, the ORs for bed sharing were similar. Sharing a room with anyone was associated with increased risk of SIDS (OR: 1.8; 95% CI: 1.2–2.6), as was sharing a room with the mother or both parents (OR: 1.6; 95% CI: 1.1–2.4) or other combinations of people (OR: 2.2; 95% CI: 1.4–3.4) in the unadjusted model. After adjustment of the significant variables for the 4 potentially confounding variables (maternal education, marital status, age, and prenatal care), all of the significant factors in the unadjusted analyses remained significant except for parental bed sharing and room sharing (Table 1).

In the final multivariate model, factors that remained significant independent risk factors were not using a pacifier, soft sleep surface, maternal smoking in pregnancy, prone sleep position, pillow use, and bed sharing in combinations other than the parents alone (Table 2). When this analysis was limited to blacks only, results were similar. On the basis of the final full-sample multivariate model, the population attributable risks were not using a pacifier, 56%; soft sleep surface, 39%; maternal smoking in pregnancy, 37%; prone sleep position, 33%; pillow use, 17%; and bed sharing in combinations other than with the parent(s) alone 17% (Table 2). Usual sleep practices within the last 2 weeks before death for cases or the 2 weeks before the reference sleep period for controls, including sleep position, location, softness of the sleep surface, use of a pillow, room sharing, and bed sharing, were not associated with increased risk of SIDS after adjusting for maternal education, marital status, age, and prenatal care.


The Chicago Infant Mortality Study, 1993-1996: ORs and Population Attributable Risks for Risk Factors in the Sleep Environment*

A significant positive interaction was found between prone sleep position and soft bedding surface (P = .05), ie, the combined presence of both factors had a greater effect than would be expected by simply multiplying the effects of each factor alone. The OR for prone sleep and soft surface, adjusted for the 4 confounding variables, was 21.0 (95% CI: 7.8–56.2). Similarly, an interaction was found for prone position and pillow use (P = .04), resulting in an OR of 11.8 (95% CI: 4.0–34.4). Softness of the sleep surface and pillow use were only weakly correlated. Other possible interactions were examined; there was none between bed sharing and surface softness or between bed sharing and maternal smoking either during pregnancy or postpartum.


Results from the Chicago Infant Mortality Study provide clues to explain the higher rate of SIDS among black infants compared with white infants. Placing infants on a soft surface for sleep, particularly in conjunction with prone positioning, may contribute to this disparity. The association of prone sleeping with SIDS has been well-established in countries outside the United States2,2729 but less so for the United States.3032 In this study, prone sleep position was confirmed as a risk factor for SIDS in univariate and multivariate analyses. The high rates of prone sleeping among both case and control infants in this predominantly black sample are consistent with reports that blacks were more likely to use the prone position even after the Back to Sleep campaign was under way.4,3338 This may explain, in part, the higher rate of SIDS in this population.20

We also found sleeping on a soft surface to be a strong independent risk factor for SIDS. Although the assessment of softness was subjective, participants were provided with guidelines and the questions were identical for case and control mothers. In addition, because little was known among the public about mattress softness and SIDS risk when this study was conducted, the responses here were not likely to be biased. The combination of soft sleeping surface and prone position was extremely hazardous, as found also in a Tasmanian case-control study of SIDS.39

As previously observed,40,41 pillow use was associated with an increased risk of SIDS. Various mechanisms have been proposed to explain the association between soft sleep surfaces and SIDS. Asphyxia may occur when sleeping prone on soft surfaces as a result of rebreathing of expired air4246 or of blockage of external airways.41 Another potential mechanism is hyperthermia, either directly or in conjunction with other thermoregulatory interactions.47 Prone infants, especially those in contact with soft underbedding, may have reduced ability to lose heat, making them more susceptible to hyperthermia.47

The relationship between bed sharing and SIDS has sparked lively debate.4852 This study makes an important contribution by examining a population in which bed sharing is common.53,54 An increased risk of SIDS was observed for bed sharing, but multivariate analysis indicated that the risk was primarily associated with bed sharing when the infant was sleeping with people other than the parents. Because there were few mother-father bed sharers, the findings for this category were driven primarily by the mother-infant dyad. These results are reassuring and consistent with laboratory studies demonstrating that more maternal inspections, more infant arousals, and less deep sleep among infants may occur when mothers and infants sleep together routinely.5557 Our results do not support a protective role for parent-infant bed sharing, however. Unlike studies in New Zealand and England, which have demonstrated that the risk of SIDS associated with bed sharing is primarily among smoking mothers,23,24,29,58 in the Chicago Infant Mortality Study, the smoking status of the mother did not influence the risk of bed sharing.

Sleeping together on a sofa occurred with 15 SIDS but no control infants, indicating that this practice may be extremely hazardous. A similar finding has been reported from England.23 Although it is possible that these infants died of asphyxiation or entrapment, infants who were found wedged or entrapped in other ways were not given a SIDS diagnosis. The SIDS diagnosis was reached after a thorough review, including a careful scene investigation. Regardless, sleeping with an infant on a sofa should be discouraged, as should bed sharing between an infant and family members other than the mother or parents.

Room sharing with parents was not associated with a reduced risk of SIDS, unlike the results of 2 other studies.23,59 This was largely reflective of the strong correlation between room sharing and bed sharing and the high rate of bed sharing in this study, resulting in few families that shared a room without bed sharing. Also, usual sleep practices in the preceding 2 weeks were not significantly associated with SIDS. In light of our findings for last sleep, this result indicates that sleep-related risk reduction measures need to be followed at all times.

Pacifier use during last sleep was found to lower the risk of SIDS substantially in this sample, a finding consistent with that of several other studies.29,6062 It is not known whether this association is reflective of behavioral characteristics of the caregiver or of the infant or is the direct effect of the pacifier during sleep, either mechanical (eg, directly maintaining an open airway or keeping the infant in a face-to-side position) or longer term maturational effects influencing airway patency. Although using a pacifier to prevent SIDS has not been recommended in the United States,3 it has been recommended in Germany for all infants63 and in the Netherlands for bottle-fed infants.60 Although some authors have argued that pacifiers should not be recommended until a physiologic mechanism that explains the association between pacifiers and the reduced risk of SIDS is identified,64 this situation is similar to avoidance of the prone position, whose physiologic mechanism in SIDS causation is still unknown. Consideration should be given to including pacifiers as a new strategy that might reduce the risk of SIDS even further while continuing research to study possible adverse consequences, including breastfeeding duration,6570 dental malocclusion,7173 otitis media,7477 and other health problems.78 In the meantime, parents who already use pacifiers for their infants and those who are not breastfeeding need not be discouraged from using them.

Breastfeeding was found to be protective against SIDS in the univariate analysis and after adjusting for sociodemographic factors, but it became nonsignificant in the multivariate model that included the other environmental factors. These results are consistent with most published reports and suggest that other factors associated with breastfeeding, rather than breastfeeding itself, are protective.29,40,79,80

As in any retrospective study, recall bias may occur if mothers of SIDS infants recall exposures more thoroughly than mothers of unaffected, healthy infants, thus yielding an apparent association when none exists. Prospectively collected data on sleep position, however, have confirmed results from other studies, indicating that recall bias has not been a major problem in case-control studies of SIDS.81,82 The length of time lapsed between the exposure and the recall has been shown to have a greater influence on recall accuracy than case or control status.83 In this study, parents of both SIDS victims and control infants were interviewed about their infant’s sleep position shortly after the last sleep (or reference) period, minimizing recall bias.

The control selection process presented numerous challenges, particularly in light of the difficulty in recruiting people of color into research studies8490 and the criteria that controls be matched to case infants on race/ethnicity, birth weight, and age. Because the number of potential controls contacted was far larger than those actually enrolled, the possibility of nonresponse bias exists. Responders were more educated, had better prenatal care, and were of lower parity than nonresponders. Adjustments to outcomes of interest were made to help control for these differences.20 However, there may have been other unmeasured differences between responders and nonresponders that could have confounded the relationships found in this study.

Parents are influenced strongly by physicians in choosing the sleep position for their infants.4,20 Other infant care practices, such as bed sharing and use of soft bedding, may also be influenced by medical providers, particularly if reinforced by the media.4 To reduce the racial disparity in SIDS rates, all families must be counseled regularly about recommendations for reducing the risk of SIDS.3,91 On the basis of the findings of this study, they should receive instruction that emphasizes supine sleeping, firm bedding, not using pillows, and not sharing a bed with other children or sleeping with another person on a sofa, while being sensitive to parental concerns and cultural traditions.

Although education about risk factors for SIDS is critical, it loses its effectiveness if financial constraints prevent families from following the recommendations. For example, parents may be unable to provide firm sleep surfaces for their infants if they cannot afford to purchase new, firmer mattresses. Similarly, bed sharing with other children may be unavoidable unless families have enough beds for their members or at least a crib for their infants. Thus, additional research is needed to determine the role that these factors may play in acceptance of the recommendations and to evaluate interventions, such as crib donation programs.

As a result of the Chicago Infant Mortality Study, a multiagency intervention was implemented in Chicago to reduce the risk of SIDS among black families, through brochures and posters, media messages, videotapes, direct parental education, and education of health providers. The greater decline in SIDS rates among blacks in Chicago than in previous years supports the effectiveness of this approach.92 The national Back to Sleep campaign has expanded its outreach to target minority populations more effectively.93 Risk factors particularly pertinent to black, as demonstrated in this study, must be addressed to reach the national goal of eliminating the racial disparity in SIDS.94


This work was supported by the National Institute of Child Health and Human Development and the National Institute on Deafness and Other Communication Disorders under contract number NO1-HD-3-3188 and the Centers for Disease Control and Prevention and the Association of Teachers of Preventive Medicine under cooperative agreement number U50/CCU300860-06.

We sincerely thank the Office of the Medical Examiner of Cook County (Chicago, IL), the Chicago Department of Public Health, and the families that generously participated in this study. Finally, we acknowledge the significant contributions made by Dr Robert Kirschner to this study and to children’s well-being in general; he will be deeply missed.


    • Received October 2, 2002.
    • Accepted December 4, 2002.
  • Reprint requests to (F.R.H.) Department of Family Medicine, University of Virginia Health System, Box 800729, Charlottesville, VA 22908-0729. E-mail: frh8e{at}
  • The content of this publication does not necessarily reflect the views or policies of the US Department of Health and Human Services or the Association of Teachers of Preventive Medicine, nor does the mention of trade names, commercial products, or organizations imply endorsement by the US government.

  • Deceased.

  • Dr Hauck is currently with the University of Virginia Health System, Charlottesville, Virginia; and Mr Donovan is currently with Covance, Princeton, New Jersey. Dr Iyasu is currently with the Food and Drug Association, Rockville, Maryland.

SIDS, sudden infant death syndromeOR, odds ratioCI, confidence interval