Changes in Sleep Position During Infancy: Title and subTitle BreakA Prospective Longitudinal Assessment FREE

SUDDEN INFANT death syndrome (SIDS) is defined as the sudden death of an infant not expected by history and unexplained by a thorough postmortem examination, death scene investigation, and medical history review. SIDS is the most common cause of death for infants between 1 and 12 months of age, and in the United States in 1995 occurred at a rate of 0.87 cases per 1000 live births.¹

The cause of SIDS is unknown, but a number of factors have been associated with an increased risk. Incidence is higher among blacks than whites, is inversely proportional to birth weight, is increased by maternal illicit drug use and smoking during pregnancy, and may be decreased by breast-feeding.² Recent studies, mostly from countries other than the United States, have suggested that infants sleeping prone (lying on their stomachs) are at increased risk.^{3 - 5} Following successful campaigns to reduce the prevalence of prone sleeping, SIDS rates have decreased by approximately 50% in these countries.^{6 - 10}

In April 1992, the American Academy of Pediatrics (AAP) recommended that healthy newborns be placed on their side or back to sleep, and, in 1994, the "Back to Sleep" campaign was initiated in the United States by a joint statement of the AAP, several government agencies, and SIDS organizations. The goal of the "Back to Sleep" effort is to reduce the prevalence of prone sleeping among infants to 10% or less.¹¹ Periodic telephone surveys of households with infants younger than 8 months indicate that the average prevalence of prone sleeping in this age group decreased from 70% in 1992 to 24% in 1996, and that the prevalence of prone sleeping was higher among minorities and older infants.¹¹

We described the prevalence of the prone sleep position among infants born in 1995 and 1996 as they aged from 1 to 6 months and sought to identify factors that predict which infants will be placed in the prone sleeping position at 3 months of age—the age at which SIDS has its peak incidence.

The Infant Care Practices study is an ongoing multicenter, prospective, longitudinal study conducted in Boston, Lowell, and Lawrence, Mass, and Toledo, Ohio. The principal aims of the study are to describe current newborn sleep practices, to document changes in sleep position with age and calendar time, and to identify determinants and health consequences of various sleep practices during the first 6 months of life. Since February 1995, mothers of newborn infants have been contacted at selected birth hospitals in these cities and invited to participate. Each working day, study personnel review the obstetrical records to identify women who delivered live infants during the previous 24 hours. Women are ineligible if they reside out of state (or plan to move out of state within 6 months), will not be caring for the child, are not fluent in English or Spanish, or have a child with a major congenital malformation or other medical condition that might influence sleep position. As time allows, eligible women whose infants weighed at least 2500 g at birth are contacted and invited to participate according to a list ordered by the terminal 2 digits of the mother's medical record number. The order on the list is random with respect to maternal and infant characteristics and time of delivery. Our goal is to enroll at least 50% of eligible mothers and infants. (A similar approach, using a separate list, is used for women delivering infants weighing less than 2500 g, but these subjects are not considered in this report.) After informed consent is obtained, mother-infant pairs are enrolled and followed up until the child's first birthday. Descriptive data (eg, maternal age, race/ethnicity, parity, the infant's birth weight) and information on SIDS risk factors (eg, history of maternal smoking during pregnancy) are collected by interview at the time of enrollment. Follow-up data are collected by mailed questionnaire when the infants are 1, 3, and 6 months of age. Mothers not responding to the mailed questionnaire are interviewed by telephone. At each follow-up, data concerning infant health status and infant care practices (including the position in which the infant was placed for sleep on the previous night and factors that influenced the mother's choice of position) are recorded.

The χ² test was used to compare proportions. Multiple logistic regression was used to evaluate potential risk factors (predictors) for sleep position, while controlling for the potential confounding effects of other factors.¹² Indicator terms were included in the logistic regression models for the following: race/ethnicity, maternal age, marital status, parity, education, smoking while pregnant, annual household income, infant's sex and postnatal age, breast-feeding, and calendar time. Linear trend was assessed using continuous terms in the logistic models. Ninety-five percent confidence intervals (CIs) for odds ratios (ORs) were calculated using the test-based method.¹³

Of 12,748 women delivered of a live infant on a day covered by the study staff at participating hospitals between February 22, 1995, and December 31, 1996, 1529 (12%) were ineligible, 1270 (10%) refused, and 1640 (13%) were not invited to participate either because the woman was discharged before she could be contacted or the recruiter's time did not permit. A total of 7797 mothers whose infants weighed at least 2500 g at birth were enrolled. At the time of this analysis, follow-up data were obtained at 1, 3, and 6 months for 6405 (82%), 5746 (74%), and 4806 (62%) infants, respectively. Data for both the 1-month and 3-month follow-up were available for 5207 infants. Overall, 92% of enrolled women provided follow-up data at least once; responses from women with incomplete follow-up were not materially different from those given by women providing complete follow-up. Descriptive characteristics of mothers, who provided follow-up data at 1 and 3 months, and their infants are shown in Table 1 and Table 2. Of the 6405 mothers providing data at 1 month, the majority were non-Hispanic whites, 654 (10%) were non-Hispanic blacks, and 697 (11%) were Hispanic. The mean age of these women was 28 years, the mean birth weight of the infants was 3520 g, and 6357 (99%) of the 6405 infants were singletons. The characteristics of those providing data at 3 and 6 months were similar to women providing data at 1 month (data for 6-month follow-up not shown). The median ages of the infants at the time of the 1-month, 3-month, and 6-month follow-up were 4 weeks, 13 weeks, and 27 weeks, respectively.

Sleep position for these infants at 1-month, 3-month, and 6-month follow-up is shown in the Figure. The proportion of infants placed in a prone position at 1, 3, and 6 months was 1171 (18%) of 6405, 1679 (29%) of 5746, and 1568 (33%) of 4806, respectively; the corresponding figures for supine sleeping were 1171 (18%) of 6405, 2139 (37%) of 5746, and 2072 (43%) of 4806. Less than 4% of the infants at any follow-up were placed to sleep in a sitting or unknown position (data not shown). A majority of infants, 2879 (55%) of 5207, were placed in the same position at both 1 and 3 months. Of the 5207 infants, 757 (15%), 682 (13%), and 1417 (27%) slept prone, supine, and on their side, respectively, at both 1 and 3 months. Among the 915 infants placed prone at 1 month, 757 (83%) continued to be placed in this position at 3 months. Among the 921 infants placed supine at 1 month, 682 (74%) continued to sleep in this position at 3 months. Among the 3188 infants placed on their sides at 1 month, 1771 (56%) were changed to another position by 3 months. Of those that changed, a third changed to the prone position; the remainder changed to supine. Infants sleeping on their sides at 1 month accounted for 579 (39%) of 1496 infants placed to sleep in a prone postion at 3 months.

Sleep position at 1 and 3 months in relation to maternal and infant characteristics is shown in Tables 1 and 2. At 1 month, prone sleeping was more common among non-Hispanic blacks and Hispanics than other racial/ethnic groups; male infants and infants born to unmarried women were more likely to sleep prone than female infants and infants born to married women. The prevalence of prone sleeping was inversely related to maternal age, education, and income and directly related to parity. At 3 months, the prevalence of prone sleeping was higher overall and generally showed relations to maternal and infant characteristics similar to those seen at 1 month. Among Asian infants, the prevalence of prone sleeping was low and did not increase between 1 and 3 months.

The results of multiple logistic regression models examining the relation of maternal and infant characteristics to prone sleeping at 1 month and the risk of changing from nonprone at 1 month to prone sleeping at 3 months are shown in Table 3. Simultaneous control for all of the listed factors revealed prone sleeping at 1 month to be significantly associated with race/ethnicity, maternal age, education, parity, and the infant's sex, but not maternal smoking history, marital status, breast-feeding, or household income. Compared with non-Hispanic whites, the adjusted OR for prone sleeping among non-Hispanic blacks was 2.1 (95% CI, 1.7-2.6) and, among Hispanics, it was 2.2 (95% CI, 1.8-2.8). The odds of sleeping prone decreased with increasing maternal age and education and increased with increasing parity (P for trend <.001 for each comparison). Compared with women delivering their first child, the ORs for women with 1 or more previous children were 1.5 and 2.0, respectively. The relations between race/ethnicity, maternal age, parity, and infant's sex and the risk of changing sleep position from nonprone to prone at 3 months are similar to those seen with prone sleeping at 1 month. The odds of changing to prone sleeping at 3 months was inversely related to maternal age and directly related to parity (P for trend <.001 for each comparison). Marital status, years of education, cigarette smoking, breast-feeding, and annual household income were not significantly associated with the change to prone sleeping at 3 months.

In a multivariate model created to examine risk factors for sleeping prone at 3 months (rather than for changing from nonprone to prone at 3 months), prone sleeping at 1 month was the strongest predictor of prone sleeping at 3 months. Compared with infants sleeping on their sides, those prone at 1 month had an OR of 21 (95% CI, 17-26) for sleeping prone at 3 months. For supine sleeping at 1 month, the OR was 0.7 (95% CI, 0.5-0.9). The point estimates for all other characteristics were similar to those shown in Table 3 (data not shown). Risk factors for prone sleeping at 6 months and for the change from nonprone sleeping at 3 months to prone at 6 months were also similar.

Fitting separate logistic models by race indicates that the effect of maternal age on the risk of prone sleeping at 1 month and the risk of changing from nonprone to prone by 3 months was limited to non-Hispanic black and non-Hispanic white women; similarly, the effect of parity on both risks was observed only among Asians and non-Hispanic whites. Annual household income of less than $35000 was associated with an increased risk of prone sleeping at 1 month among Asian women (OR, 3.3; 95% CI, 1.3-8.3) and an increased risk of changing from nonprone to prone sleeping at 3 months among Asian and Hispanic women (the ORs were 4.9 [95% CI, 1.0-2.4] and 3.3 [95% CI, 1.2-9.2], respectively). Logistic models stratified by enrollment year indicate that the factors associated with prone sleep at 1 and 3 months had not changed over the period covered by these data.

The factors reported to have the greatest influence on choice of sleep position according to parity and sleep position at 1 and 3 months are shown in Table 4. The width of the 95% CI for each estimate was±6% or less. Among mothers using nonprone sleep positions at 1 month, the advice of health care professionals was commonly cited as the most important influence on choice of sleep position, accounting for 56% and 48% of reports among primiparous and multiparous women, respectively. Between one fifth and one quarter of these women reported that printed materials (eg, magazines, newspapers, or pamphlets distributed by hospitals and clinics) were most influential. Among multiparous women, experience with an older child was reported as most influential by 20% and, among primiparous women, the category of family or friends was cited by 11%.

Among women using the prone position at 1 month, health care professionals were reported as the most important influence for 20% of primiparous and 13% of multiparous women. The most frequently cited influences among primiparous women were family or friends and the infant's behavior (eg, slept better or seemed to prefer this position), accounting for 35% and 25% of responses, respectively. Among multiparous women, experience with an older child and the study infant's behavior was cited by 48% and 22%, respectively. Among women who changed to prone at 3 months (after using a nonprone position at 1 month), the influencing factors were similar to those reported by women of the same parity who used the prone position at 1 month.

The current study documents the prevalence of prone sleeping among a large sample of infants in the United States, who are at the age of highest risk for SIDS (ie, 1-6 months), and identifies several maternal and infant characteristics associated with prone sleeping at 1 and 3 months of age and the risk of changing from nonprone to prone sleeping at 3 months. Although the prevalence of prone sleeping among US infants has decreased since the AAP made its first recommendation in 1992, more than 29% of mothers of 3-month-old infants (ie, the age of peak incidence of SIDS) born in 1995 to 1996 report placing their infants to sleep in the prone position. While it is clear that current public health efforts have had some success in influencing the initial selection of infant sleeping position and maintaining a nonprone sleep position through the age of highest risk for SIDS, the prevalence of prone sleeping at 3 months of age remains above the 10% goal set by the "Back to Sleep" program. Furthermore, efforts aimed at reducing the prevalence of prone sleeping have been least successful among women who are black or Hispanic and among non-Hispanic white women who are younger than 25 years or have other children.

This prospective, longitudinal assessment of sleep position in the United States indicates that both the prone and supine sleep positions are relatively stable (ie, few infants placed to sleep in these positions at 1 month change to another position as they grow). Prone sleeping at 1 month is the strongest predictor of prone sleeping at 3 months, when the risk of SIDS is greatest.

In virtually all other countries where a public health intervention to change infants' sleep position has been undertaken, the prevalences of prone sleeping have decreased substantially and are now generally below 10%.^{6 ,14 - 19} Following publicity advising side sleeping, the prevalence of prone sleeping among controls enrolled in the New Zealand Cot Death study in late 1990 was 7.7%.¹⁴ In a separate survey of 200 mothers in South Aukland designed to test the effectiveness of the National Cot Death Prevention Programme, 2.5% reported their infants slept prone.¹⁵ Following a public health campaign in Australia in 1991, the prevalence of prone sleeping at 1 month of age decreased to 5.4%.¹⁶ In data obtained from a case-control study of SIDS conducted in a single Norwegian county, 10 (8%) of 123 control infants (mean age, 5.7 months) born after a 1990 campaign to discourage prone sleeping usually slept prone.¹⁷ However, a national survey also conducted in Norway following the intervention found that the prevalence of prone sleeping at 3 months of age was 28%.⁶ In the United Kingdom, 7% of 3-month-olds were reported to sleep prone following the national "Back to Sleep" campaign.¹⁸ Data from a mail survey conducted in Hong Kong in the spring of 1994 indicate that the prevalence of prone sleeping was 3% at approximately 1 month of age.¹⁹ No intervention campaign was ever conducted in this jurisdiction. When we compare the US experience with that of other countries, it is important to recognize that the United States had the highest prevalence of prone sleeping prior to formal recommendations to change sleep position, and the "Back to Sleep" campaign was only begun in 1994.

These data are consistent with the US national telephone survey and regional reports that indicate the prevalence of prone sleeping has decreased from nearly 70% in 1992 to 1994 to about 25% in 1996.^{11 ,20 - 23} These data confirm that a mother's experience with an older child increases the risk that she will use the prone sleep position,²⁰ but do not confirm that risk is increased for unmarried women after controlling for other factors.^{22 ,24} Compared with earlier reports, the present study includes a larger sample of minority infants, resulting in more stable estimates and allowing us to detect significant associations between race/ethnicity and prone sleeping. Because a representative sample of all women delivered of a newborn at a participating hospital was eligible for the present study, our data are less subject to selection bias than telephone surveys, which by definition are limited to households with telephones. Furthermore, the unique longitudinal design of this study provides the first opportunity to observe changes in sleep position as children grow and to identify those infants who change to the prone position between 1 and 3 months.

There are several limitations to these data. Mothers were enrolled only in eastern Massachusetts and northwest Ohio, and compared with the US population, a higher proportion of study mothers had some postsecondary education. As such, these women may not be representative of all US women who gave birth during this time. In particular, the Hispanic women in our study were predominantly of Puerto Rican origin and may not reflect attitudes and behavior of other Hispanic groups (eg, Mexican Americans), and the total number of Asian women (predominantly of Southeast Asian origin) was small. Although only 74% of enrolled women provided follow-up at 3 months, there was no evidence that this biased the data. While prone sleeping at 1 month is the strongest predictor of prone sleeping at 3 months, the data do not indicate whether it is possible to change the habits of infants who sleep prone at 1 month.

Data from 1995 to early 1997 suggest that the goal of less than 10% of all infants sleeping prone at 3 months of age has not yet been met in the United States. The prevalence of prone sleeping at 1 month of age is less than 20%, but a substantial proportion of mothers change to prone by the time their infants are 3 months old. As a consequence, during the period of maximum SIDS risk, the prevalence of prone sleeping is still 29%. The risk of prone sleeping at 3 months was greatest among infants who sleep prone early in life, who are black or Hispanic, and whose mothers are young (<25 years), not well educated, or have 1 or more older children. Among Asian women, low income was associated with prone sleeping. Women choosing the prone position reported that this decision had been influenced most by the infant's behavior, family and friends (if this was their first child), or experience with an older child (if this was not their first child).

It seems reasonable that efforts intended to further reduce the prevalence of prone sleeping should be designed to target the population groups who are at particular risk for using this practice. It should also be recognized that antenatal and/or neonatal strategies intended to influence the initial choice of sleep position may be insufficient to maintain nonprone sleeping through the peak SIDS risk period. Well-child visits during the first weeks of life provide an additional opportunity to discuss sleep position as a risk factor for SIDS and potentially influence parents' care practices (ie, encourage the use of supine sleep position and discourage switching to prone) before the infant enters the peak SIDS risk period.