Sudden unexpected death in infancy (SUDI) is a broad term used for all sudden unexpected infant deaths ranging from those that remain unexplained after a full investigation (unexplained SUDI) to those where a full explanation of the death is found during subsequent investigations (explained SUDI). Sudden infant death syndrome (SIDS) and “cot death” have previously been used to describe the “unexplained SUDI” group where SIDS is defined as “the sudden and unexpected death of an infant under 1 year of age, with onset of the lethal episode apparently occurring during sleep, that remains unexplained after a thorough investigation including performance of a complete autopsy, and review of the circumstances of death and the clinical history” [1]. In developed countries, unexplained SUDI represents the highest proportion of all post-neonatal deaths [2].

In the late 1980s and early 1990s, education programs and campaigns, commonly referred to as the “Back to Sleep” campaigns, were started after several risk factors for SIDS were discovered, most importantly prone (front) sleeping, smoking, bed sharing (particularly in the presence of maternal smoking), and not breastfeeding. These programs were followed by a dramatic decline in unexplained SUDI [3, 4] supporting the idea that such risk factors might be causally related [5]. In New Zealand (NZ), rates of unexplained SUDI have continued to decline, but have always been relatively high compared to other countries [6]. The unexplained SUDI rates per 1000 live births amongst Māori (the indigenous people of NZ) have been 5 times that of non-Māori [7]. However a recent decline in unexplained SUDI and among Māori in NZ has been reported [8], the reasons for which remain to be determined.

The first aim of this study was to determine the extent to which infant care practices for prevention of SUDI are being followed in a NZ community, as much of what is known about trends in practices linked to SUDI reduction in NZ has been derived from surveys conducted in the 1990s and early 2000s. The second aim was to develop a SUDI risk assessment instrument that could be used to identify maternal, infant and household variables predictive of SUDI risk. Such a risk score could be used in identifying cases where additional support for families is needed and might indicate useful points for interventions to target, as well as facilitating comparisons between communities and within communities over time.

Information was collected about the infant care practices of 209 Dunedin, NZ, parents of infants born between June 2009 and February 2011. Participants were families comprising the control group (n =209) of a 4-arm randomized controlled trial (RCT): the Prevention of Overweight in Infancy (POI) study (total n =802). Parents were recruited antenatally from the single maternity hospital servicing Dunedin city. Infants were excluded if they lived outside the study area, were born before full term (36.5 weeks), or if a congenital abnormality or a physical or intellectual disability likely to affect feeding, physical activity or growth was identified. In total, 1458 of 2946 women were eligible to participate in the RCT. After declines (n =511) and post-birth exclusions (n =45), 802 enrolled in the main trial (58% response rate). However only data from the control group (n =209) were included here because the intervention arms received education and support on infant sleep and/or breastfeeding. Thus the control group received standard care, whereas the 3 intervention groups received 1) breastfeeding, activity and complementary feeding education and support, 2) infant sleep education, or 3) both interventions. The RCT study details including group allocation methods have been published [9]. The New Zealand Lower South Regional Ethics Committee approved the study (Project Key: LRS/08/12/063) and all participants gave written informed consent.

The present study demonstrates that the main safe sleep messages for SUDI prevention are highly practiced within this NZ community sample. Risk scores calculated for our population confirmed that the majority (72%) were at either the lowest or slightly elevated risk (SUDI Risk Ratio Score ≤1.5) for SUDI through safe sleep and feeding practices. Importantly, this score identified 5 predictors of SUDI risk related to the mother: young maternal age, low education, not employed in the third trimester of pregnancy, high parity, and high depression scores antenatally. Most of these factors would be expected, but are difficult to modify. However, the increased SUDI risk score amongst mothers with high depression scores on the EPDS antenatally is of particular interest, given the strong association between antenatal and postnatal depression [21], the adverse consequences of depression for quality maternal-infant interactions (reviewed in [22]), and the reported association of postnatal depression with SUDI [23]. Although there is a high degree of health surveillance antenatally in NZ, screening for maternal depression is not routine.

The SUDI risk assessment instrument is, as far as we are aware, unique within the SUDI literature in that it incorporates a range of risk factors without assuming that each has the same impact. It provides a single, generalizable, easily used score for SUDI risk that does not require clinical measurements and can be easily updated as new evidence becomes available through meta-analyses or large studies. Potential applications include its use as a tool for cross-study comparisons (e.g., across different cultures), and for identifying temporal changes (e.g., assessing the impact of public health campaigns).

One similar tool exists, the SIDS risk index score of Conroy and Marks [24] derived from five high risk SUDI sleep practice variables identified from the Confidential Enquiry into Stillbirths and Deaths in Infancy study [25] in a sample of 66 disadvantaged families, the majority of whom were from an ethnic minority group. Bed sharing was not included, and all risk factors were treated as additive, unlike the current study, which has treated the risk factors as multiplicative as the risk estimates were obtained from odds ratios. Two main findings were similar: not being first-born, and higher psychological vulnerability of the mother, were independent predictors of a high SIDS risk index score [24]. Their measure of psychological vulnerability included the EPDS score at 2 months post-partum. In both studies, the EPDS score was treated as a continuum, rather than as a presence or absence of depression defined using clinical cutoffs. Thus, it is the symptoms of depression, rather than clinical depression per se, that the current study and that of Conroy and Marks [24] emphasize as being an important predictor.

Maternal Māori ethnicity was almost statistically significant as a predictor of a high SUDI risk in the unadjusted model (P =0.053), but the association was greatly attenuated after adjusting for other variables (adjusted OR of 1.21 compared to unadjusted OR of 1.79) and was no longer a tendency in the adjusted model. Neither low family income nor household deprivation were independent predictors of SUDI risk.

The prevalence of prone sleeping in our study was low at 1.3% and 2.8% of 3 week and 19 week old infants respectively. Nationwide figures for the prevalence of prone sleeping in 3 month old infants before the “Back to Sleep” campaign were 33% [26], dropping to 3.0% afterwards [27]. The prevalence of side sleeping nationally was 73% after the Back to Sleep campaigns [27]. Side sleeping was later recognized as an additional risk factor for SIDS and education to discourage the practice resulted in a fall in the prevalence with local reports of a prevalence of 21% in 2003 [28]. The present study suggests a further reduction with 12% and 7.6% of infants sleeping on their side at 3 weeks and 19 weeks of age respectively. With the reduction in both side sleeping and prone sleeping, the prevalence of supine sleeping here was 90%, appreciably higher than the 62% reported in 2003 [29]. Major reasons cited for non-supine sleeping positions are infant preference, infant comfort, and parental fear of choking [30]. However, it is clear that the safe sleep position messages prevail within our community.

The prevalence of mothers smoking during pregnancy in this study (8%) is slightly lower than the national average of 11%, although the prevalence is significantly higher nationally amongst Māori women (34%) and in women with lower markers of socioeconomic status [31]. However these groups are not strongly represented in this sample. Meta-analyses published in 1997 [32] and in 2013 [33] concluded that maternal smoking doubles the risk of SIDS. The risks attributed to passive smoking remain unclear, however one study has reported that the risk from postnatal exposure increases with the number of smokers in the household, or with the daily hours the infant is subjected to an environment with cigarette smoke [34]. Encouragingly, in our study, figures for smoking in the car and smoking inside the home were very low.

The advice to not bed share, particularly if the mother smoked in pregnancy or currently, was also being adhered to in this study with only one case of bed sharing in a mother who smoked. However, bed sharing was twice as common at 3 weeks of age as at 7 weeks and older. The reason for this higher rate in younger infants is unknown but may be due to parental choice, maternal sleep needs, or infant feeding practices. A large meta-analysis identified that bed sharing in the absence of maternal smoking was still a risk for SUDI (at 2 weeks, odds ratio =2.4) but was only significant during the first 8 weeks of life [17]. Furthermore, the risk of bed sharing infants dying in a maternity hospital bed within the first few days of birth has received attention [35]. Room sharing with the baby sleeping in its own cot or bassinette for the first 6 months of life is the recommended practice. Anecdotally, some parents say this is difficult to comply with and Figure 1 illustrates the shift in this practice as the infant gets older, with 68% room sharing at 3 weeks declining to 38% at 23 weeks.

Pacifier use at sleep time is associated with a lower SUDI risk through a yet to be determined mechanism [36]. Despite some countries recommending pacifier use for SUDI prevention [37], NZ does not include pacifier use as part of its safe sleep messages. We found pacifier use more than doubled from 3 weeks of age [9.6%] to 15 weeks, [19%], similar to the 19% of 3 month old infants in this region using pacifiers in 2001/2 [29]. This suggests no significant effect of international recommendations on pacifier use in this community, and no broader social effect even though pacifier use is much more common in other countries [32% to 71%] [38].

The suggestion that breastfeeding may be protective against SUDI has been controversial, with many, although not all, studies reporting an association [39]. However, a recent meta-analysis suggests that breastfeeding is protective, and that the effect is even stronger when breastfeeding is exclusive [19]. Eighty-nine percent of infants in the current study were breastfed at 2 months of age, compared to 79% of NZ infants aged approximately 6–9 weeks [40]. Although exclusive breastfeeding is likely to provide greater protection against SUDI [19], we used “any breastfeeding” at 2 months as the marker for protective breastfeeding behaviour. This was because the only estimate available in the literature of the relative risk of SUDI amongst infants who are breastfed was for “any breastfeeding” at 2 months [19].

The strengths of this study include the comprehensive longitudinal dataset, and the availability of earlier data with which to compare and describe changes in child-care practices over time. When exploring independent predictors for SUDI risk scores, this study was able to take into account a wide range of maternal infant and family characteristics. Limitations include, first, the respondents were from areas of low SUDI risk and predominantly Caucasian, thus we don’t know how well the score would translate to a study of high-risk infants. Second, the SUDI risk assessment instrument included some unadjusted odds ratios. As data with adjusted odds-ratios for all variables become available from future meta-analyses or high-quality studies, a better estimate of the risk will be possible. Third, the calculation of the total SUDI risk score assumes multiplicative and not additive risks. While this seems appropriate given that the score is based on odds ratios, which are multiplicative in nature, we don’t know if this is the case for all factors considered here and it is possible that correlations between risk factors not adjusted for in the analyses found in the literature may lead to very high risk estimates for some families. Again, further calibration of the instrument will be possible when better estimates become available. Finally, missing data limited the calculation of risk scores to 176 of the 209 participants recruited into this arm of the study (84.2%). Missing alcohol data reduced the number of observations available for that unadjusted model by a further 36 and biases in answering this question may have affected the association reported here. However, as AUDIT-C was not included in the adjusted model, this did not substantially affect the sample size available for that model which was reduced by only 6 (3.4%) due to missing covariates. It seems plausible that missing risk score data would be largely missing at random after conditioning on the covariates included in the adjusted model. However, the presence of missing data increased the widths of the presented confidence intervals and some non-statistically significant predictors (ethnicity and income) have confidence intervals which do not rule out important associations.

This study has identified groups within the community as priorities for education about safe sleep practices beyond standard care: young mothers, mothers with high parity, mothers with low educational levels, and mothers with symptoms of depression antenatally. The antenatal period is an opportune time to screen for symptoms of depression, providing the chance for early intervention and treatment before birth. We reinforce earlier data suggesting maternal depression is a marker for SUDI risk [23] and within the context of our findings, related to infant care practices. Further research is required to determine whether the findings also apply to infants at high risk of SUDI, particularly in ethnic minority populations, with different cultural infant care practices.