To study factors associated with sudden infant deaths occurring with the external airways (ie, nose and mouth) covered by bedding.
Case-comparison study of infants dying with vs those dying without the external airways covered.
Death-scene investigation and reconstruction at the site of death using an infant mannequin; 18 metropolitan areas.
Caregivers for a consecutive sample of infants who died of sudden infant death syndrome (SIDS). Complete data from 206 of 382 eligible cases.
Main Outcome Measures
Among infants dying suddenly and unexpectedly, an analysis of whether sociodemographic risk factors for SIDS, sleep practices, or bedding increased the risk of dying with the external airways covered.
Data were analyzed by using univariate and 2 types of multivariate risk analysis, logistic regression and latent class. Of the victims, 59 (29%) were found with the external airways covered. Conventional risk factors for SIDS did not affect the risk of death with the external airways covered. Factors increasing the risk of death with the external airways covered included prone sleep position (odds ratio [OR], 2.86) and using soft bedding (OR, 5.28), such as comforters (OR, 2.46) and pillows (OR, 3.31). Infants at low risk for death with the external airways covered slept in the prone position, but rarely on a pillow, comforter, or other bedding that allowed a pocket to form beneath the face. All 9 infants who were positioned supine or on one side for sleep and found with the external airways covered had turned and were found dead in the prone position.
Sudden infant deaths with the external airways covered were common in the United States when most infants slept prone. Soft bedding, including pillows and comforters, increased the risk that an infant who died would be found with the external airways covered. Therefore, these items should not be placed near infants, regardless of the sleep position.
RESEARCH conducted in other countries shows that sleeping with the head and face covered,1- 3 using certain items of soft bedding,3- 5 and sleeping in the prone position2,6,7 increase the risk of sudden infant death syndrome (SIDS). There are no controlled studies from the United States of the interactions among head and face covering, bedding, and sleep position. However, case series from the United States suggest that infants are at increased risk for death when they sleep on certain items of bedding, eg, waterbeds and sleeping bags, because the infants were found with part of the head covered, particularly the face, nose, and mouth.8,9 Physiological studies have shown that many of the bedding items cause substantial thermal stress or rebreathing of exhaled gases.9- 16 Thus, epidemiological data from abroad and limited clinical and physiological studies from the United States and elsewhere suggest that soft bedding covering the entire head or just the face can pose a risk for sudden death among infants. Indeed, before the recent "Back-to-Sleep" campaign, perhaps 30% of sudden infant deaths in the United States occurred with the infant prone and with the external airways (ie, nose and mouth) covered.10,17- 19
We report a case-comparison study conducted in the United States of infants dying suddenly and unexpectedly. For the study, we divided the infants into 2 groups. Group 1 included infants (n=59) who died with their external airways covered and group 2 included infants (n=147) who died with the external airways uncovered. We chose the case-comparison approach to compare the circumstances of their deaths because it increased the likelihood that we would identify factors associated with death with the external airways covered. We did not use the familiar case-control method commonly used in SIDS research to identify factors associated with death per se. Rather, we chose the case-comparison approach, with deceased rather than living control infants, because we were interested in determining whether certain sleep practices and bedding factors, such as prone sleep position and soft bedding, and risk factors for SIDS, such as winter season and prematurity, might distinguish the 2 groups.
Our results show that sudden infant deaths with the head covered, in the United States, occurred most often with parts of the front, rather than the back, of the head covered, and that deaths with the nose and mouth covered were more frequent among infants sleeping in the prone position on soft bedding. Our results emphasize the potential role of bedding and are pertinent to modifiable SIDS-related sleep practices in the United States.
The comparison groups in this case-comparison study were 2 groups of infants who had died suddenly and unexpectedly; group 1 infants died with their external airways covered, and group 2 infants died with their external airways uncovered. The procedures used to study each group were the same. Living control infants were not studied.
Cases from 2 parallel studies were combined: a multiple-site study conducted by the US Consumer Product Safety Commission (CPSC) and a study conducted in St Louis, Mo.10 The methods used in the CPSC and St Louis studies were essentially the same, with a few exceptions as noted. Both studies involved face-to-face interviews at the death scene with the person(s) who found the infant dead.
For the CPSC study, medical examiners or coroners in metropolitan areas near regional CPSC offices were contacted. The CPSC study was conducted from February 1, 1992, through December 31, 1993, in Georgia (4 sites), California (3 sites), Florida (3 sites), New Jersey (3 sites), Illinois (1 site), New Mexico (1 site), Texas (1 site), and Wisconsin (1 site). The St Louis study was conducted from May 27, 1991, to May 26, 1992. Sites were chosen where the medical examiner or coroner was willing to participate and where complete postmortem studies were done for all deaths due to SIDS.20
To assure case ascertainment, CPSC investigators were in weekly contact with the medical examiner's or coroner's office and also reviewed the office logs of deaths on a monthly basis (quarterly in Georgia).
The CPSC study interviews were conducted by investigators experienced with face-to-face interviews about product-related matters. In addition, the CPSC staff attended a 2-day program given by the St Louis investigators and other specialists in SIDS.7 The program emphasized how to conduct a thorough, but empathetic, death-scene investigation10,21,22 and how to avoid bias in the interview.
A protocol was developed that sought information on sociodemographic factors related to SIDS risk23- 25 (Table 1) and information about the infant's sleep environment, particularly at the time of death. Data sought included the usual bedding near the infant, the usual and final sleep position, whether the infant's external airways were covered by bedding, and the type of bedding. The description of the position at the time of death involved the use of an infant mannequin placed by the caregiver in the same position in which the infant was found dead (Figure 1). The mannequin weighed 5.0 kg, had movable limbs, and its head could be turned from side to side but not moved in the sagittal plane.10 During the investigation, the mannequin was handed to the caregiver who found the infant dead, and the caregiver was asked to place the mannequin on the bedding in the position in which the infant was found. The mannequin was left in the position for a minimum of 15 seconds. Particular attention was paid to head position, and extra care was taken by asking repeated questions about head orientation to document the probable position of the external airways. Photographs also were taken to demonstrate the position of the external airways in relation to the sleep surfaces and bedding. For infants in group 1, the person who found the infant dead had placed the mannequin with the external airways completely covered. For infants in the prone position, if the face was straight down, the position was recorded as "vertical"; if the head was 45° or less from the vertical position, the position was recorded "nearly face down"; 45° or more from the vertical position was recorded as "to the side." With the mannequin in position, if both nostrils and the mouth were covered, the death was recorded as "external airways covered" (Figure 1 and Figure 2); if either nostril or the mouth was visible in the reconstruction, the death was recorded as "external airways not covered." The scenes shown in Figure 1 and Figure 2 are typical of a quarter or more of the infant deaths in the United States; ie, face down, with few or no blankets over the trunk or back of the head, and much of the lateral aspect of the face exposed. In all cases in which the external airways were covered, the bedding was reported to be in physical contact with the mannequin's nose and mouth. The presence of a pocket that persisted after the mannequin was lifted off the bedding suggested that the bedding was soft and malleable.10,12 The depth of the pocket was measured, and the absence of any discernible impression was recorded as "no pocket." The St Louis study did not document the presence or absence of a pocket created by the mannequin in all the study bedding, and less information was available about antecedent illness. For most of the deaths in group 1, the bedding was tested as a cause of rebreathing of exhaled gases. The methods and results of these tests have been reported in detail.9,12,26,27
A reconstructed death scene that shows how 30% of victims of sudden infant death syndrome are found in the United States. An infant mannequin was positioned on a pillow, which was covered by a pillow case, that had been placed on top of a mattress to provide the infant with "more softness and comfort." The person who found the infant dead reconstructed the scene. The mannequin's head can be turned from side to side, but the infant in this case was found with the face directly down.
The corpse of an infant who died in the position shown, with the face buried in a comforter. The photograph was made by the Consumer Product Safety Commission investigator within hours of the infant's death. For the reenactment, the infant was positioned with the help of the mother, who found the infant dead.
Data from the CPSC study were reviewed for consistency with data from medical examiners' investigations. Although differences were unusual (3 of 59 in group 1), when there were important discrepancies, the data from the mannequin reconstruction were used.
Analyses were based on comparisons between the 59 infants in group 1 and the 147 infants in group 2, and the results are shown in Table 2, Table 3, Table 4, and Table 5. Unadjusted odds ratios were calculated for group 1 vs group 2 by using the standard risk factors for SIDS listed in Table 123- 25 and by using the possible risk factors associated with bedding. The bedding factors considered are shown in Table 4 and Table 5.2,7,10,12,26,28
Both univariate and adjusted odds ratios were computed using SPSS for Windows 6.14 (SPSS, Inc, Chicago, Ill).29 The latent class analyses were done using software described in detail.30
Two types of multivariate analyses were used: logistic regression and latent class analysis. These analyses were used to conservatively address 4 principal hypotheses and questions about group 1 deaths (external airways covered), namely, whether (1) a link exists between the prone sleep position and the death of infants in group 1; (2) deaths in group 1 were more likely on bedding with the physical properties of softness and malleability; (3) specific items of bedding used by infants from the United States increased the risk of death of group 1 infants; and (4) the SIDS-related sociodemographic risk factors affect the risk of death for group 1 infants.
From the multivariate logistic regression, 2 models were developed, with 3 steps in the development of each model (Table 5). From the latent class analysis, 3 subgroups were identified within the total sample of 206 (Figure 3).
Results of the latent class analysis. The association of soft bedding and death with the external airways covered (AWC) is emphasized. The prone sleeping position was common in all 3 subgroups. A, Subgroup with low risk for death with the AWC; 48% of 206 infants. None of the bedding in the low-risk subgroup allowed pocket formation, and no infants in that subgroup died with the AWC. B, Risk-pillow subgroup; 24% of 206 infants. C, Risk-comforter subgroup; 28% of 206 infants. Prone indicates sleeping in the prone position; pillow, sleeping on a pillow; pocket, sleeping on bedding that permitted pocket formation beneath the infant's face; comforter, sleeping on a comforter; and minority, black (76.0%), Hispanic (20.9%), or other (3.9%).
Multivariate logistic regression analysis29,31 with stepwise selection was used to identify variables associated with the group 1 deaths for 2 separate models (Table 5). Dichotomous variables (eg, prematurity and sex) were coded as 1 for present or 0 for absent. Cases from St Louis with missing data for pocket formation and illness were coded as zero (ie, "no pocket formation" and "no illness"). Two models were then developed, and the relevant results are shown in Table 5.
Model 1 (Table 5) assessed the significance of bedding-related variables after the effects of significant risk factors for SIDS had been held constant.23- 25 Three steps were used in developing this model, using stepwise procedures on each step. The first-step variables included all risk factors listed in Table 1, except sleeping in the prone position. The second-step variables included variables related to the infant's sleep environment: a blanket, a comforter, a pillow, or a sheet under the infant; bedding over the infant; formation of a pocket when the mannequin was positioned; bed sharing with another person; and a prone sleeping position. The third-step variables included interactions among the significant first- and second-step variables. Model 2 (Table 5) assessed the significance of bedding-related variables after the effects of all risk factors for SIDS had been held constant.
Latent class analysis 32,33 was used as an aid to interpret the data because it has the potential to identify subgroups (latent classes) within the sample of 206 cases. Infants within the subgroups identified by latent class analysis were at more (or less) risk for death with the external airways covered because of similar sleep practices. The latent class analysis was based on 6 dichotomous variables, the 5 risk factors identified in the logistic regression analysis (Table 5), and the outcome variable, external airways covered (yes or no). The latent class analysis yields a prediction of the fraction of the total sample with low risk or varying degrees of increased risk for death with external airways covered.32,33
In the CPSC study, 414 potential cases were identified in which SIDS was the initial or preliminary diagnosis. Of the original 414 cases, 64 were not included in the study, as further review showed that in 13 cases, death was due to suffocation (eg, wedging of the head and body, head entrapment, or overlying by a person in the same bed), and in 51 cases, other causes of death were found by examining the histological specimens (eg, respiratory infections, 27; "undetermined," 6; hypoxic-ischemic encephalopathy, 5; heart disease, 3; and dehydration, 2). Cases in which another diagnosis became apparent after the examination of histological specimens were referred from 11 sites, and no site-specific differences were apparent. Of the 350 remaining CPSC cases, death-scene investigations were conducted for 176. The remaining 174 cases were not included in the study for the following reasons: (1) the person who found the infant dead could not be contacted (67 cases [38.5]); (2) the person who found the infant dead was unwilling to be interviewed (73 cases [42.0]); (3) the infant's face position was not established reliably (15 cases [8.6]); and (4) miscellaneous reasons (19 cases [10.9]). Each CPSC site contributed approximately the same proportion to the cases that were included or not included in the study.
In St Louis, 30 of 32 potential cases were included, bringing the total number of cases for analysis to 206, ie, 206 cases in which complete postmortem and death-scene investigations led to the conclusion by the medical examiner or coroner that the death was due to SIDS. The median interval from death until a CPSC interview was 16 days; 89% were done within 2 months (ie, 61 days; range, 0-203 days); for St Louis, the mean±SD interval was 9.2 ± 7.7 days. For the CPSC data, the set of SIDS and bedding variables used in the multivariate analysis was unrelated to the interval between the death and the site visit (multiple regression, F23,167=1.24). Delays resulted from families being difficult to contact (eg, having no telephones or moving) and factors such as allowing time for the partial emotional recovery of caregivers. There were no significant differences (P≥.05) between cases of SIDS in the present study and cases in other large series (Table 2), with 1 exception. The mothers of the infants for the present study were significantly less likely to be smokers; however, our rate of 38.2% smoking is similar to the rate in more recent preliminary reports (42.0%).34 The mean±SD age of infants in the present study was 91.0 ± 53.5 days. Of the infants in minority groups, 76.0% were black, 20.9% were Hispanic, and 3.9% were other minorities.
Detailed information on sleep practices was unavailable for the cases not included because a home visit was not done. However, information pertinent to SIDS that was obtained from the medical examiners' and coroners' records shows that, as a group, these cases were not significantly different from the study cases described in Table 4. The characteristics of the cases not included were younger than 23 weeks of age, 87.1%; placed in the prone position before death, 72.9%; never breast-fed, 63.6%; death during winter (ie, between September 21 and March 20), 52.6%; boys, 64.3%; black, 54.1%; maternal smoking, 30.1%; mother younger than 20 years, 26.2%; birth weight less than 2500 g, 24.8%; and gestation less than 37 weeks, 22.5%.
The risk factors for SIDS rarely had a significant effect on whether an infant was found with the external airways covered (Table 1). However, there were 2 important exceptions: (1) In general, infants in minority groups were less likely to die with the external airways covered. (2) Being placed prone for the "last sleep" before death increased the odds ratio (OR) for death with the external airways covered to almost 3. Changes in behavior suggestive of illness tended to be more common in group 1 infants (OR, 1.87; 95% confidence interval [CI], 0.97-3.60, not significant). Almost the same percentage of infants in group 1 (43 ) as infants in group 2 (94 ; χ2=4.22, P≤.24) habitually slept in the prone position. This finding is analogous to one noted in several other studies of sleep practices and SIDS, ie, there is a weaker effect on the risk for SIDS if the infant habitually slept prone compared with having slept prone on the night of death.35- 37
Table 3 shows, in greater detail, that being positioned prone for the last sleep is associated with a much higher rate of dying with the external airways covered compared with being positioned on one side or supine (χ2=6.79, P≤.009). All 9 infants in group 1 who were positioned to sleep on one side or supine had turned to the prone position but were not found with external airways covered.
Although the data are not shown, only 2 infants in group 1 were found on one side or in the supine position. Of the infants in group 1 who were found in the prone position, 5 had their heads turned to the side; 52 infants in group 1 who were found in the prone position had their heads in the vertical or nearly face down position, and, thus, were positioned face down, or nearly so, at the time of death.
Table 4 shows the unadjusted ORs for bedding factors linked to deaths in group 1. Of the infants in group 1, 27 (46%) were sleeping on comforters; only 34 (23%) of the infants in group 2 were sleeping on comforters. Of group 1 infants, 14 (24%) were sleeping on pillows; only 19 (13%) in group 2 were sleeping on pillows. In group 1, after the mannequin was lifted off the bedding, a pocket 1 to 4 inches deep was found in 21 cases (36%); in 6 cases (10%), the pocket was between 1/2 and 1 inch deep; thus, a pocket was left in 27 cases (46%). In group 2, a pocket was discernible in only 21 (14.3%) of 147 deaths. Multiple layers of bedding beneath the infant were associated with increased risk (≥2 layers vs 1 or no layer on the mattress or fixed sleeping surface); layers of bedding over the infant (≥1 vs none) were not associated with an increased risk of death with the external airways covered.
Four deaths (2 from each group) occurred while the infants were sleeping on sheepskins.4,11 Only 82 (40%) of the 206 infants were found in sleeping environments designed for infants (eg, bassinets or cribs). The remainder were sleeping on adult beds (74 infants ), sofas (21 ), waterbeds (6 ), or the floor (6 ) or in playpens (10 ). Of the 206 infants, 70 (34%) were sharing a bed with another person at the time of death,38- 42 including 48% of black infants, 21% of white infants, and 22% of Hispanic infants (χ2=8.97, P ≤.003 for all subgroups). None of the sleep arrangements had a significant effect in group 1 deaths except sharing a bed; fewer infants who shared a bed were included in group 1 (corrected χ2=4.54, P≤.03)
Bedding from infants in group 1 was collected to study the relation to rebreathing of exhaled gases in 36 cases (61%). By using animal and mechanical testing models, 24 (67%) of the bedding caused marked rebreathing.9,10,26,27
Table 5 shows the adjusted OR for the 2 models derived from the logistic regression analysis. Three steps were used in the development of each model. For model 1, in the first step, only race, among 13 conventional risk factors for SIDS, was significantly related to death in group 1 after the 13 risk factors were entered using stepwise procedures. In the second step, holding race constant and using stepwise procedures, the risk of death in group 1 was significantly increased if the infant was positioned on a soft item that left a pocket, prone for the last sleep, on top of a pillow, or on a comforter. In the third step, there were no significant interactions among the first- and second-step variables shown in Table 5. For model 2, all 13 standard risk factors for SIDS were forced into the model in step 1. Results were essentially the same as with model 1; the same 4 bedding-related variables were significantly related to death in group 1 after all risk factors for SIDS had been held constant. However, in step 1 no risk factor for SIDS was significantly related to death in group 1; therefore, with the exception of race, these results are not shown in Table 5.
The latent class analysis was done to identify subgroups of infants with similar sleep practices within the sample of 206 cases. Results are shown in Figure 3. The subgroup and the predicted size of each were based on 6 dichotomous variables: sleeping in the prone position, 4 sleep practice variables from Table 5, and the outcome variable, external airways covered (yes or no). We calculated 2-, 3-, and 4-class latent class models for the 6-way frequency table. The 3-class model was chosen because it provided the best fit to the frequencies (χ247=42.86, P =.64), with no improvement in fit using a 4-class model.
The 3 subgroups of infants found by latent class analysis were as follows: (1) a low-risk subgroup for death with the external airways covered, with few infants sleeping on soft bedding; (2) a "risk-pillow subgroup," in which infants were more likely to be found sleeping on top of a pillow; and (3) a "risk-comforter subgroup," in which infants were more likely to be found sleeping on a comforter.
The subgroups had several noteworthy characteristics. Although the low-risk subgroup that represented 48% of cases had a high prevalence of sleeping in the prone position (>60%), there were no deaths with external airways covered and no infants found dead on a pillow or pocket-forming bedding, and the incidence of deaths of infants sleeping on comforters was low. For the risk-pillow subgroup (24% of cases), the prevalence of sleeping in the prone position was also high (62%), but more than 40% of infants were found with external airways covered. Compared with the low-risk subgroup, infants in the "risk-pillow subgroup" were much more likely to be found dead on a pillow (67%) and more likely to be found dead on pocket-forming bedding (37%). For the risk-comforter subgroup (28% of cases), the prevalence of sleeping in the prone position was very high (93%), and 67% were found with external airways covered. More than half of these infants were found dead on a comforter and on pocket-forming bedding. The representation of infants classified as minorities in the subgroups was as follows: low-risk, 70%; risk-pillow, 80%; and risk-comforter, 36%.
The 3-group latent class model predicted that 42% of the infants in the risk-pillow subgroup and 67% of the infants in the risk-comforter subgroup would be in group 1. Thus, the model predicted that 29.0% of the 206 cases would be in group 1 (actual rate, 28.6% [59/206]).
Results of death-scene investigations of 206 sudden unexpected infant deaths in the United States are presented. We document factors associated with death when the external airways are covered by bedding. These findings are the first direct evidence from epidemiological studies that infants positioned to sleep prone are more likely to die with the external airways covered. Although this may seem predictable, linkage between sleeping in the prone position and death with the external airways covered has not been established. That prone positioning increased the likelihood for death with the external airways covered was assumed but not studied earlier.10,11,13 We also explored the effect of risk factors for SIDS and the type of bedding used and whether the risk factors for SIDS or the type of bedding explained the risk of dying with the external airways covered.
Compared with infants in group 2, bedding beneath the infants in group 1 was soft and more likely to be sufficiently malleable to form a pocket beneath the infant (Table 4 and Table 5, Figure 3, B, and Figure 3, C). These findings were consistent with the studies of the physical properties of the bedding in cases of SIDS12 and with reports showing that pillows and soft bedding increased the risk for SIDS.28 Furthermore, the results shown in Table 4 and Table 5 and Figure 3, B, and Figure 3, C, identify specific soft and malleable items and indicate that comforters and pillows enhance the risk of death with the external airways covered. These results are analogous to the results of studies conducted in New Zealand and Australia, in which soft bedding, such as sheepskins and items filled with natural fibers, caused marked increases in the OR for sudden death while in the prone position.5
Individual risk factors for SIDS did little to increase the risk of death with the external airways covered. However, a trend was noted toward an increased risk of death with the external airways covered with changes in behavior within 48 hours of death, including irritability and sleeping more; perhaps these changes were associated with blunting of arousal,43 abnormal behavioral responses to carbon dioxide, or thermal stress in the sleep environment.44 Among factors having nonsignificant effects, the rate of death with the external airways covered was not higher during the colder months. This is surprising, because reduction in use of the prone sleeping position has been followed, in other countries, by loss of the winter peak in rates of SIDS.7,45 In the present study, the percentage of deaths during the winter months was similar (59.7%) to the percentage in an earlier US study (58.7%).25 One explanation for no winter peak in deaths with the external airways covered is that infants in the United States do not use bedding during winter that is softer or that causes more rebreathing of exhaled gases than the bedding used during summer.46 If this is correct, and rebreathing exhaled gases is an important explanation for deaths in the prone position, the loss of the winter peak in SIDS may be less clear in the United States, when few infants sleep prone.
Nine infants died with the external airways covered after turning to the prone position from one side or the supine position (Table 3).3 This suggests that caregivers should use firm bedding and that soft bedding is dangerous, regardless of the initial sleep position of the infant.
One third of the infants died while they shared a bed with another person, including 48% of black infants. Nevertheless, significantly fewer infants died with external airways covered while sharing a bed than while sleeping alone. One interpretation of this finding is that sharing a bed may interrupt the sequence leading to death with the external airways covered. It also is plausible that the external airways of infants who died while sharing a bed were covered at the time of death, but the position of the external airways was changed before they were found dead. Our findings about sharing a bed and death with the external airways covered merit study because it is our impression from previous studies that shared adult beds are more likely than standard infant beds to be softer and to have other physical properties associated with deaths in the prone position.
In some cases, the interval between death of the infant and the site visit was several days. Because there are no reports on effects of recall bias on recollection of death with the external airways covered,47,48 we can only speculate whether the person who found the infant dead would preferentially report that the infant died with the external airways covered. One study of sleep position showed no bias in favor of recalling that the infant was placed in the prone position; whether the external airways were covered was not the focus in that study, however.47
We used a case-comparison approach, rather than a case-control study. For a case-control study, appropriate controls would have been unexpected deaths not diagnosed as SIDS; however, this type of control group is unacceptable for the present study because it is very small and heterogeneous. More important, the case-comparison approach worked, by identifying factors associated with death with the external airways covered. Finally, if another case-control study had been done with living infants as controls, the results would have likely been consonant with ours, because only 2% to 4% of living infants habitually sleep in the prone position with their faces straight down,49 while 28.6% of our cases were found dead with the external airways covered, and almost all had their faces in the straight down or nearly straight down position. Why 20% to 50% of infants are found dead with the face down when only 2% to 4% sleep that way is likely explicable by a peculiar susceptibility of some infants who eventually lie face down on bedding that is soft or forms a pocket.
Some have claimed that the face-down posture is agonal and, thus, not instrumental in sudden death with the external airways covered.50 However, reports of witnessed unexpected deaths of 26 infants did not mention the infants turning into the face-down position.51 Moreover, acceptance of the face-down posture as agonal requires an explanation of why it is more likely to occur on soft items that form pockets. Finally, Waters et al52 showed that living infants in their own beds spontaneously positioned their faces so that their external airways were covered. This position sometimes caused striking abnormalities in oxygen saturation and partial pressure of carbon dioxide.52
We used latent class analysis (Figure 3), a type of multivariate risk analysis, to detect subtle subgroups within the 206 cases.32,33 The subgroups were based on the risk for death with the external airways covered, the type of bedding used, the sleep position, and status within a minority group. We thus described 2 at-risk subgroups and 1 low-risk subgroup. The at-risk subgroups (risk-pillow and risk-comforter) accounted for virtually all deaths in group 1. The 2 subgroups associated with increased risk differed in important ways from the low-risk subgroup. Infants at low risk for death with the external airways covered often slept in the prone position but rarely on a pillow, comforter, or bedding that allowed a pocket to form (Figure 3, A). Compared with the low-risk subgroup, the risk-pillow subgroup (Figure 3, B) was more likely to have used bedding that allowed formation of a pocket, to have slept on a pillow, and to be from a minority group. Infants in the risk-comforter subgroup were most likely to have been positioned for sleep on a comforter or other bedding that permitted pocket formation (Figure 3, C). Thus, latent class analysis suggests simple strategies to reduce deaths with the external airways covered among minorities, although being a minority infant reduced the general risk for death with the external airways covered. As shown in Figure 3, B, and Figure 3, C, many minority infants died with the external airways covered on pillows and comforters (80% and 36%, respectively).
Interest in factors affecting the risk for minority infants is warranted because the proportion of SIDS in many cities in the United States is much higher among poor black infants,10,23,53 even since initiation of the "Back to Sleep"6,54 campaign. Furthermore, black infants continue to sleep in the prone position on bedding, including pillows, that increases risk of death with the external airways covered.46 Therefore, the latent class analysis suggests that, because black parents persist in positioning their infants prone for sleeping, they should not use pillows, comforters, or any malleable item that allows pocket formation.
Accepted for publication December 18, 1997.
Supported in part by a research grant from the American Lung Association, New York, NY.
The opinions expressed are those of the authors and do not necessarily represent the views of the Consumer Product Safety Commission.
We thank Bradley T. Thach, MD; Edward E. Lawson, MD; Sally Davidson-Ward, MD; and Manon Boudreault, MPH, for their helpful comments during the preparation of this article.
Editor's Note: This is one case for soft findings (eg, pillows, comforters) being of great significance.—Catherine D. DeAngelis, MD
Reprints: James S. Kemp, MD, Cardinal Glennon Children's Hospital, 1465 S Grand Blvd, St Louis, MO 63104 (e-mail: email@example.com).
Scheers NJ, Dayton CM, Kemp JS. Sudden Infant Death With External Airways CoveredCase-Comparison Study of 206 Deaths in the United States. Arch Pediatr Adolesc Med. 1998;152(6):540-547. doi:10.1001/archpedi.152.6.540