Factors Associated With Sudden Unexpected Postnatal Collapse

We conducted a retrospective cross-sectional study of all SUIDs from all 41 states contributing data to the NFR-CRS during the period 2010–2020 and approved for analysis by researchers, corresponding to versions 5.0–6.0 of the NFR-CRS.⁹ The NFR-CRS is funded by the Maternal and Child Health Bureau of the US Department of Health and Human Services. The NFR-CRS includes detailed information of deaths reviewed by state child death review teams. NFR-CRS does not include SUPC deaths occurring in the birth hospital. It does contain SUPC deaths occurring after home birth or discharge from the birth hospital. The NFR-CRS includes infant demographic and clinical characteristics, parental demographics, household characteristics, prenatal history (including prenatal care, chronic health conditions, medications taken during pregnancy, tobacco use, substance use, prior live births), breastfeeding, and the conditions at the time of death (including sleep area, sleep position, objects in the sleep area, room sharing, sharing a bed or other surface, swaddling, and whether the caregiver fell asleep while feeding the infant), and cause of death (Supplemental Table 4). The NFR-CRS has been described in more detail elsewhere.^9,10 This study was deemed nonhuman subjects research by the institutional review board of the University of Virginia.

Following our previously published methodology,¹¹ we included all SUID deaths, which we defined as deaths of infants not having reached their first birthday and occurring unexpectedly in sleep or the sleep environment and not explained by non–sleep-related causes (eg, known medical conditions, intentional injuries). We did not have access to Apgar scores, which is also occasionally used to define SUPC. The study population of SUID was stratified into 2 groups: SUPC and non-SUPC deaths. We defined SUPC to be the death of an infant <7 days old who dies suddenly and unexpectedly after being born at term or near term (34–42 weeks’ gestation, inclusive).^1,7 The non-SUPC group included SUID occurring from 7 days old but not having reached their first birthday.

The main dependent variable was whether the death was defined as SUPC or non-SUPC. Our independent variables were 95 variables contained in the NFR-CRS (Supplemental Table 4), including child demographics (3 variables), parent demographics (4), household characteristics (3), infant health history (9), pregnancy and birth history (18), substances (including smoking and alcohol) taken during pregnancy (5), history of maltreatment (2), circumstances at the time of death and other sleep-related factors (34), and cause of death (2). Race and ethnicity were treated as social constructs, not biological characteristics, and were included, as there are known disparities in infant mortality by race and ethnicity. They were collected from the infant’s death certificate.

We performed a post hoc sensitivity analysis to test the robustness of our findings. We sought to exclude associations found in the main analyses that represent proxies for infant age as opposed to possible associations with SUPC. To overcome this limitation, we restricted the study population in our post hoc sensitivity analysis to infants <3 months old and stratified the group into SUPC (as defined previously) and 7 days–3 months of age.

We used simple descriptive statistics (frequencies and percentages) and compared the characteristics of infants with SUPC to the characteristics of the older group using χ² tests and OR [95% CI]. We then performed multiple imputation using the Markov Monte Carlo method to impute missing data of variables used in the analyses.^12,13 Multiple imputation has been previously used for analyses of the NFR-CRS database.^{10,11,14–17} Variables with ≥50% missingness were excluded from the analyses. Ten imputed datasets were created. All subsequent analyses used pooled estimates from imputed datasets. We performed individual unadjusted logistic regressions for each independent variable. Using variables with P < .10 in the single variable logistic regressions and stepwise selection, we built a multivariable logistic regression to calculate adjusted OR [95% CI]. Because associations detected through stepwise selection can be influenced by the order of variable entry, we also performed backward selection. Backward selection resulted in the same selection of variables for the final regression model. A P < .05 was considered to be statistically significant. All analyses were performed using SAS v.9.4 (SAS Institute, Cary, NC).

Of the 6051 SUID deaths meeting the inclusion criteria, 1.6% (n = 98) met the definition for SUPC (Table 1). Among SUPC deaths, the median age was 4 days, with 67% occurring on days of life 4–6. The most common place of SUPC death was the child’s home (85%). Compared with non-SUPC deaths, a higher percentage of infants who died of SUPC were non-Hispanic White (59.2% SUPC versus 41.9% non-SUPC; OR 2.35 [95% CI, 1.42–3.89]) and had a mother aged 35–39 years old (13.3% versus 7.0%; OR, 3.83 [95% CI, 1.24–11.84]). There were no differences in SUPC compared with non-SUPC deaths in the cause of death (eg, accidental suffocation, strangulation in bed) (40.8% versus 41.7%; OR, 1.08 [95% CI, 0.69–1.68]). There were also no differences between SUPC and non-SUPC deaths in the percentage of infants placed prone for sleep (13.3% versus 17.2%; OR, 0.84 [95% CI, 0.46–1.55]), but a lower percentage of SUPC deaths (compared with non-SUPC deaths) were found prone (15.3% versus 35.7%; OR, 0.35 [95% CI, 0.19–0.063]) (Table 2). Compared with non-SUPC deaths, a lower percentage of SUPC deaths had never breastfed (24.5% versus 30.2%; OR, 0.62 [95% CI, 0.39–0.99]). Although an equal percentage of SUPC and non-SUPC deaths occurred in an adult bed (55.1% versus 55.8%), a higher percentage of SUPC deaths occurred in a bassinet compared with non-SUPC deaths (11.2% versus 6.7%; OR, 3.33 [95% CI, 1.22–9.08]). SUPC deaths (compared with non-SUPC deaths) were more likely when the infant was surface sharing with an adult or another child (73.5% versus 59.6%; OR, 2.74 [95% CI, 1.59–4.73]) and when their caregiver fell asleep while feeding the infant (23.5% versus 8.1%; OR, 4.06 [95% CI, 2.49–6.64]). Compared with non-SUPC deaths, a higher percentage of infants who died of SUPC were swaddled at the time of death (36.7% versus 17.2%; OR, 3.61 [95% CI, 2.25–5.80] of SUPC). Although there were no differences in exposure to prenatal smoking, infants dying of SUPC (compared with non-SUPC deaths) were more likely to have a mother who took medications during pregnancy for diabetes mellitus (OR, 4.55 [95% CI, 1.38–15.00]) or epilepsy (OR, 5.66 [95% CI, 1.32–24.26]), but the infant was less likely to have a history of being maltreated (OR, 0.30 [95% CI, 0.11–0.82]) or to have a mother who smoked marijuana during pregnancy (OR, 0.38 [95% CI, 0.15–0.96]); however, the number of infants in these categories was extremely small, so these results should be interpreted with caution.

Of the 95 independent variables, 21 variables had P < .10 in bivariate analyses and were included in the multivariable logistic regression to generate odds ratios with 95% CI (Supplemental Table 5). Stepwise selection resulted in 6 variables included in the final logistic regression model: maternal age, live birth order, whether the caregiver fell asleep while feeding the infant, room sharing, swaddling, and the presence of a pillow in the sleep environment (Table 3). Compared with non-SUPC deaths, infants who died of SUPC had 13 times’ higher odds of having a parent ≥40 years (aOR, 13.1 [95% CI, 3.3–51.4]) and 7 times’ higher odds of having a parent aged 35–39 years (aOR, 7.2 [95% CI, 2.2–23.0]), compared with having a parent <20 years old. Compared with non-SUPC deaths, infants dying of SUPC had 4 times’ higher odds of being the first live birth (aOR, 4.0 [95% CI, 2.4–6.9]), compared with being the third or higher birth order. Among the circumstances at the time of death, infants dying of SUPC had approximately 2.5 times’ higher odds of dying after their caregiver fell asleep while feeding the infant (aOR, 2.6 [95% CI, 1.6–4.4]) and while being swaddled (aOR, 2.7 [95% CI, 1.7–4.1]) compared with non-SUPC deaths. Infants who died of SUPC had nearly 7 times’ higher odds of room sharing at the time of death (aOR, 6.9 [95% CI, 2.8–17.3]) compared with non-SUPC deaths. In contrast, they had half the odds of having a pillow in the sleep environment (aOR, 0.5 [95% CI, 0.3–0.8]) compared with non-SUPC deaths.

The results of our sensitivity analyses were also similar to our main analyses, except that presence of a pillow was no longer associated with SUPC (Supplemental Tables 6–9).

In this retrospective cross-sectional study of SUID deaths from 41 states during the period 2010–2020, we characterized 98 SUPC deaths not occurring in the birth hospital and compared them to nearly 6000 non-SUPC deaths. We found that 1.6% of the SUID deaths were attributable to SUPC. Of the 95 possible risk factors examined, SUPC deaths were more likely to occur when the mother was aged ≥35 years and when infants were the first live birth. In addition, infants dying of SUPC were more likely to be room sharing, swaddled, and to have their caregiver fall asleep while feeding the infant. Understanding differences in risk factors for SUPC and SUID at older ages is important for better elucidating any etiologies specific to SUPC and for providing the most salient safe sleep guidance to parents and caregivers based on the infant’s age.

Our findings are consistent with prior research on SUPC in finding that older maternal age and lower birth order are risk factors for SUPC.^2,4,6,8 Specifically, our findings regarding birth order are consistent with the only other large-scale study of SUPC.^6,8 Those analyses, after restricting their study population to gestational ages ≥35 weeks and to infants with Apgar scores ≥7, found that SUPC was more likely in first-born infants.⁶ 

The highly detailed data source used in the present study allowed us to uncover novel risk factors for SUPC, including room sharing, swaddling, and a caregiver who fell asleep while feeding the infant. Because of differences in which data were collected in our study compared with other prior research, it is unclear if there is overlap between falling asleep while feeding the infant and skin-to-skin contact, a risk factor identified in previous case series of SUPC that included SUPC in birth hospitals. However, unlike previous case series, we found no associations with breastfeeding and smoking. Additionally, although we conducted sensitivity analyses to try to ascertain if specific factors were proxies for infant age, it is still possible that room sharing was higher among SUPC deaths because of the combined high percentages of surface sharing and parents who fell asleep while feeding their infant. In light of previous research supporting room sharing as an important protective factor against SUID, caution should be taken in interpreting this study’s results as suggesting a relationship between room sharing and SUPC.^18,19 Because our study population included only infants after home birth or discharge from the birth hospital, our findings are not directly applicable to concerns about the practice of rooming-in at birth hospitals.⁵ However, our findings are applicable to hospital personnel regarding recognition of parental fatigue, and when, where, and how the infant is sleeping, especially in those parents who are inexperienced in the care of infants and who may require additional attention and guidance while in the birth hospital.

Our sensitivity analyses also sought to determine if swaddling was a proxy for age, rather than a risk factor for SUPC. It is unclear if our sensitivity analyses were able to fully account for the association of swaddling with infant age. Therefore, this finding should be viewed in light of that important limitation. Swaddling is also commonly used in newborns, and many birth hospitals model this practice for parents as a way to soothe infants and promote sleep.^20,21 However, there is an increase for sudden and unexpected death if a swaddled infant is placed or rolls into the prone position.^21,22 Thus, when swaddling is used, infants should always be placed wholly on the back. Additionally, care must be taken when feeding a swaddled infant so that the infant is placed on the back when they fall asleep. Future studies should attempt to ascertain if swaddling is a risk factor for SUPC or merely a proxy for infant age.

Although these findings illuminate possible unique risk factors associated with SUPC (as opposed to SUID at older ages), they do not provide absolute risk. For instance, in the majority of SUPC deaths, the parent was aged 22–34 years, was not primiparous, and the caregiver did not fall asleep while feeding the infant. Additionally, although there were no differences in SUPC and non-SUPC deaths, nearly three quarters of infants dying of SUPC were surface sharing at the time of death, more than half were sleeping in an adult bed, nearly half had not been placed in the supine position for sleep, and one third had never breastfed. Therefore, health care providers should continue to reinforce the recommendations of the American Academy of Pediatrics to reduce sleep-related infant deaths, including SUPC deaths.²³ However, our findings suggest that clinicians should discuss the importance of recognizing when one is getting sleepy while feeding an infant and assuring that the infant is in the safest location possible if and when the parent falls asleep. Having discussions about the safest places for infants to sleep will help tired, first-time parents proactively plan for situations when they feel that they will fall asleep during a feeding.

This study’s novel discoveries are due to its data source and methodological approach. The NFR-CRS allowed us to examine far more potential risk factors than previous research on SUPC. Similar to some previous studies,^6,8 we took an exploratory approach to discovering possible new risk factors rather than testing specific risk factors a priori. Our use of stepwise selection of covariates contains inherent limitations. For instance, stepwise selection can be influenced by the order in which variables are added to the regression model. As a result, we also performed backward selection. Although backward selection resulted in the same variables being included in the final regression model, it remains a limitation to our analytic approach. Other dimensional reduction techniques, including machine learning techniques, should be considered in future research.

There are other limitations that should be considered. Our data source relies on voluntary reporting to the National Center for Fatality Review and Prevention. Not all states provide data, and some states may not report all deaths. For instance, our dataset did not include SUPC deaths occurring in the birth hospital. All SUPC deaths occurred after home birth or discharge from the birth hospital. As a result, it is impossible to eliminate the possibility of selection bias and our findings may not be generalizable to deaths occurring in the hospital setting. Furthermore, we did not have access to Apgar scores, which would have helped to further refine our definition of SUPC deaths. It is also not possible to calculate death rates. The dataset also does not include information on infants who did not die. This may be especially important in the present study because many SUPC-like events may occur in a hospital setting, resulting in successful resuscitation of the infant.

Well-known risk factors, including surface sharing and prone sleep position, are common in SUPC deaths. However, compared with SUID at older ages, SUPC is associated with older and primiparous mothers, swaddling, and when caregivers fall asleep while feeding their infant. Therefore, providers should reinforce all recommendations of the American Academy of Pediatrics for reducing sleep-related deaths, while also helping parents to plan for situations when they feel they will fall asleep during a feeding.

aOR

adjusted odds ratio

CI

confidence interval

NFR-CRS

National Fatality Review Case Reporting System

SUID

sudden and unexpected infant death

SUPC

sudden unexpected postnatal collapse