Factors Associated With the Transition to Nonprone Sleep Positions of Infants in the United States:  The National Infant Sleep Position Study FREE

SUDDEN INFANT DEATH syndrome (SIDS) is the leading cause of postneonatal infant mortality (deaths between 1 month and 1 year of age) in the United States. In 1992, 4891 infants died with this diagnosis, corresponding to a rate of 1.20 deaths per 1000 live births.¹ The rate of SIDS in the United States changed little during the decade prior to 1992, despite significant drops in overall infant mortality. Studies in the United States have identified factors associated with 2-fold to 4-fold increased risk for SIDS, including preterm birth, low birth weight, young maternal age, high parity, late or no prenatal care, smoking and substance abuse during pregnancy,² and postnatal exposure to environmental cigarette smoke.³ Most of these factors are also associated with other causes of infant mortality⁴ and have proven to be difficult to modify in populations at risk.^5- 6

High SIDS rates in New Zealand, Australia, and the United Kingdom prompted epidemiological studies that focused on investigating risks in the infants' environment. These studies found a strong association between SIDS and placing infants prone (on their stomachs) for sleep, with reported odds ratios (ORs) ranging from 4.5 to 8.8.^7- 10 This knowledge, as well as prior clinical research in Hong Kong and the Netherlands,^11- 12 led to the initiation of public health programs in Australia, New Zealand, England, Denmark, and Norway to advocate the supine (on the back) or lateral (on the side) sleep position. Within 1 to 2 years after the public health campaigns, SIDS rates dropped 50% or more in these countries, with a concomitant reduction in the prevalence of the prone sleep position.^12- 16

In June 1992, the American Academy of Pediatrics (AAP) Task Force on Infant Positioning and SIDS published a recommendation that healthy full-term infants be placed laterally or supine to sleep.¹⁷ Over the next year, controversy regarding the AAP recommendation emerged, based on medical concerns about the change to nonprone sleeping.^18- 19 In January 1994, additional information on the health outcomes of nonprone sleeping infants led to a consensus to strengthen and accelerate risk reduction efforts.^20- 21 A coalition was formed among the US Public Health Service (PHS), the AAP, the Association of SIDS Program Professionals, and the SIDS Alliance for the planning, development, and implementation of a "Back to Sleep" national public education campaign, which was launched in June 1994.

In 1992, the National Institute of Child Health and Human Development (NICHD), on the advice of international experts, initiated population-based studies on sleep practices and health outcomes. This article reports on data from one of these studies, the National Infant Sleep Position (NISP) Study, which tracked changes in infant sleep position between 1992 and 1996 in the United States and assessed population characteristics associated with infant sleep position.

From May 1992 through 1996, Datastat Inc, Ann Arbor, Mich, conducted telephone interviews by randomly sampling households with infants younger than 8 months from a nationally representative list. The list, which was purchased from Metromail, Lincoln, Neb, is based on public information generated from birth records, infant photography companies, and infant formula companies. Metromail guarantees a 50% eligibility rate with the targeted subject population (in this case, households with infants). Interviews were completed in households that responded affirmatively to the question, "Is there an infant in this house that was born in the last 7 months; that is, on or after (date)?" Interviews with the infant's nighttime caretaker were requested and completed (80%-85% of those interviewed were the infant's mother). From 1993 to 1996, households in which the mother of the infant did not complete high school were oversampled after the national sample was complete.

Between 1002 and 1008 interviews were completed each year. In 1992, 2068 calls were attempted to complete 1002 interviews (1015 infants because some were twins) within the 48 contiguous states. Of the 2068 calls, 16.2% were to nonhouseholds (including nonworking numbers and businesses), 76.7% were to households where eligibility could be determined, and 7.1% were to persons who refused to be interviewed. For those 1586 households for which eligibility could be determined, 66.6% were eligible (N=1057; 1002 were interviewed and 55 refused after the interview began).

An exact response rate cannot be calculated because eligibility could not be determined for those who refused to be interviewed. An estimate of the response rate was made based on the assumption that the eligible proportion of the households who refused was the same as the eligible proportion of those for whom we could determine eligibility. For 1992, the estimated response rate was 86.8%: 1002 completed interviews divided by 1154 calls (1057 eligible and 97 estimated to be eligible). Using the same method, the estimated yearly response rates for 1993 through 1996 were 81.7%, 81.6%, 86.4%, and 84.8%, respectively.

A sample size of 1000 infants was targeted based on the criterion of detecting a minimum 10% change in the prevalence of the prone sleep position in infants younger than 8 months, from a starting value of 70%. At α=.01, a sample size of 827 has 0.95 power to detect a 10% change.

The interview was developed specifically for this study. The interview mainly focused on (1) whether the caregiver had a usual specific position in which she or he placed the child to sleep at night; (2) if so, what the position was (stomach, side, or back); (3) whether an infant placed in a given position changed positions during the night based on caregiver observations; (4) reasons for position placement; and (5) an array of sociodemographic information and sleep environment characteristics that could be associated with infant position. Pilot testing was carried out to make sure questions were clear to respondents and the interview was not too long or burdensome. The sleep position data are limited to the usual position in which the nighttime caretaker placed the infant to sleep. Specific information about daytime naps and positions used by other caregivers was not obtained. The interview lasted an average of 10 minutes and was administered by the computer-assisted telephone interview method.

A telephone survey has the limitation of reaching those who have a telephone and, therefore, underrepresents the economically disadvantaged population. When compared with the 1992 natality statistics published by the National Center for Health Statistics (NCHS),²² the NISP survey sample underrepresents the following maternal characteristics: black race (5.6% [NISP] vs 16.2% [NCHS]), Hispanic ethnicity (4.5% [NISP] vs 15.8% [NCHS]), age, younger than 20 years (5.3% [NISP] vs 12.8% [NCHS]) and less than 12 years of education (6.3% [NISP] vs 23.2% [NCHS]). Infants with low birth weight also were underrepresented (4.3% [NISP] vs 7.2% [NCHS]).

The simple analysis of trends in lateral, prone, and supine placement was evaluated for significance with the χ² test for trend. Predictors of prone and supine placement were studied using multiple logistic regression analysis.^23- 24 Starting in 1993, some variables of interest were added to the interview and, thus, survey years 1993 through 1996 were used for the logistic analysis. Consistency of the time trends in the multiple logistic models for decreased prone placement and increased supine placement was studied by evaluating the significance of interactions among survey year and other predictor variables in the multiple logistic models. From 1993 on, we attempted to compensate for the limitation of the sample in the underrepresentation of the economically disadvantaged population by oversampling for households with mothers who did not complete high school. Additional multiple logistic regression analyses were done to include this oversample, but since the results were essentially unchanged, the results reported here do not include the oversample.

SIDS rates and infant mortality rates were obtained from the annual published supplements to the Monthly Vital Statistics Reports of the NCHS.^25- 26 Deaths from SIDS are reported as those deaths with SIDS as the underlying cause in the International Classification of Diseases, Ninth Revision (ICD-9) code 798.0. The preliminary SIDS rate for 1996 is based on approximately 84% of the deaths.²⁶

When asked, "Do you have a position that you usually place the baby in to sleep?" more than 97% of respondents in each year replied affirmatively. The proportion of infants younger than 8 months placed prone to sleep declined dramatically from 1992 to 1996 as follows: 70%, 58%, 43%, 29%, and 24% (χ² test for trend,P <.001). From 1993 on, those placing their infants prone were asked to clarify whether the infant was placed with the face down or with the head turned to the side. The percentage who placed the infant with the face down was 1% for 1993 through 1995, and 0.5% in 1996. Supine placement increased during the 5-year period (except in 1996) as follows: 13%, 17%, 27%, 38%, and 35%. Lateral placement increased over the 5-year period as follows: 15%, 22%, 27%, 32%, and 39%.

The position in which the infants were placed varied by the race or ethnicity of the mother. Among whites, prone placement decreased from 71% in 1992 to 22% in 1996. Conversely, supine placement increased from 13% to 37% (Figure 1). A larger proportion of black mothers placed their infants prone (82% in 1992, declining to 43% in 1995 and 1996). A large increase in supine placement occurred among blacks between 1993 and 1994 (from 2% to 17%), but there were no substantial changes thereafter (Figure 1, bottom). The decline in prone placement among blacks between 1994 and 1995 is largely accounted for by an increase from 20% to 35% in lateral sleeping (Figure 1, middle). Infants of Asian or other ethnicity were placed prone in lower proportions and placed supine in higher proportions in all years compared with whites, with lateral sleep position increasing slightly (to 30%) in 1996. The year-to-year variability shown among Hispanics may be because of the low sampling proportion and the heterogeneity of this ethnic group. In 1996, 19% of Hispanic mothers placed their infant prone, 49% placed them laterally, and 28% placed them supine.

The time trends in specific age groups for each of these positions are shown in Figure 2. By 1996, 13.0% of infants younger than 8 weeks were placed prone (Figure 2, top), 72% were placed laterally (Figure 2, middle), and 13% were placed supine (Figure 2, bottom). Prone prevalence in infants 8 to 15 weeks of age dropped by two thirds between 1993 and 1996, from 60% to 18%, and lateral prevalence doubled from 26% to 50% (χ² test for trend, P<.001 for prone and lateral). Within 1 year after the initiation of the "Back to Sleep" campaign, there was a doubling in supine placement among infants 8 to 15 weeks of age, from 17% to 35% (P<.001), with no further increase in 1996. Between 1992 and 1995, the year-to-year changes in the patterns of sleep position placement for infants 16 weeks or older were significant at P<.001. By 1996, 28% of infants in this age group were placed prone, 26% were placed laterally, and 43% were placed supine.

Multiple logistic regression analysis of time trends for prone or supine sleep position placement, and the association with particular maternal and infant characteristics, was performed on the combined data sets for 1993 through 1996, with year as a categorical variable (Table 1). The probability that an infant was placed prone decreased by about half each year between 1993 and 1995. By 1996, an infant was only one fifth as likely to be placed prone as an infant in 1993. The probability of an infant's being placed supine increased almost 3-fold between 1993 and 1995, with little difference occurring between 1995 and 1996.

With a few exceptions, major characteristics significantly and independently associated with the probability that an infant was placed prone were also associated in the opposite direction with being placed supine (Table 1). After adjustment for other sociodemographic and infant characteristics, black mothers were still significantly more likely to place their infants prone (OR, 2.34), and much less likely to place them supine (OR, 0.39), compared with whites. The adjusted ORs for prone and supine among infants of Hispanic or Asian or other race did not reach statistical significance. After adjustment for other factors, including maternal characteristics, younger infants were significantly less likely to be placed to sleep prone (OR, 0.63 for <8 weeks and 0.82 for 8-15 weeks) or supine (OR, 0.24 for <8 weeks and 0.57 for 8-15 weeks) compared with infants 16 weeks or older (Table 1). Significant interactions with regard to prone sleep position were observed between year and the age of the infant (P=.03), because there was more change over time in the 8- to 15-week age group than in either the younger or older infants.

Mothers aged 20 to 29 years were about 30% more likely to place their infants prone and 20% less likely to place them supine compared with mothers 30 years or older. Maternal education of less than 16 years was a significant independent predictor for reduced use of the supine sleep position. Parity was also a significant independent factor: half of the mothers had a previous child, and those mothers were 1.68 times as likely to place the current infant prone and 0.72 times less likely to place the infant supine.

The change in sleep position placement varied by region. Respondents living in mid-Atlantic and southern states were significantly more likely to place their infants prone (1.41 and 1.47 times more likely, respectively), compared with those living in the Midwest and New England. Conversely, they were significantly less likely (0.72 and 0.65 times less likely, respectively) to place them supine.

Infants placed to sleep at night in the last 2 weeks in settings other than the crib were significantly less likely to be placed prone. This is in part because of an interaction (P=.06) between the infant's age and the place slept; younger infants were more likely to be placed in bassinets or other places, including cradles. There was also an interaction between year and the place the infant slept (P=.04), in which the change in prone sleep position placement was most rapid for infants who were placed in cribs. Infants who were usually placed in an adult bed were 1.80 times more likely to be placed supine. Infant sex and birth weight were not significant predictors of sleep position placement in this model.

When asked, "Does the baby usually change position during the night?" between 52% and 55% of respondents answered affirmatively each year. The respondents were then asked, "In what position do you commonly find the baby when you check on him or her during his or her sleep?" The percentage of infants found prone was always similar to the percentage placed prone (eg, 28% vs 24% in 1996), with 0.4% to 2% found with their face down in the bed. However, more infants were found supine than were placed supine (eg, 50% vs 35% in 1996). Therefore, the change in infant sleep position reported by the nighttime caregiver was analyzed to determine the developmental stability of the position in which the infant was placed to sleep.

The probability that an infant placed in one position would be found in another was the same each survey year despite the different prevalences of the position placed. Therefore, the data from all 5 years were combined to increase statistical power (Table 2). Prone position was found to be a stable position. From birth through 15 weeks of age, infants placed prone had a .96 probability of being found prone. The probability declined significantly with age but remained .73 at 24 weeks or older. Supine position was as stable as prone position through 15 weeks of age. The probability that an infant remained supine declined significantly thereafter, but a large percentage (68%) of the infants were still on their backs at 24 weeks and older. The probability that supine-placed infants moved to the lateral position was greater than the probability that they moved to the prone position, but the trend to prone increased significantly with age; at 16 to 23 weeks it was .06 and at 24 weeks or older, .14. Lateral sleep position was found to be the least stable. Even at younger than 8 weeks, the probability of a laterally placed infant remaining on his or her side was .72. This probability declined steadily with age, leveling to .30 at 16 weeks and older. Most laterally placed infants rolled to their backs, but there was a finite probability of infants rolling to their stomachs that increased with age: .05 at younger than 16 weeks, .12 at 16 to 23 weeks, and .18 at 24 weeks or older. The increasing trend of laterally placed infants rolling to their stomachs with age is statistically significant.

The SIDS rates for 1985 through 1996 are shown in Figure 3. Prior to 1992, the SIDS rate was relatively constant, with an annual rate of reduction of only 1.3%. A sustained drop in deaths from SIDS has occurred since 1992. Between 1992 and 1995, the SIDS rate declined 28.5%, from 1.2 per 1000 live births to 0.87 per 1000 live births. Most of this decline took place between 1993 and 1995, with a decline of 12% between 1993 and 1994 and a decline of 15.5% between 1994 and 1995. There were 1494 fewer SIDS deaths in 1995 than in 1992. The preliminary rate of 0.74 SIDS deaths per 1000 live births for 1996 indicates that the downward trend in SIDS rates has continued.²⁶

The postneonatal mortality rate decreased by 13%, from 3.1 per 1000 live births in 1992 to 2.7 per 1000 live births in 1995. The preliminary rate in 1996 was 2.5 per 1000 live births. Since more than 95% of SIDS deaths occur in the postneonatal period, the decline in SIDS accounted for about three fourths of the decline in the postneonatal death rate between 1992 and 1995.

Between 1992 and 1995, the SIDS rate for white infants declined 30% and the SIDS rate for black infants declined 18%. The black-white ratio has increased for SIDS from 2.16 in 1992 to 2.52 in 1995. Both the preliminary SIDS rate for black infants (1.54 per 1000 live births) and the preliminary SIDS rate for white infants (0.6 per 1000 live births) for 1996 are about 15% lower than the race-specific final SIDS rates for 1995, and thus, the black-white ratio remained about 2.5.

The annual surveys conducted since 1992 as part of the NISP Study serve as an evaluation of the national response to the AAP recommendation and the "Back to Sleep" public health education campaign. These surveys are conducted during the same time frame each year, with similarly constituted samples and response rates, providing a consistent analysis of time trends in infant sleep position and risk factors associated with sleep position choice in the United States. This information is a valuable guide for future intervention efforts to further decrease prone sleep position.

Practical constraints, including the need to obtain baseline information quickly pending the release of the AAP recommendation in 1992, necessitated the use of telephone interviews from a national, commercially available sample enriched for households with young infants. The resulting survey population underrepresents the economically disadvantaged. However, to address this concern, an oversample for mothers with less than 12 years of education was instituted beginning in 1993. When additional multiple logistic regression analyses were performed including the oversample, the results were essentially the same.

The values of prone prevalence in the national survey are similar to those reported by others for specific geographic locations using other survey techniques. For example, a survey in Toledo, Ohio, conducted by pediatricians with parents of infants younger than 7 months found that in 1993, 54% of infants were placed prone to sleep, 31% were placed laterally, and 15% were placed supine.²⁷ Gibson et al²⁸ reported that surveys conducted in public clinics and private practice settings in Philadelphia, Pa, showed a 62% prone prevalence in 1993, which declined to 41% in the spring of 1994.

The multiple logistic models indicate that those infants at greatest risk for SIDS by maternal sociodemographic criteria (ie, young age, low educational level, black race, and high parity) are also those at greatest risk for not being placed in the supine or lateral sleep position. The higher SIDS rates among blacks in recent years compared with whites may be a reflection of higher prevalences of prone sleeping, as well as the contributions of other risk factors such as higher rates of low birth weight, prematurity, and young maternal age.²⁹

More research is needed to understand why mothers do not place their infants supine to sleep. Are they receiving the message from the public health campaign? How do they perceive the risk of SIDS and the nonprone sleep interventions? Are there cultural or practical barriers to implementation? A study conducted in Louisville, Ky, between November 1993 and March 1994 found that among parents in a predominantly white private practice setting, 72% reported that their physician recommended nonprone sleep position compared with 48% in a predominantly black inner-city clinic.³⁰ The compliance with the nonprone recommendation for those in the clinic was lower (54%) than those attending the private practice (72%). Although this is a restricted sample, it illustrates discrepancies in the delivery and implementation of the "Back to Sleep" message.

The decline in the SIDS rate in the United States correlates in time with the decline in prone sleeping. Assuming an OR of 2.0 for prone sleep position³¹ and a prone prevalence of 70%, the population attributable risk³² can be estimated at 41% in 1992. Based on this calculation, the SIDS rate should have declined by a little more than 20% between 1992 and 1995, when in fact it declined by 29%.

There are other SIDS risk factors with relative risks of similar magnitude to prone sleep position; however, their prevalences in the population have been lower and more stable. To illustrate using US natality statistics,^22,29 12.7% of the births in 1992 and 13.1% in 1995 were to mothers younger than 20 years, 5.2% and 4.2% were to mothers with late or no prenatal care, 7.1% and 7.3% had low birth weight, and 11% in both years were preterm. Mothers who reported smoking during pregnancy decreased from 16.9% in 1992 to 13.9% in 1995. The decline in cigarette smoking during pregnancy and public health efforts to reduce prenatal and postnatal smoking may have contributed to, but cannot account for, the recent substantial decline in SIDS deaths.

Another potential explanation for the decline in SIDS rate could be changes in assigning SIDS as the underlying cause of death. In 1991, a National Institutes of Health (NIH) expert panel recommended a revised definition of SIDS for research purposes, which stated that a death scene investigation should be part of the diagnosis.³³ As a result, some cases lacking a scene investigation may have been certified as "cause undetermined." However, the autopsy rate for SIDS cases in the United States is high and has remained unchanged in recent years (93.4% in 1990 and 95.7% in 1994).³⁴ Also, an analysis of other causes of death related to SIDS, including "cause undetermined," could not account for the decline in the rate between 1990 and 1994.³⁴

In this survey and others,^27- 28 parents reported that the infant changed position during the night. The NISP survey responses show consistent patterns of developmental change in the position found during the night at different ages. Prone and supine sleep position are much more stable than lateral sleep position. The stability of the prone position is of concern, particularly in the presence of soft bedding, which increases the risk for SIDS, possibly by asphyxiation.^35- 36 The instability of the lateral position should be noted because a small proportion of infants on their side roll to their stomach, and the risk of that happening increases with the age of the infant. Multivariate analysis of a recent case-control study of stillbirths and infant deaths in the United Kingdom has shown that infants placed to sleep laterally were 1.84 times more likely to die of SIDS than infants placed to sleep supine.³⁷ Lateral sleep position has also been reported to confer more than twice the risk relative to supine position in univariate analysis of the New Zealand Cot Death Study³⁸ but not in multivariate analysis of the case-control study of SIDS in Tasmania.³⁹

In November 1996, the AAP revised its sleep position statement to emphasize that supine is the preferred position, although lateral is an acceptable alternative since it confers protection relative to prone position.⁴⁰ This recommendation has been incorporated into the "Back to Sleep" campaign materials. It is encouraging that after the initiation of the "Back to Sleep" campaign there was a substantial increase in supine placement for infants 8 to 15 weeks of age, when the peak incidence of SIDS occurs.² There were no gains, however, in supine prevalence between 1995 and 1996, indicating the need for sustained efforts to promote supine sleeping. England's recent experience, in which 75% of infants 3 months of age are reported to be placed supine to sleep⁴¹ and in which SIDS rates are 70% below preintervention levels,⁴² is an example of what may be accomplished.

There are many challenges to public health interventions in the United States. The United States lacks both the nationalized health care system and broadcast media that promoted dissemination of the sleep position message overseas. Also, the population is extremely large and heterogeneous. Despite these challenges, the changes in infant sleep position since 1992 have been dramatic. The success achieved in reducing deaths from SIDS will, hopefully, give added impetus to further dissemination and implementation of this simple intervention.