Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: Soc Sci Med. 2020 Dec 6;269:113585. doi: 10.1016/j.socscimed.2020.113585

SOCIAL NETWORK INFLUENCES ON NEW MOTHERS’ INFANT SLEEP ADJUSTMENTS

Benjamin Cornwell 1, Xuewen Yan 2, Rebecca F Carlin 3, Linda Fu 4, Jichuan Wang 5, Rachel Y Moon 6
PMCID: PMC7794038  NIHMSID: NIHMS1654812  PMID: 33333376

Abstract

Despite public awareness campaigns, some parents continue to engage in infant sleep practices that are considered risky by health experts, such as bedsharing or placing their infants on their stomachs. This study examines the role parents’ social networks play in shaping their responsiveness to new information and/or suggestions about how they should place their infants for sleep, paying attention to the respective effects of health professionals and their close interpersonal ties. We collected data from a sample of 323 new mothers in Washington, D.C., where they reported their infant sleep practices and perceived personal social networks. We find evidence that mothers’ social networks play a significant role in the likelihood that they adjust their infant sleep practices within the first few months of their infants’ lives. Mothers are more likely to change sleep practices when health professionals and/or (lay) family members advise them to do so. The influence of network members is not always positive. For mothers endorsing initially safe practices, their probability of change increases if their network members substantially espouse unsafe practices. Among mothers with initially unsafe practices, network members’ level of support for safe sleep practices is not predictive of the likelihood of sleep practice change. Implications for potential interventions are discussed.

Keywords: social networks, network influence, parenthood, infant sleep practices, sudden infant death syndrome, behavioral change

Introduction

The leading cause of death among infants between one month and one year of age in Western countries is sudden unexpected infant death (SUID). Approximately 3,600 deaths per year in the U.S. are classified as SUIDs, attributable to the more widely known sudden infant death syndrome (SIDS), as well as accidental suffocation and strangulation in bed (ASSB), and other ill-defined causes of death (Centers for Disease Control and Prevention 2019). These generally occur during sleep or in a sleep environment and are referred to as sleep-related infant deaths. Furthermore, there are significant racial/ethnic disparities in the prevalence of such sleep-related deaths (Hauck et al. 2002; Kroll et al. 2018; Parks, Erck Lambert, and Shapiro-Mendoza 2017). The incidence of SUIDs has barely declined since 1999, though there is state-level variation in this trend (see Erck Lambert, Parks, and Shapiro-Mendoza 2018). And, despite increased knowledge and public messaging campaigns for safe infant sleep practices (see Colson et al. 2005; Moon 2016; Ottolini et al. 1999; Rasnski et al. 2003), there is evidence that the proportion of parents who place infants on their stomachs (“prone”) and in other high-risk sleep environments (e.g., bedsharing) has increased in recent years (Colson et al. 2009, 2013; Shapiro-Mondoza et al. 2015; Willinger et al. 2003).

To design effective interventions that increase safe infant sleep practices, we need to better understand how parents make decisions regarding how they put their babies to sleep. The notion that parents respond to information from health-related experts (e.g., nurses, pediatricians) and institutions has led to public messaging campaigns, such as the “Back to Sleep” campaign of the mid-1990s, whose impact on infant sleep practices appears to have plateaued (Moon 2011). But in recognizing that many professional care providers themselves do not consistently model safe infant sleep practices (Moon, Calabrese, and Aird 2008; Moon and Oden 2003; Price et al. 2008), and that parents are apt to mimic these professionals’ behavior (e.g., Brenner et al. 1998), some campaigns have focused on educating health professionals such as hospital maternity ward personnel (Colson and Joslin 2002; Kellams et al. 2017; Macklin et al. 2019; Mason et al. 2013; Moon, Hauck, and Colson 2016).

Likewise, there is increasing recognition among scholars that parents’ adherence to the advice of people who are not health professionals can also lead to unsafe infant sleep practices. This had led to a shift in the focus of research toward understanding more informal channels of influence on parents’ infant sleep practices; namely, parents’ personal social networks. It is argued that one reason parents’ infant sleep practices are not easily altered via education is that parents are not always swayed by (and are sometimes skeptical of) scientific/professional sources of information and are in some cases more open to advice from family, friends, neighbors, and co-workers (Moon et al. 2019; c.f., Hwang et al. 2016).

There are several reasons why these contacts influence new parents’ infant sleep practices, even when they have received professional advice to the contrary. One’s close social ties, or personal social network members provide a reference group (Merton 1968) by which one perceives and defines norms of common and acceptable behavior (e.g., see Schultz et al. 2007). A given person’s network members have the capacity to informally socially sanction one’s behavior that falls outside the boundaries of these norms, especially when those network members are connected to each other (see Coleman 1988). They also furnish role models, to whom people look when deciding how to behave in uncertain or fraught situations (e.g., when a baby continues to cry late into the night). Social network members constitute sources of advice and information about best practices and other behaviors in these situations.

Finally, one’s close interpersonal contacts are often embedded in a broader community that exercises certain cultural practices (Crane, Denise, and Ball. 2016), some of which may involve unsafe infant sleep practices such as use of thick bedding and bedsharing (see Moon, Hauck and Colson 2016). It has already been found that social networks influence a wide variety of health outcomes and health-related behaviors, such as obesity (Christakis and Fowler 2007; De la Haye et a. 2010), risky sexual behavior (Choi and Gerorich 2009; Scheider et al. 2013), alcohol intake among college students (Polonec, Major, and Atwood 2006; Reifman, Watson, an McCourt 2006), and needle sharing among drug users (Davey-Rothwell and Latkin 2007).

In a similar vein, parenting practices such as breastfeeding initiation and continuation (Swanson and Power 2005) and vaccinating children (Brunson 2013) are also likely shaped by personal network members’ practices and network norms. Indeed, research suggests that new parents’ infant sleep practices tend to parallel the beliefs and practices of their close personal contacts. A recent study of mothers in Washington, D.C. (Moon et al. 2019) shows that those whose kin and friends reportedly advocated supine positioning, room sharing without bedsharing, and avoidance of soft bedding were themselves much more likely to engage in these practices. The study hypothesizes that social network members can exert normative influences on mothers’ infant sleep practices. With regard to advice about sleep practices, several studies find that some mothers consider their personal social networks to be a more trustworthy resource than medical providers (Colson et al. 2005, 2006; Epstein and Jolly 2009; Oden et al. 2010). These studies note that professional providers’ advice may be less trusted, especially when mothers are embedded in a social environment different than the white middle-class culture represented by health professionals. In general, it appears that when the norms of the social network are contrary to recommendations, they can be major barriers to safe and healthy infant care practices.

While the findings above are suggestive of the role personal network members could play in parents’ infant sleep practices, we believe they can be complemented in two ways. First, research so far has relied on a cross-sectional design, and yet the influence of networks shall, by definition, occur over time. In other words, we lack a longitudinal perspective that would allow us to directly ascertain whether the prior configurations of parents’ interpersonal ties are predictive of their subsequent infant sleep practices. While such a longitudinal design by itself cannot establish causality, it sheds significant light on the temporal sequence and process of the observed correlation between parents’ personal networks and their infant sleep practices.

Second, despite the recognition that norms of personal social networks may contradict professional recommendations, no study to our knowledge has explicitly gauged the relative importance of health professionals versus personal ties in parents’ infant sleep choices. As such, we are yet to understand how advice from personal social network members and health professionals alike might combine to influence parents’ infant sleep practices. The latter is a noteworthy possibility to explore, especially given evidence of such a synergistic effect for some other health-related behaviors (e.g., Shin and Shim 2019).

In this paper, we view social networks as broadly involving all types of contacts new mothers regularly interact with—including friends and family members in their personal social network as well as health professionals. Following the above two leads, our study will examine new mothers’ propensity to actually change their infant sleep practices during the first few months of their baby’s life in response to the influences from social networks. Using two waves of longitudinal survey data, our goal is to assess whether and to what extent the advice and practices of mothers’ personal social network members—relative to those of health providers—shape the likelihood that mothers will alter infant sleep practices, especially as they pertain to sleep position and bedsharing.

Materials and Methods

Survey design

This study uses data from a study of mothers of young infants in the Washington, D.C. region. To enroll a sample with sufficient numbers of mothers who were adherent and non-adherent to safe infant sleep recommendations, the study team used a combination of hospital-based and chain-referral recruitment strategies. Because this was expected to be a relatively small sample with short interviews, the team employed restrictions involving immigrant status and race. In terms of immigrant status, we only recruited mothers with US-born parents in order to reduce unobserved heterogeneity that could be due to cultural differences in infant sleep practices. Regarding race, we target Black or White mothers to capture racial variability in our study. On the one hand, the Washington D.C. population are predominantly White or Black (US Census 2019). On the other, we consider Black mothers an important demographic to include because existing research has shown higher incidences of SIDS among African American communities (Colson et al. 2005, 2009). Beyond this, potential participants were excluded from participation if: (1) they were not the custodial parent (e.g., infant in foster care); (2) they were unable to complete an interview in English; (3) their infant had a condition precluding use of the supine sleeping position (e.g., recent spinal surgery); or (4) their infant was born prematurely (< 36 weeks gestation). Overall, mothers were eligible to participate if they were ⩾ 18 years old, English-speaking, self-identified as Black or White, and had U.S.-born parents.

Recruitment focused on mothers of infants who were younger than six months of age. (For consistency, we refer to these as “new” mothers, regardless of whether they have older children as well.) This resulted in an initial pool of 405 respondents who were custodial mothers of young infants, within which 345 were mothers of younger-than-three-month infants, and 66 were mothers of infants of three-to-six months.

At baseline, these respondents completed a staff-administered survey that addressed new mothers’ infant sleep practices, the composition of their personal social networks, and basic socio-demographic information (described below). Roughly two and a half months after the baseline interview, the team re-contacted the 345 mothers whose infants were younger than three months old at the time of the initial survey. These respondents were invited to complete a 5–10 minute follow-up telephone survey, which elicited additional information about any changes in mothers’ infant sleep and other child care practices. A total of 327 mothers of the younger infants agreed to participate in this interview, yielding a 94.5% cooperation rate for the follow-up survey. This study was approved by the institutional review boards of Children’s National Medical Center, MedStar Washington Hospital Center, and the University of Virginia.

Dependent variable

As previously noted, our interest lies in overtime changes in mothers’ infant sleep practices. Therefore, instead of modeling infant sleep practices, we construct a dichotomous dependent variable that indicates whether the mother in question made any changes between the initial survey and the follow-up survey with respect to either of two infant sleep practices: (1) sleeping position (non-back versus back [recommended]); and (2) bedsharing (sleeping on same sleep surface with parents versus not sleeping with parents [recommended]). In other words, those who changed either infant sleep practice are coded as 1 for the outcome, regardless of whether the adjustment occurred toward safe or unrecommended practices. Importantly, while our dependent variable ignores the safety direction of practice changes, the actual safety status at the follow-up stage remains a relevant part of our analysis: our models will consistently condition on mothers’ adoption of safe infant sleep practices at the initial survey.

Social network data

The operationalization of our key independent variables relies extensively on personal, or “egocentric” social network data. By a personal or egocentric social network, we mean in this paper a network involving the respondent of interest (“ego”) and all the contacts she has close and strong interpersonal ties with (“alters,” or “network members”). More specifically for our survey, we follow the “realist” approach to egocentric social networks (Laumann et al. 1983), where we rely on respondents’ reports or perceptions to specify the characteristics of their personal networks. This approach serves our purpose well, as research has shown that it is egos’ subjective perceptions of their networks that tend to actually influence their behavior (Krackhardt 1987).

Our network survey questions drew substantially, in terms of scope and content, on the social network modules of the National Social Life, Health, and Aging Project and the General Social Survey (see Cornwell et al. 2009; Marsden 1987). The majority of our personal social network data were collected at baseline, where new mothers completed a staff-administered survey that used established, validated questions about their close interpersonal contacts.

For starters, mothers were asked to list the first names or initials of up to eight people with whom they “interact with on a regular basis” and “most often discuss things that are important to” them. In the remainder of the network module, we collected data about two main aspects of new mothers’ personal social networks. The first involves a set of common questions about egocentric networks: each network member’s socio-demographic traits, the type of relationship they have to ego (open-ended description), their level of connectivity (1 = once/year to 5 = daily) with ego and between each other.

The second component directly elicits information about network members’ practice of infant sleep. Regarding sleep position and bedsharing, respondents were presented with two statements describing unrecommended practices (“Have ever placed babies on their side or stomach for sleep”; “Have slept with babies in their bed before”) and asked whether, for each alter sequentially, the latter engaged in any of these practices. In case a new mother does not know about a specific network member’s practices, our questionnaire also provided the option of “don’t know”, although we note that no respondents in our survey chose that option for either the sleep position or bedsharing question. One potential issue here is how new mothers will evaluate network members who do not have infants. Unfortunately, we do not have data on alters’ parenting history. However, we suspect this is not a serious problem, given the overwhelmingly more elderly age of the network members (egos are on average 37.1 years younger than the mean age of their alters).

During the follow-up survey, the above questions about respondents’ personal social networks were not repeated. Instead, the network module asked respondents, for each aspect of infant sleep practices (e.g., back sleep, bedsharing), whether anyone (this may or may not be a network member as defined in the baseline survey) has suggested that they change their behavior since they were last interviewed, and if yes, who. For the latter question, respondents provided free-text responses describing the person’s relationship to them, such as “doctor”, “baby’s grandmother”, or “friend from the church”. Also, the survey allows respondents to mention up to four people who advised behavioral change, and the “who” question was required each time.

Independent variables

Social network predictors.

Our main explanatory variables relate to the characteristics of the egocentric social network in which a new mother is embedded, which potentially influenced her decision-making regarding infant sleep practices. We measure several features of these networks. First, we specify a variable “alter-ego behavior difference” that quantifies the proportion of alters in egos’ networks whom respondents perceived as engaging in infant sleep practices that were different from their own at baseline. Our theoretical intuition for constructing this variable is as follows: if new mothers indeed model or mimic their alters’ infant sleep practices over time, their inclination to change their behavior during the follow-up period may be larger when their initial practices deviated more substantially from their alters’.

To compute the alter-ego behavior difference score, we draw on respondents’ own practices and the second part of the network module in the initial survey. As already discussed, the survey asked new mothers to assess whether each of their alters implemented unsafe infant sleep practices (respectively non-back sleep and bedsharing, 1 = yes, 0 = no for individual alters). We first obtain the mean values of all alters for each practice to derive the proportion of a new mother’s alters using an unsafe practice, and then averaged these means for a combined estimate for sleep position and bedsharing (Cronbach’s α = .62). Then, for mothers initially adopting safe infant sleeping practices (defined as using both back sleep and non-bedsharing), this proportion is their alter-ego behavior difference score. For mothers starting with unsafe practices (defined as using either non-back sleep or bedsharing), we determine the final difference scores by subtracting the corresponding proportion from 1. For example, when 60% of a respondent’s alters engage in unsafe practices, a mother will have a score of .60 for alter-ego behavior difference if she adopted recommended practices in the initial survey period, and .40 if her practices were unsafe.

We also create a set of indicators that capture who, if anyone, suggested to the respondent between the initial and follow-up survey that she change her infant sleep practices at some point, using information from the network module of the follow-up survey (described earlier). We categorize new mothers’ respondent’s free-text responses regarding their advice givers into three types: (1) health professional; (2) kin; and (3) anyone else (e.g., friend, neighbor, co-worker), and for each type, we construct a dummy variable measuring whether ego received advice from the corresponding source type (1 = yes, 0 = no). Note that our coding strategy is only one of the several possible ways of constructing the desired variables, because respondents could name up to four advice givers. Our substantive results remain unaltered, when we test an alternative approach of constructing mutually exclusive measures (those who only received kin’s advice, only health professionals’ advice, mixed sources of advice, or none). Also, similar to our measure of the dependent variable, these dummy indicators combine information about sleep position and bedsharing, and are coded for a given category if at least one of the two sleeping practices fulfills the criteria.

Key control variable.

An essential control variable is the degree to which the mother’s infant sleep practices were safe to begin with. This is a dichotomous variable that takes the value of 1 if a mother initially implements recommended practices for both sleep position and bedsharing (0 otherwise). Conditioning on initial safety allows us to make inferences about the direction of the behavioral change. For instance, a positive coefficient for the main effect of initial safety practice would imply that mothers were more likely to shift from recommended to unsafe practices, versus the other way around.

Other control variables.

We also control for several relevant structural and individual-level characteristics. For one, we include network size, because people who have larger networks have more potential influences. On the assumption that tie strength and past family practices can shape network members’ influence on mothers, we include a measure of the proportion of alters who are kin (by blood or marriage). Because network members are generally more capable of exercising influence when they are interconnected (e.g., see Coleman 1988), we also control for the density of the respondent’s network (calculated as the proportion of alter-alter dyads in the network who interact with each other). Finally, we add respondents’ demographic attributes that were previously shown to be associated with women’s child health outcomes: respondent’s age (in decades) (Molina-García et al. 2019), educational attainment (college or less than college) (Desai et al. 1998, Keats 2018), race (White or African-American) (Colson et al. 2005, 2009), and whether they are first-time mothers (Aston 2002).

Analytic strategy

The aim of our study is to explore the respective roles of personal social networks and health professionals in affecting new mothers’ probability of changing their infant sleep practices over time. In what follows, our analyses will revolve around two aspects. First, we examine whether receiving suggestions from kin, health professionals, or other types of contacts (measured by three dummy predictors) increases the likelihood of behavioral adjustments. Second, we test whether more substantial deviance from network members’ norms of practices (measured by ego-alter behavior difference scores) is also associated with higher propensity to shift one’s infant sleep practices.

We first conduct bivariate analyses, using t-tests or chi-square tests for statistical inference. We then move to multivariate logistic regression to tease out the independent effect of each main predictors.

Results

Descriptive statistics regarding the main variables of interest are presented in Table 1. Nearly one-quarter (22.0%) of mothers in some way changed their infant sleep practices between the initial survey and the follow up. As for the direction of the change, 6.8% shifted from unrecommended to safe practices, while 9.3% converted their infant sleep practices in the opposite direction.

Table 1.

Descriptions of Key Variables Used in the Analysesa

Variable Mean or Proportion S.D.
Dependent variable
Overtime change in sleep practices R changed at least one of the two practices for her baby: sleep position (back versus non-back) and bedsharing (no versus yes), between Waves 1 and 2. {1= Yes, 0 =No} .22 .41
Independent variables
Health professional suggested change R received suggestions to change infant sleeping practice from a health professional between Waves 1 and 2. {1=Yes, 0=No} .11 .31
Kin suggested change R received suggestions to change infant sleeping practice from kin between Waves 1 and 2. {1=Yes, 0=No} .09 .28
Others suggested change R received suggestions to change infant sleeping practice from anyone other than kin and health professional between Waves 1 and 2. {1=Yes, 0=No} .05 .22
Alters’ use of unsafe practices at baseline Average of 3 standardized items (a = .62) each representing the proportion of R’s alters who, in R’s assessment, implement unsafe practice of sleep position or bedsharing. (e.g., “Have placed babies on their side or stomach before.”) Range: 0 to 1. Range: 0 to 1. .09 .14
Alter-ego difference Proportion of R’s alters whose own practice about safe infant sleeping were contradictory to ego’s practice at Wave 1. For Rs implementing safe practices at Wave 1, this equals the variable “alters’ unsafe practices”. For Rs implementing unsafe practices at Wave 1, this equals one minus that variable. Range: 0 to 1. .28 .30
Initial safety status R adopts safe infant sleeping practices, using both back position and non-bedsharing at Wave 1. {1 = Yes, 0 =No} .74 .44
Proportion kin Proportion of R’s alters who are kin. Range: 0 to 1. .62 .28
Network density Proportion of R’s network members who know each other. Range: 0 to 1. .61 .31
Network size Number of R’s alters. Range: 1 to 8. 4.49 1.24
Age (in decades) R’s age at Wave 1 divided by 10. Range: 1.8 to 4.6. 2.90 .60
College R attended at least some college {1 =Yes, 0 =No} .47 .50
Married R is currently married {1 =Yes, 0 =No} .48 .50
White R is White {1 = Yes, 0 = African American} .29 .46
First time mother The infant is R’s first born {1 = Yes, 0 = No} .58 .49
Open in a new tab
a

Estimates are restricted to respondents who participated in both Waves 1 and 2 surveys.

Bivariate analysis of any behavior change

Table 2 presents results from bivariate analyses of associations between key predictors and the measure of behavior change. This table provides preliminary evidence that being advised to change one’s practices by either health professionals or kin significantly increases a mother’s probability of altering her infant sleep practices. Among mothers who did not receive suggestions from a health professional about approaching their baby’s sleep differently, 20.3% modified their sleep practices – substantially fewer than those who did receive suggestions from a health professional (52.9%; χ2 = 10.0, p < .01). Similarly, absent suggestions from kin, 18.6% of respondents altered their practice, in contrast to the corresponding figure of 44.2% among cases where kin did suggest a change (χ2 = 14.3, p < .001). This pattern is not evident for sources of suggestions other than kin or health professionals (23.5% vs 21.9%; χ2 = .03, p = .87).

Table 2.

Differences in Rates of Change in Infant Sleeping Practice between Waves 1 and 2, by Social Network Featuresa

Overtime change
 Test statistic N
Yes No
Health professional suggested change 52.9% 20.3% χ2 = 10.0** 323
Kin suggested change 44.2% 18.6% χ2 = 14.3*** 323
Anyone else suggested change 23.5% 21.9% χ2 = 0.03 323
Alter-ego behavior difference M = 45.5% M = 23.6% t = − 5.58*** 318
Proportion kin M = 69.4% M = 59.6% t = − 2.65** 323
Open in a new tab
*

p < .05,

**

p < .01,

***

p < .001 (two-tailed tests)

Note: For rows representing sources of suggestions, the “Yes” versus “No” columns represent the proportion of mothers who shifted their behavior, depending on whether they received suggestions from a given source. For the rows on behavioral differences and proportion kin, the “Yes” versus “No” columns indicate the mean score of the corresponding row, grouped by whether mothers changed their infant sleep practices.

Two-sample t-tests also provide support for the hypothesis that mothers are significantly more likely to change their practices when they are embedded in a social network in which alters conduct practices that are different from their own. Among mothers who changed their behavior, 45.5% of their alters, on average, had been using sleeping practices contradictory to their own initial practices. Among mothers who did not change their behavior, the figure is an average of 23.6%. (t = −5.58, p < .001). Finally, note the positive association between the proportion of kin in a mother’s social network and her probability of changing her infant sleep practices. Those who did not change their behavior reported that, on average, 59.6% of their network members were kin, which is 9.8% lower than the average proportion of kin among mothers who changed their initial infant sleep practices (t = −2.65, p < .01).

Multivariate analysis of behavior change

Results from multivariate logistic regression analyses (reported in Table 3) confirm the potential relevance of receiving network members’ suggestions to mothers’ behavioral change in infant sleeping practices. Coefficients are shown on the log-odds scale. We note first that initial safety status emerges as the strongest predictor of sleep practice adjustment in the model. Net of covariates, using safe practices in Wave 1 decreases the odds of overtime change by 400% (e1.38 − 1). Equivalently, holding all other predictors at their average values, 11.9% of mothers are expected to shift from safe to unsafe practices, whereas the proportion is 40.2% if the change were to occur towards the direction of safety.

Table 3.

Coefficients from Logistic Regression Analyses Predicting Whether the Respondent Changed Her Baby’s Sleeping Practice between Waves 1 and 2a

Predictor Model 1 Model 2 Model 3 Model 4
Health professional suggested change 1.382* 1.297 1.474* 1.472*
(.647) (.708) (.679) (.694)
Kin suggested change .980* .774 .997* .842
(.396) (.426) (.434) (.447)
Anyone else suggested change .058 −.026 .300 .276
(.665) .678) (.714) (.724)
Alter-ego behavior difference .561 .934 .629 −.973
(.679) .865) (.617) (.817)
Initial safety status (ref = unsafe) −1.608*** −3.017*** −1.372*** −2.825***
(.473) .693) (.416) (.640)
Proportion kin .940 1.051 .608 .573
(.589) (.618) (.771) (.786)
Initial safety status × Alter-ego behavior difference 3.670** 3.687**
(1.323) (1.229)
Network density −.141 −.035
(.614) (.625)
Age (in decades) −.026 −.032
(.039) (.040)
College (ref = no college) −.187 −.320
(.488) (.509)
White (ref = African American) −.814 −.777
(.559) (.567)
Network size .088 .118
(.140) (.145)
Married (ref = unmarried) .067 .159
(.446) (.462)
First time mother (ref = no) .169 (.372)
Intercept −1.331 −0.435 −.762 .306
(.685) (.695) (1.569) (1.638)
N 318 318 314 314
Tjur’s R2 .20 .22 .21 .24
AIC 283.1 276.1 286.7 279.5
BIC 309.4 306.2 335.5 332.0
Open in a new tab
a

Standard error in parentheses

Note: We use an intuitive R2 measure for logistic regression proposed by Tjur (2009), which calculates the difference between the means of the predicted probabilities for each of the two categories in the dependent variable. For example, here, for all cases with actual overtime change, Model 1 predicts the probability of change to be .366, and the corresponding number is .169 for cases without change, and therefore, the Tjur’s R2 is .20 (= 0.366 – .169).

Model 1 shows the additive effects of the main social network predictors. First, we note that there is a significant association between behavior change and the reception of suggestions from all sources combined (χ2 = 10.2; p < .05). More specifically, having received suggestions from a health professional or from kin significantly increases a mother’s likelihood of changing her baby’s sleeping position or bedsharing (p < .05, CI = 0.22 ~ 2.55; p < .05, CI = 0.19 ~ 1.77, respectively). For example, a mother who received advice about behavioral change from health professionals is four times (e1.38) more likely to change her infant sleep practices than a mother who did not receive such advice from a health professional. Figure 1 plots the predicted probabilities of behavior change that are associated with different sources of advice. Holding all covariates at their mean values, the average increase in the probability of adjusting infant sleep practices is 17.1% when mothers were advised by kin to change their practices, compared to 27.1% when the source of advice was a health professional, and less than 1% for those who received suggestions from other sources.

Figure 1:

Figure 1:

Open in a new tab

Predicted Probabilities of Change in New Mothers’ Infant Sleep Practices, by Source of Advice for Change

Note: In the prediction equation, all covariates are held at their mean. The coefficients from model 1 (n = 318) are used.

The proportion of network members who are kin is not a significant predictor of change in in infant sleep practices in any of the models. But the models show that alter-ego behavioral differences matter under particular conditions. The non-significance of the variable in model 1 suggests that the bivariate finding that mothers tend to change infant sleep practices as they deviate more strongly from their alters’ practices is spurious. Additional tests (available upon request) indicate that the inclusion of initial safety status in the model explains the bivariate association. Without the predictor of initial safety status, alter-ego differences in practices prove a significant and positive (b= 2.11, p < .001, CI = 1.20 ~ 3.02) predictor of mothers’ probability of adjusting her infant sleep practices.

In Model 2, we add an interaction between ego-alter difference and initial safety. The model reveals a large and significant interaction effect (b = 3.67, p <. 01, CI = 1.28 ~ 6.06), as well as an improvement in model fit in comparison with Model 1. This confirms that the association between ego-alters’ differences in infant sleep practices and the likelihood of ego changing her practices depends on whether she initially engaged in safer infant sleep practices from the start. This interaction is depicted in Figure 2, which plots the predicted probability of mothers changing their infant sleep practices between interviews among those who initially engaged in safe infant sleep practices. The association is curvilinear upwards, suggesting that the influence of network members with initially contradictory practices is particularly strong when mothers perceive an overwhelming majority of such behavior among their network members. The slope is not significant for those who initially engaged in unsafe infant sleep practices (not shown).

Figure 2.

Figure 2.

Open in a new tab

Predicted Probability of Change in Infant Sleep Practices among Mothers who Initially Reported Safe Practices

Note: Alter-ego differences in practice is measured by the proportion of alters who implements contradictory practices from ego’s in Wave 1. The coefficients from model 2 (n = 318) are used.

Models 3 and 4 are full models for Models 1 and 2, respectively, in that they also include sociodemographic and additional network-related control variables. The directions and magnitudes of our main explanatory variables remain similar for both the additive and interaction-effect models. Judging by both the BIC and AIC metrics, model fit is not improved after the inclusion of additional controls, meaning that the network features considered in Models 1 and 2 likely comprise the most important predictors for overtime change.

Robustness check

The coefficients that are generated in a given model are sometimes sensitive to model specification, or changes in the control variables added to the model (e.g., endogeneity due to omitted variable bias). Therefore, we conducted robustness checks for our main predictors in Tables 1 and 2 using a method developed by Young and Holsteen (2015). For a given predictor, estimated coefficients for all the possible model specifications using all combinations of predictors are calculated, yielding a distribution of all estimates. Results from this analysis suggest that the findings regarding the significance of key predictors in our analyses are robust to different combinations of covariates.

While the robustness check was conducted for all main predictors, we present here one particularly important result – the positive interaction effect between alter-ego differences and mothers’ initial safety status. The figure plots the distribution of the estimates for the interaction term, calculated from 1,024 (= 210) models that exhaust the possible combinations of the 10 control variables in Model 4. The vertical line identifies the coefficient estimate from Model 2, which is our preferred model. For all model specifications, the direction of the coefficient is consistently larger than 3, meaning that the association is always strongly positive (see Appendix Figure A1). But note that the magnitude of the effect is not unimodal. If the true parameter falls to the right of the vertical line, our coefficient from Model 2 may be an overestimation. The substantive interpretation, however, remains robust to model uncertainty concerns.

Conclusion and Discussion

New parents’ infant sleep practices play a major role in shaping risk of sudden unexpected infant death. This study provides some insight into some of the informal social mechanisms by which infant sleep practices are shaped in those first fragile months of babies’ lives. We find that while the majority of new mothers engage in safe infant sleep practices from the start, nearly one-quarter of them changed their practices in some way over a three-month follow-up period.

The prospect of change in infant sleep practices during this early period is associated with some key social network factors. For one, mothers who received suggestions from either health professionals or kin were more likely to change their infant sleep practices. These appear to be separate (additive) sources of change in mothers’ behavior. To extrapolate, this suggests that interventions aimed at changing mothers’ infant sleep practices may be more successful if educational messages are relayed not just via health professionals, but also via mothers’ personal social network members. Changing mothers’ network members’ beliefs may reduce unrecommended infant sleep practices by indirectly shaping mothers’ perceptions of acceptable and normal practices. A related approach would be to develop group-level interventions that include mothers and their family and friends learning together from health care professionals. Research suggests that people respond more positively to professional advice and educational messaging when they come from multiple, disparate sources (see Shin and Shim, 2019). When health care professionals, family and friends consistently recommend safe sleep practices, the effect is additive (Von Kohorn et al. 2019). This is also consistent with recent research that shows that mothers who have more expansive networks (as opposed to restricted networks that include mainly close-knit kin) tend to adopt safer infant sleep practices (Moon et al. 2019). However, research that actually examines the effectiveness of in-person interventions is needed.

One expectable finding from our analyses is that most new mothers are not likely to change their infant sleep practices if they initially engaged in safe infant sleep practices from the start. But among these mothers, behavior change was significantly more likely among those for whom a vast majority of their network members engaged in unsafe infant sleep practices. This suggests that the notion that interventions should only target mothers who engage in unsafe practices from the start is limited. Public messaging campaigns should also encourage mothers who engage in safe infant sleep practices to “stay the course” despite messages to the contrary from their closest social contacts.

This study has a number of limitations that are important to note. For one, our data rely on parents’ reports of their social network members’ beliefs and behaviors, which may invite perception bias. Consistent with other research on network effects (e.g., Martens et al. 2006), we believe that mothers’ perceptions of network members’ behaviors are inherently influential, regardless of what network members actually believe or do. Likewise, we have no ability to assess the extent of social desirability bias among mothers who may not want to admit that they engage in unsafe sleep practices.

Another issue is that while the follow-up survey adds a longitudinal element to the data, we cannot firmly establish causality. It is possible that one’s propensity to change parenting practices are connected to one’s willingness to listen to certain types of contacts in the first place, which would create dual causation. Likewise, it is possible that mothers who engage in more unsafe infant sleep practices were less likely to be surveyed. Some other limitations include the small sample size, our inability to generalize too far beyond this sample, and the fact that we were only able to collect (perceived) data about a relatively small number of mothers’ social network members.

Despite the limitations of this study, our findings provide some insight into potential informal social network bases of influence on mothers’ infant sleep practices. We encourage researchers and practitioners to continue to explore mothers’ networks as potential targets for intervention. In particular, we suggest two directions for future research. First, it will be worthwhile to investigate whether specific attributes or types of network members can be especially predictive of mothers’ choice of infant sleep practices. For example, do the gender, race, or age of a given network member matter? Or does the level of influence depend on the type of relationship that new mothers have to the alter? Compared with our study that used new mothers or egos as our unit of analysis, future research conducting alter-level analysis will be ideally suited to answer these questions.

Second, an important question that our study left unanswered is potential racial differences in network effect on mothers’ infant sleep practices. In our study, we did not detect Black-White differences in infant sleep adjustments net of network effects, but due to sample size limits, we were unable to robustly fit interaction effects to test whether the relationship between networks and infant sleep practices differ by race. Particularly in view of the relative prevalence of SIDS among African American communities, future research with better data should be interested to survey this possibility.

Highlights.

Social networks affect mothers’ propensity to adjust infant sleep practices.

Mothers tend to adjust their practices when health providers advise them to do so.

Mothers also tend to adjust their practices in response to family members’ advice.

The influence of personal network members may be negative for infants’ sleep safety.

Acknowledgments

The data collection and research were funded by the National Institute on Minority Health and Health Disparities (1R01MD007702). The analysis, interpretation of data, writing the report, and the decision to submit the report for publication are of the authors only.

We thank Erin York Cornwell and Alec McGail for reading early drafts of this paper and providing valuable feedback. We are also indebted to Yao Iris Cheng, Anita Mathews, and Rosalind P. Oden for their work on the survey.

Appendix

Figure A1.

Figure A1.

Open in a new tab

Model Robustness for the Interaction Term between Alter-Ego Differences and Initial Safety Status

Note: The graph shows the distribution of estimated coefficients for the interaction term across 1,024 models with all possible combinations of the 10 control variables in Model 4 (respondent n = 314). The vertical line represents our preferred estimate for the interaction term according to Model 2.

Footnotes

Conflicts of interest

The authors declare no conflict of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Contributor Information

Benjamin Cornwell, Department of Sociology, Cornell University, Ithaca, NY.

Xuewen Yan, Department of Sociology, Cornell University, Ithaca, NY.

Rebecca F. Carlin, Goldberg Center for Community Pediatric Health, Children’s National Health System and Department of Pediatrics George Washington University School of Medicine and Health Sciences, Washington, D.C.

Linda Fu, Goldberg Center for Community Pediatric Health, Children’s National Health System and Department of Pediatrics George Washington University School of Medicine and Health Sciences, Washington, D.C..

Jichuan Wang, Center for Translational Science, Children’s National Health System and Department of Epidemiology and Biostatistics, George Washington University, Washington, D.C..

Rachel Y. Moon, Department of Pediatrics, University of Virginia, Charlottesville, VA

References

  1. Aston ML, 2002. Learning to be a normal mother: empowerment and pedagogy in postpartum classes. Public Health Nursing, 19(4), pp.284–293. [DOI] [PubMed] [Google Scholar]
  2. Brenner RA, Simons-Morton BG, Bhaskar B, Mehta N, Melnick VL, Revenis M, Berendes HW, Clemens JD 1998. Prevalence and predictors of the prone sleep position among inner-city infants. JAMA 280 (4), 341–346. [DOI] [PubMed] [Google Scholar]
  3. Brunson EK 2013. The impact of social networks on parents’ vaccination decisions. Pediatrics. 131 (5), e1397–404. [DOI] [PubMed] [Google Scholar]
  4. Centers for Disease Control and Prevention. 2019. Sudden unexpected infant death and sudden infant death syndrome. http://www.cdc.gov/SIDS/index.htm. Accessed July 22, 2019.
  5. Choi KH, Gregorich SE 2009. Social network influences on male and female condom use among women attending family planning clinics in the United States. Sex Transm Dis. 36 (12), 757–762. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Christakis NA, Fowler JH 2007. The spread of obesity in a large social network over 32 years. N Engl J Med. 357 (4), 370–379. [DOI] [PubMed] [Google Scholar]
  7. Coleman JS 1998. Social capital in the creation of human capital. American J Soc. 94, S95–S120. [Google Scholar]
  8. Colson ER, Joslin SC 2002. Changing nursery practice gets inner-city infants in the supine position for sleep. Arch. Pediatr. Adolesc. Med. 156 (7), 717–720. [DOI] [PubMed] [Google Scholar]
  9. Colson ER, Levenson S, Rybin D, Calianos C, Margolis A, Colton T, et al. 2006. Barriers to following the supine sleep recommendation among mothers at four centers for the Women, Infants, and Children Program. Pediatrics. 118 (2), e243–250. [DOI] [PubMed] [Google Scholar]
  10. Colson ER, McCabe LK, Fox K, Levenson S, Colton T, Lister G, et al. 2005. Barriers to following the back-to-sleep recommendations: Insights from focus groups with inner-city caregivers. Ambul Pediatr. 5 (6), 349–354. [DOI] [PubMed] [Google Scholar]
  11. Colson E R, Rybin D, Smith LA, Colton T, Lister G, Corwin MJ 2009. Trends and factors associated with infant sleeping position: the national infant sleep position study, 1993–2007. Arch Pediatr Adolesc Med. 163 (12), 1122–1128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Colson ER, Willinger M, Rybin D, Heeren TC, Smith LA, Lister G, et al. 2013. Trends and factors associated with infant bed sharing, 1993–2010: the National Infant Sleep Position study. JAMA Pediatr. 167 (11), 1032–1037. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cornwell B, Schumm LP, Laumann EO and Graber J, 2009. Social Networks in the NSHAP Study: rationale, measurement, and preliminary findings. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 64(suppl_1), pp.i47–i55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Crane Denise, and Ball Helen L.. 2016. “A Qualitative Study in Parental Perceptions and Understanding of SIDS-Reduction Guidance in a UK Bi-Cultural Urban Community.” BMC Pediatrics 16:23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Davey-Rothwell MA, Latkin CA 2007. Gender differences in social network influence among injection drug users: perceived norms and needle sharing. J Urban Health. 84 (5), 691–703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. De la Haye K, Robins G, Mohr P, Wilson C 2010. Obesity-related behaviors in adolescent friendship networks. Soc Networks. 32 (3), 161–167. [Google Scholar]
  17. Epstein J, Jolly C 2009. Credibility gap? Parents’ beliefs about reducing the risk of cot death. Community Pract. 82 (11), 21–24. [PubMed] [Google Scholar]
  18. Erck Lambert AB, Parks SE, Shapiro-Mendoza CK 2018. National and state trends in sudden unexpected infant death: 1990–2015. Pediatrics. 141 (3), e20173519. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hauck FR, Moore CM, Herman SM, Donovan M, Kalelkar M, Christoffel KK, et al. 2002. The contribution of prone sleeping position to the racial disparity in sudden infant death syndrome: the Chicago Infant Mortality Study. Pediatrics. 110 (4), 772–780. [DOI] [PubMed] [Google Scholar]
  20. Hwang SS, Rybin DV, Heeren TC, Colson ER, Corwin MJ 2016. Trust in sources of advice about infant care practices: the SAFE study. Maternal and Child Health Journal. 20 (9), 1956–1964. [DOI] [PubMed] [Google Scholar]
  21. Keats A, 2018. Women’s schooling, fertility, and child health outcomes: Evidence from Uganda’s free primary education program. Journal of Development Economics, 135, pp.142–159. [Google Scholar]
  22. Kellams A, Parker MG, Geller NL, Moon RY, Colson ER, Drake E, Corwin MJ, McClain M, Golden WC, Hauck FR 2017. TodaysBaby quality improvement: safe sleep teaching and role modeling in 8 US maternity units. Pediatrics. 140 (5), e20171816. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Krackhardt D, 1987. Cognitive social structures. Social networks, 9(2), pp.109–134. [Google Scholar]
  24. Kroll ME, Quigley MA, Kurinczuk JJ, Dattani N, Li Y, Hollowell J 2018. Ethnic variation in unexplained deaths in infancy, including sudden infant death syndrome (SIDS), England and Wales 2006–2012: national birth cohort study using routine data. J Epidemiol Community Health. 72 (10), 911–918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Laumann EO, Marsden PV and Prensky D, 1989. The boundary specification problem in network analysis. Research methods in social network analysis, 61, p.87. [Google Scholar]
  26. Macklin JR, Gittelman MA, Denny SA, Southworth H, Arnold MW 2019. The EASE project revisited: improving safe sleep practices in Ohio birthing and children’s hospitals. Clinical Pediatrics. 8 (9), 1000–1007. [DOI] [PubMed] [Google Scholar]
  27. Marsden PV, 1987. Core discussion networks of Americans. American Sociological Review, pp.122–131. [Google Scholar]
  28. Martens MP, Page JC, Mowry ES, Damann KM, Taylor KK Cimini MD 2006. Differences between actual and perceived student norms: An examination of alcohol use, drug use, and sexual behavior. Journal of American College Health. 54 (5), 295–300. [DOI] [PubMed] [Google Scholar]
  29. Mason B, Ahlers-Schmidt CR, Schunn C. 2013. Improving safe sleep environments for well newborns in the hospital setting. Clinical Pediatrics. 52 (10), 969–975. [DOI] [PubMed] [Google Scholar]
  30. Merton RK 1968. Social theory and social structure. New York: Free Press. [Google Scholar]
  31. Molina-García L, Hidalgo-Ruiz M, Cámara-Jurado AM, Fernández-Valero MJ, Delgado-Rodríguez M and Martínez-Galiano JM, 2019. Newborn Health Indicators Associated with Maternal Age during First Pregnancy. International journal of environmental research and public health, 16(18), p.3448. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Moon RY 2011. American Academy of Pediatrics, Task Force on Sudden Infant Death Syndrome. Technical report-SIDS and other sleep-related infant deaths: expansion of recommendations for a safe infant sleeping environment. Pediatrics. 129 (5), e1341–e1367. [DOI] [PubMed] [Google Scholar]
  33. Moon RY 2016. Task Force on Sudden Infant Death Syndrome, SIDS and other sleep-related infant deaths: Updated 2016 recommendations for a safe infant sleeping environment. Pediatrics. 138 (5), e20162938. [DOI] [PubMed] [Google Scholar]
  34. Moon RY, Calabrese T, Aird L. 2008. Reducing the risk of sudden infant death syndrome in child care and changing provider practices: lessons learned from a demonstration project. Pediatrics. 122 (4), 788–798. [DOI] [PubMed] [Google Scholar]
  35. Moon RY, Carlin RF, Cornwell B, Mathews A, Oden RP, Cheng YI, Fu LY, Wang J. 2019. Implications of mothers’ social networks for risky infant sleep practices. The Journal of Pediatrics. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Moon RY, Oden RP 2003. Back to sleep: can we influence child care providers? Pediatrics. 112 (4), 878–882. [DOI] [PubMed] [Google Scholar]
  37. Oden R, Joyner BL, Ajao TI, Moon R. 2010. Factors influencing African-American mothers’ decisions about sleep position: a qualitative study. J Natl Med Assoc. 102 (10), 870–880. [DOI] [PubMed] [Google Scholar]
  38. Ottolini MC, Davis BE, Patel K, Sachs HC, Gershon NB, Moon RY 1999. Prone infant sleeping despite the “Back to Sleep” campaign. Arch Pediatr Adolesc Med. 153, 512–517. [DOI] [PubMed] [Google Scholar]
  39. Parks SE, Erck Lambert AB, Shapiro-Mendoza CK 2017. Racial and ethnic trends in sudden unexpected infant deaths: United States, 1995–2013. Pediatrics. 139 (6), e20163844. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Price SK, Gardner P, Hillman L, Schenk K, Warren C. 2008. Changing hospital newborn nursery practice: results from a statewide “Back to Sleep” nurses training program. Maternal and Child Health Journal 12 (3), 363. [DOI] [PubMed] [Google Scholar]
  41. Polonec LD, Major AM, Atwood LE 2006. Evaluating the believability and effectiveness of the social norms message “most students drink 0 to 4 drinks when they party.” Health Commun. 20 (1), 23–34. [DOI] [PubMed] [Google Scholar]
  42. Rasinski KA, Kuby A, Bzdusek SA, Silvestri JM, Weese-Mayer DE 2003. Effect of a sudden infant death syndrome risk reduction education program on risk factor compliance and information sources in primarily black urban communities. Pediatrics. 111 (4), e347–354. [DOI] [PubMed] [Google Scholar]
  43. Reifman A, Watson WK, McCourt A. 2006. Social networks and college drinking: probing processes of social influence and selection. Pers Soc Psychol Bull. 32 (6), 820–832. [DOI] [PubMed] [Google Scholar]
  44. Schneider JA, Cornwell B, Ostrow D, Michaels S, Schumm P, Laumann EO, et al. 2013. Network mixing and network influences most linked to HIV infection and risk behavior in the HIV epidemic among black men who have sex with men. Am J Public Health. 103 (1), e28–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Schultz PW, Nolan JM, Cialdini RB, Goldstein NJ, Griskevicius V. 2007. The constructive, destructive, and reconstructive power of social norms. Psychol Sci. 18 (5), 429–434. [DOI] [PubMed] [Google Scholar]
  46. Shapiro-Mendoza CK, Colson ER, Willinger M, Rybin DV, Camperlengo L, Corwin MJ 2015. Trends in infant bedding use: national infant sleep position study, 1993–2010. Pediatrics. 135 (1), 10–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Shin E, Shim JM 2019. Listen to doctors, friends, or both? Embedded they produce thick knowledge and promote health. Journal of Health Communication. 24 (1), 9–20. [DOI] [PubMed] [Google Scholar]
  48. Swanson V, Power KG 2005. Initiation and continuation of breastfeeding: theory of planned behaviour. J Adv Nurs. 50 (3), 272–282. [DOI] [PubMed] [Google Scholar]
  49. Tjur T. 2009. Coefficients of determination in logistic regression models – A new proposal: The coefficient of discrimination. The American Statistician 63 (4), 366–372. [Google Scholar]
  50. US Census 2019. Retrieved from https://www.census.gov/quickfacts/DC, Accessed March 20, 2020
  51. Von Kohorn I, Corwin MJ, Rybin DV, Heeren TC, Lister G, & Colson ER (2010). Influence of prior advice and beliefs of mothers on infant sleep position. Archives of Pediatrics & Adolescent Medicine, 164(4), 363–369. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Willinger M, Ko CW, Hoffman HJ, Kessler RC, Corwin MJ 2003. Trends in infant bed sharing in the United States, 1993–2000: the National Infant Sleep Position study. Arch Pediatr Adolesc Med. 157 (1), 43–49. [DOI] [PubMed] [Google Scholar]
  53. Young C, Holsteen K. 2017. Model uncertainty and robustness: A computational framework for multimodel analysis. Sociological Methods & Research 46 (1), 3–4 [Google Scholar]

RESOURCES