OBJECTIVES:

To determine mediators of improvements in infant safe-sleep (SS) practices in a mobile health intervention.

METHODS:

In a cluster-randomized controlled trial, mothers received SS intervention or breastfeeding control videos for 60 days. Maternal responses about infant sleep position and location (outcomes) and mediators (attitudes, perceived social norms, and perceived control) from the theory of planned behavior were assessed. Intervention effects on mediators and association between mediators and outcomes were examined.

RESULTS:

Of 1600 recruited, 1263 mothers participated. Mothers receiving SS videos were more likely to have positive attitudes and norms for supine sleep (attitudes: adjusted odds ratio [aOR] = 2.35 [95% confidence interval (CI) 1.72 to 3.20]; norms: aOR = 1.75 [95% CI 1.27 to 2.42]) and recommended sleep location (attitudes: aOR = 1.91 [95% CI 1.54 to 2.36]; norms: aOR = 1.37 [95% CI 1.13 to 1.66]). Positive attitudes and norms toward supine sleep and room-sharing without bed-sharing were associated with higher odds of both practices (supine: aOR = 8.25 [95% CI 4.72 to 14.43] for positive attitudes and aOR = 6.67 [95% CI 4.25 to 10.46] for norms; room-sharing: aOR = 7.14 [95% CI 5.35 to 9.53] for positive attitudes and aOR = 4.44 [95% CI 3.03 to 6.51] for norms). Both positive attitudes and positive norms mediated the effect of the intervention.

CONCLUSIONS:

The intervention achieved success in improving adherence to SS recommendations by changing maternal attitudes and norms about supine sleeping and room-sharing without bed-sharing. Recognition that these attitudes and norms appear to be the main drivers of mothers’ choices regarding infant-sleep practices should inform health messaging strategies to promote SS.

What’s Known on This Subject:

A safe-sleep mobile health intervention, with videos sent via text or e-mail, resulted in higher rates of supine infant sleep positioning (adjusted percentages: 89.1% vs 80.2%) and room-sharing without bed-sharing (adjusted percentages: 82.8% vs 70.4%).

What This Study Adds:

The mobile health intervention achieved success in improving adherence to safe-sleep recommendations by changing maternal attitudes and norms about supine sleeping and room-sharing without bed-sharing. Discussion of parental concerns and consistent messaging may positively impact maternal attitudes and norms.

Despite the well-documented association of infant-sleep practices (including nonsupine positioning and bed-sharing) with sleep-related infant deaths (such as sudden infant death syndrome, accidental suffocation and strangulation in bed, and ill-defined deaths),1 US rates of adherence to recommendations for safe infant sleep are low,2,4 with rates of nonsupine sleep2 and bed-sharing3 having increased in recent years. Authors of previous studies have reported that the theory of planned behavior (TPB),5 which posits that beliefs are linked to behavior, can be used to identify factors associated with maternal decisions about infant sleep practices6 and vaccination.7 Specifically, maternal attitudes that a particular practice is good for the infant, perception of social norms around practices, and maternal perception of control over the decision are important in understanding maternal practices. However, it is unclear whether these factors are equally important.

We have shown that a safe-sleep (SS) mobile health (mHealth) intervention, with videos sent via text or e-mail, resulted in higher rates of supine infant sleep positioning (adjusted percentages: 89.1% vs 80.2%) and room-sharing without bed-sharing (adjusted percentages: 82.8% vs 70.4%),8 practices recommended by the American Academy of Pediatrics to reduce the risk of sleep-related infant deaths.9 This was a multifaceted intervention, based on the TPB, designed to impact maternal attitudes, perceived social norms, and maternal perceptions of control regarding SS behavior. Of the 2 components of the intervention (quality improvement [QI] and mHealth), the mHealth intervention was most effective in improving SS outcomes.9 Our aim for this article is to better understand why the mHealth intervention was effective, and which components of the intervention were responsible for the improved SS behavior. We hypothesized that maternal attitudes, perceived social norms, and maternal perception of control would be mediators of improved SS behaviors (ie, the mHealth intervention affected these mediators, which in turn affected SS behaviors). Additionally, no previous studies have quantitated the contribution of each of these mediators to changes in SS behavior. Therefore, we conducted a mediation analysis using Baron and Kenny’s10 steps to demonstrate that the mHealth intervention had an effect on the mediators; that the mediators, after controlling for the intervention, predicted the outcome of SS behaviors; and that the effect of the mHealth intervention was explained by controlling for the mediators.

The design and methods for the Social Media and Risk Reduction Training intervention have been described previously.8 Briefly, mothers of healthy term infants were recruited March 2015 to May 2016 from 16 US birth hospitals (see Acknowledgments) for a 4-armed cluster-randomized controlled trial of the impact of complementary health education interventions on SS (intervention) or breastfeeding (control) practices. The 16 hospitals were stratified into the 4 study arms such that they were matched on prestudy rates of SS practices, any breastfeeding, and exclusive breastfeeding. At the time of the study, 1 hospital had been designated Baby Friendly, and 3 were in the initial designation process. Nursing staff at each hospital were trained on SS or breastfeeding education in a QI intervention. Additionally, mothers received an mHealth intervention, branded as TodaysBaby,* in which they received videos via e-mail or text message for 60 days after hospital discharge. Hospitals were randomly assigned to 1 of 4 intervention combinations: breastfeeding QI and breastfeeding mHealth, SS QI and breastfeeding mHealth, breastfeeding QI and SS mHealth, or SS QI and SS mHealth; all mothers enrolled from a hospital received the same interventions. Mothers were excluded if they did not speak English, did not live in the United States, would not have custody of the infant, or could not receive daily e-mail or text messages or if the infant required hospitalization for >3 days, had contraindications to breast milk feeding or following SS guidelines, or was deceased. After providing written informed consent, mothers completed an initial demographics survey.

Institutional review board approval was obtained at Boston University, Yale University, University of Virginia, and all participating hospitals.

When the infant was ≥60 days old, mothers completed a survey about infant care practices, attitudes, subjective social norms, and perceived control with questions for which a Likert scale was used (1 [strongly disagree] to 7 [strongly agree]). Survey questions were the same as those used in previous studies.11,13 Development of the original survey questions included extensive pilot testing and assessment of test-retest reliability. For each set of questions, the average response was calculated; responses >4 were categorized as positive (responses in Supplemental Tables 5 and 6). Responses were analyzed for infant sleep position (supine position versus other) and infant sleep location (room-sharing without bed-sharing versus other) on the basis of maternal report of usual practice over the past 2 weeks.

Questions used to assess attitudes toward sleep position included the mothers’ ratings regarding if she believed that each infant sleep position (back, side, and stomach) made the infant healthy, safer, and more comfortable and kept the infant from choking. Questions on attitudes toward sleep location (bed-sharing and room-sharing without bed-sharing) were used to assess whether the location was pleasant for the infant and/or mother, safer for the infant, more comfortable for the infant and/or mother, and kept the infant from choking. Positive attitudes were defined as having positive attitudes toward the recommended behavior and not having positive attitudes toward other behaviors (eg, having positive attitudes toward both supine and side sleep would lead to a categorization of not having positive attitudes toward supine sleep only).

Social norms were assessed by asking if the people most important to the mother thought that the infant should sleep in each position or location. Positive social norms were defined as having positive norms toward the recommended behavior and not having positive norms toward other behaviors.

To assess perceived control for supine sleep, mothers were asked, by using a 7-point Likert scale, to rate the statement, “How I place my infant to sleep is mostly up to me”; responses >4 were categorized as positive. For sleep location, mothers were asked to rate the statements, “Choosing to sleep in the same bed with my infant is mostly up to me” and “Choosing to sleep in the same room with my infant (but not in the same bed) is mostly up to me”; subjects with both responses >4 were categorized as positive.

Statistical methods accounted for the intrahospital correlation structure of our cluster-randomized design. Cluster randomization does not guarantee balance on individual characteristics, and preliminary analyses were used to compare demographic characteristics of mothers assigned to the 2 interventions by using the Rao-Scott14 design-adjusted χ2 test. The statistical analysis accounted for the intrahospital correlation structure of our design. QI and mHealth interventions were represented through separate indicator variables in the analysis.

To examine factors from the TPB as mediators of the mHealth intervention effects on SS practices, the 4 steps of Baron and Kenny10 were followed. The association between the mHealth intervention and SS practices has been previously published (step 1).8 Effects of the mHealth intervention on potential mediators were examined through generalized estimating equation (GEE) multiple logistic regression and described through adjusted odds ratios (aORs) and 95% confidence intervals (CIs) (step 2). The association between the potential mediators and SS-practice outcomes were examined through GEE multiple logistic regression (step 3), and the mHealth intervention effect from this analysis, controlling for mediators, was compared with the mHealth effect from step 1 (step 4). The following individual-level characteristics were included as covariates: infant age and sex; maternal age, parity, race and/or ethnicity, educational level, and marital status; household income; and feeding methods (only breast milk, mostly breast milk or equally breast milk and formula, and mostly or only formula). Models were also controlled for QI assignment and baseline hospital rates of the SS outcome to control for baseline hospital differences and to parallel previously published analyses.

As a second approach to the mediation analysis, we performed causal mediation analyses on the basis of the counterfactual model (MPlus).15 This approach estimates the total effect of an intervention on a categorical outcome as the difference in the proportion with the outcome for those who received the intervention and were positive on the mediator versus those who received the control and were negative on the mediator. The total natural indirect effect through the mediator is the difference in the proportion with the outcome for those who received the intervention and were positive on the mediator versus those who received the intervention and were negative on the mediator. The pure natural direct effect is the difference in the proportion with the outcome for those who received the intervention and were negative on the mediator versus those who received the control and were negative on the mediator. Significance is determined through 95% CIs, with CIs not including 0.0 indicating significant effects. Comparing the total natural indirect effect and the total effect provides the percent of the intervention effect that is mediated. This approach to causal mediation was used to examine individual mediators separately, and separate analyses were conducted for each potential mediator, controlling for individual-level covariates.

Analyses were conducted by using SAS version 9.416 and MPlus version 7.3117 statistical software systems.

Of 1600 mothers who provided written consent, 1263 (78.9%) completed the ≥60-day survey (Table 1). Compared with nonrespondents, respondents were more likely to be ≥30 years old, not African American, and married and were more likely to have attended college (P < .0001 for all). Mothers in the intervention and control groups did not significantly differ on demographic characteristics and feeding practices (Table 1). Mothers had a mean age of 28.1 years (SD 5.8 years), with 51% being 20 to 29 years old and 42.3% being ≥30 years old. One-third (32.8%) were non-Hispanic white, 32.3% were Hispanic, and 27.2% were non-Hispanic African American. Ninety-three percent had a high school diploma, and 33.3% had a 4-year college degree; 50.7% were married. The mean infant age was 11.2 weeks (SD 4.4 weeks) at time of survey completion; 51.2% were girls. At the time of the survey, 36.2% of infants were only receiving breast milk, 49.7% were mostly or only receiving formula, and 23.4% were receiving mostly breast milk or equally breast milk and formula. Mothers who received the mHealth SS intervention were more likely to follow recommendations for sleep position (aOR = 1.99; 95% CI 1.43 to 2.79) and room-sharing without bed-sharing (aOR = 2.05; 95% CI 1.65 to 2.54).

TABLE 1

Demographic Characteristics of Respondents (N = 1263)

Controla (n = 608), n (%)mHealth SS (n = 655), n (%)
QI intervention   
 Yes 303 (49.8) 320 (48.9) 
 No 305 (50.2) 335 (51.1) 
Infant age at survey, wk   
 8–11 419 (68.9) 498 (76.0) 
 12–15 93 (15.3) 80 (12.2) 
 16–19 50 (8.2) 36 (5.5) 
 20+ 46 (7.6) 41 (6.3) 
Infant sex   
 Female 315 (51.8) 332 (50.7) 
 Male 293 (48.2) 323 (49.3) 
Parity   
 1 237 (39.0) 289 (44.1) 
 2 204 (33.5) 215 (32.8) 
 3+ 167 (27.5) 151 (23.1) 
Mother’s age, y   
 <20 39 (6.4) 46 (7.0) 
 20–29 321 (52.8) 323 (49.3) 
 ≥30 248 (40.8) 286 (43.7) 
Mother’s race and/or ethnicity   
 White 155 (25.5) 259 (39.5) 
 African American 202 (33.2) 142 (21.7) 
 Hispanic 199 (32.7) 209 (31.9) 
 Other 52 (8.6) 45 (6.9) 
Mother’s education   
 Less than HS 46 (7.6) 42 (6.4) 
 HS or GED 172 (28.4) 140 (21.4) 
 Some college 206 (34.1) 232 (35.5) 
 College or more 181 (29.9) 239 (36.6) 
Marital status   
 Married 290 (48.2) 350 (54.2) 
 Never married 290 (48.2) 262 (40.6) 
 Single, divorced, or widowed 22 (3.6) 34 (5.3) 
Household income, $   
 <20 000 96 (15.8) 85 (13.0) 
 20 000–49 999 121 (19.9) 118 (18.0) 
 ≥50 000 169 (27.8) 266 (40.6) 
 Unknown 222 (36.5) 186 (28.4) 
Feeding type at survey   
 Only breast milk 213 (35.0) 244 (37.2) 
 Mostly breast milk or equally breast milk and formula 163 (26.8) 133 (20.3) 
 Mostly or only formula 227 (37.4) 275 (42.0) 
 Unknown 5 (0.8) 3 (0.5) 
Controla (n = 608), n (%)mHealth SS (n = 655), n (%)
QI intervention   
 Yes 303 (49.8) 320 (48.9) 
 No 305 (50.2) 335 (51.1) 
Infant age at survey, wk   
 8–11 419 (68.9) 498 (76.0) 
 12–15 93 (15.3) 80 (12.2) 
 16–19 50 (8.2) 36 (5.5) 
 20+ 46 (7.6) 41 (6.3) 
Infant sex   
 Female 315 (51.8) 332 (50.7) 
 Male 293 (48.2) 323 (49.3) 
Parity   
 1 237 (39.0) 289 (44.1) 
 2 204 (33.5) 215 (32.8) 
 3+ 167 (27.5) 151 (23.1) 
Mother’s age, y   
 <20 39 (6.4) 46 (7.0) 
 20–29 321 (52.8) 323 (49.3) 
 ≥30 248 (40.8) 286 (43.7) 
Mother’s race and/or ethnicity   
 White 155 (25.5) 259 (39.5) 
 African American 202 (33.2) 142 (21.7) 
 Hispanic 199 (32.7) 209 (31.9) 
 Other 52 (8.6) 45 (6.9) 
Mother’s education   
 Less than HS 46 (7.6) 42 (6.4) 
 HS or GED 172 (28.4) 140 (21.4) 
 Some college 206 (34.1) 232 (35.5) 
 College or more 181 (29.9) 239 (36.6) 
Marital status   
 Married 290 (48.2) 350 (54.2) 
 Never married 290 (48.2) 262 (40.6) 
 Single, divorced, or widowed 22 (3.6) 34 (5.3) 
Household income, $   
 <20 000 96 (15.8) 85 (13.0) 
 20 000–49 999 121 (19.9) 118 (18.0) 
 ≥50 000 169 (27.8) 266 (40.6) 
 Unknown 222 (36.5) 186 (28.4) 
Feeding type at survey   
 Only breast milk 213 (35.0) 244 (37.2) 
 Mostly breast milk or equally breast milk and formula 163 (26.8) 133 (20.3) 
 Mostly or only formula 227 (37.4) 275 (42.0) 
 Unknown 5 (0.8) 3 (0.5) 

None of the differences between the control and mHealth SS interventions were significant at P < .05. GED, general equivalency diploma; HS, high school.

a

Received mHealth breastfeeding intervention.

Maternal attitudes toward health, comfort, safety, and choking risk were highly correlated, with all pairwise correlations between attitude items for the same sleep position or location >0.50 (Supplemental Tables 5 and 6).

Compared with mothers who received the mHealth control intervention, mothers who received the mHealth SS intervention were more likely to have positive attitudes and perceived positive social norms for both supine sleep and sleep location (Table 2). The mHealth SS intervention had a stronger impact on attitudes (aOR = 2.35; 95% CI 1.72 to 3.20) than norms (aOR = 1.75; 95% CI 1.27 to 2.42) for supine sleep; similarly for room-sharing without bed-sharing, the effect was stronger for attitudes (aOR = 1.91; 95% CI 1.54 to 2.36) than for norms (aOR = 1.37; 95% CI 1.13 to 1.66). Thus, regarding both attitudes and norms, the mHealth SS intervention had a greater effect for supine sleep than for room-sharing without bed-sharing. There was no significant effect of the mHealth SS intervention on maternal control for either sleep position or sleep location.

TABLE 2

The Effect of mHealth SS Intervention on Potential Mediating Factors From the TPB

Control (n = 608), %mHealth SS (n = 655), %aOR (95% CI)a
Potential mediators of the supine sleep effect    
 Positive attitudes toward supine sleep 51.6 75.0 2.35 (1.72 to 3.20) 
 Positive social norms for supine sleep 54.6 71.2 1.75 (1.27 to 2.42) 
 Maternal control over sleep position 83.7 85.8 1.15 (0.87 to 1.53) 
Potential mediators of the room-sharing without bed-sharing effect    
 Positive attitudes toward room-sharing without bed-sharing 50.4 67.5 1.91 (1.54 to 2.36) 
 Positive social norms for room-sharing without bed-sharing 48.8 53.5 1.37 (1.13 to 1.66) 
 Maternal control over sleep location 73.4 71.6 0.80 (0.60 to 1.05) 
Control (n = 608), %mHealth SS (n = 655), %aOR (95% CI)a
Potential mediators of the supine sleep effect    
 Positive attitudes toward supine sleep 51.6 75.0 2.35 (1.72 to 3.20) 
 Positive social norms for supine sleep 54.6 71.2 1.75 (1.27 to 2.42) 
 Maternal control over sleep position 83.7 85.8 1.15 (0.87 to 1.53) 
Potential mediators of the room-sharing without bed-sharing effect    
 Positive attitudes toward room-sharing without bed-sharing 50.4 67.5 1.91 (1.54 to 2.36) 
 Positive social norms for room-sharing without bed-sharing 48.8 53.5 1.37 (1.13 to 1.66) 
 Maternal control over sleep location 73.4 71.6 0.80 (0.60 to 1.05) 
a

Controlling for demographics, breastfeeding, and QI assignment through GEE logistic regression.

Analyses of the total effect of the intervention, controlling for demographic factors but not mediators, revealed that mothers receiving the mHealth SS intervention were more likely to place their infant supine (aOR = 1.99; 95% CI 1.43 to 2.79) and to room-share without bed-sharing (aOR = 2.05; 95% CI 1.65 to 2.54; Table 3).

TABLE 3

The Effect of mHealth SS Intervention and Potential Mediating Factors on Supine Sleep and Room-Sharing Without Bed-sharing

Model With mHealth SS Intervention OnlyModel With mHealth SS Intervention and Potential Mediators
aOR (95% CI)aaOR (95% CI)
Outcome: supine position   
 mHealth SS 1.99 (1.43 to 2.79) 1.37 (0.97 to 1.95) 
 Positive attitudes toward supine sleep — 8.25 (4.72 to 14.43) 
 Positive social norms for supine sleep — 6.67 (4.25 to 10.46) 
 Maternal control over sleep position — 1.12 (0.63 to 1.99) 
Outcome: Room-sharing without bed-sharing   
 mHealth SS 2.05 (1.65 to 2.54) 1.62 (1.26 to 2.09) 
 Positive attitudes toward room-sharing without bed-sharing — 7.14 (5.35 to 9.53) 
 Positive social norms for room-sharing without bed-sharing — 4.44 (3.03 to 6.51) 
 Maternal control over sleep location — 1.25 (0.83 to 1.90) 
Model With mHealth SS Intervention OnlyModel With mHealth SS Intervention and Potential Mediators
aOR (95% CI)aaOR (95% CI)
Outcome: supine position   
 mHealth SS 1.99 (1.43 to 2.79) 1.37 (0.97 to 1.95) 
 Positive attitudes toward supine sleep — 8.25 (4.72 to 14.43) 
 Positive social norms for supine sleep — 6.67 (4.25 to 10.46) 
 Maternal control over sleep position — 1.12 (0.63 to 1.99) 
Outcome: Room-sharing without bed-sharing   
 mHealth SS 2.05 (1.65 to 2.54) 1.62 (1.26 to 2.09) 
 Positive attitudes toward room-sharing without bed-sharing — 7.14 (5.35 to 9.53) 
 Positive social norms for room-sharing without bed-sharing — 4.44 (3.03 to 6.51) 
 Maternal control over sleep location — 1.25 (0.83 to 1.90) 

—, not applicable.

a

Controlling for demographics, breastfeeding, QI assignment, and hospital baseline rates of the outcome.

Controlling for intervention assignment, positive attitudes and positive social norms toward only supine sleep were associated with higher odds of mothers placing their infant supine (aOR = 8.25; 95% CI 4.72 to 14.43 for positive attitudes; aOR = 6.67; 95% CI 4.25 to 10.46 for positive social norms; Table 3). Maternal control over sleep position was not associated with placing the infant supine. Similarly, positive attitudes and positive social norms toward room-sharing without bed-sharing were associated with higher odds of this practice, whereas maternal control over sleep location was not.

The effect of the mHealth SS intervention on supine sleep position was attenuated after controlling for positive attitudes, positive social norms, and maternal control toward supine sleep only (aOR = 1.99; 95% CI 1.43 to 2.79 when not controlling for potential mediators; aOR = 1.37; 95% CI 0.97 to 1.95 after controlling for potential mediators; Table 3). Similarly, the effect of the mHealth SS intervention on room-sharing without bed-sharing was also attenuated after controlling for potential mediators.

These analyses suggest that both positive attitudes and positive perceived social norms toward SS behavior mediated the effect of the mHealth SS intervention. The mHealth intervention increased the percentage of mothers with positive attitudes and social norms, and the increased positive attitudes and social norms led to increased compliance with recommended SS behavior.

The causal mediation analysis also revealed that positive attitudes and social norms mediated the effect of the mHealth intervention on SS behavior (Table 4). For example, when treating positive attitudes toward supine sleep only as the mediator, the total mHealth intervention effect was to raise the percentage of mothers placing their infant supine by 16 percentage points. Thus, 6.25 women would need to receive the mHealth intervention to have 1 more mother place her infant supine. Ten percentage points of this effect (63%) could be explained by positive attitudes. Note that in causal mediation, the total intervention effect accounts for the mediator, and so the total mHealth intervention effect differs when examining different mediators. When treating social norms as the mediator, 27% of the total mHealth intervention effect could be explained by positive social norms.

TABLE 4

Causal Mediation Analysis of the mHealth SS Intervention

Total Effect of mHealth SSaTotal Natural Indirect Effect via MediatoraPure Natural Direct EffectaEffect Mediated, %
Potential mediators of the supine sleep effectb     
 Attitudes 0.16 (0.09 to 0.23) 0.10 (0.05 to 0.15) 0.06 (−0.01 to 0.13) 63 
 Social norms 0.14 (0.07 to 0.22) 0.04 (0.01 to 0.06) 0.10 (0.03 to 0.17) 27 
 Maternal control 0.16 (0.08 to 0.25) 0.00 (−0.01 to 0.01) 0.16 (0.08 to 0.25) 
Potential mediators of the room-sharing without bed-sharing effectb     
 Attitudes 0.11 (0.01 0.21) 0.03 (−0.01 to 0.07) 0.08 (0.01 to 0.14) 31 
 Social norms 0.10 (0.01 to 0.20) 0.01 (0.01 to 0.02) 0.09 (0.01 to 0.18) 10 
 Maternal control 0.11 (0.01 to 0.21) 0.00 (−0.01 to 0.01) 0.11 (0.01 to 0.21) 
Total Effect of mHealth SSaTotal Natural Indirect Effect via MediatoraPure Natural Direct EffectaEffect Mediated, %
Potential mediators of the supine sleep effectb     
 Attitudes 0.16 (0.09 to 0.23) 0.10 (0.05 to 0.15) 0.06 (−0.01 to 0.13) 63 
 Social norms 0.14 (0.07 to 0.22) 0.04 (0.01 to 0.06) 0.10 (0.03 to 0.17) 27 
 Maternal control 0.16 (0.08 to 0.25) 0.00 (−0.01 to 0.01) 0.16 (0.08 to 0.25) 
Potential mediators of the room-sharing without bed-sharing effectb     
 Attitudes 0.11 (0.01 0.21) 0.03 (−0.01 to 0.07) 0.08 (0.01 to 0.14) 31 
 Social norms 0.10 (0.01 to 0.20) 0.01 (0.01 to 0.02) 0.09 (0.01 to 0.18) 10 
 Maternal control 0.11 (0.01 to 0.21) 0.00 (−0.01 to 0.01) 0.11 (0.01 to 0.21) 
a

Effects are differences in proportions practicing supine sleep position with 95% CIs in parentheses. CIs that do not contain the null difference of 0.0 indicate significant effects.

b

Mediators were evaluated separately, and in analyses, we controlled for demographics, breastfeeding, QI intervention, and hospital baseline rates of SS behavior.

The Social Media and Risk Reduction Training mHealth intervention had a significant impact on maternal attitudes and perceived social norms regarding choice of infant sleep position and sleep location. Mothers receiving the intervention were more likely to have positive attitudes about supine positioning and room-sharing without bed-sharing and were more likely to state that the people most important to them encouraged these practices. These changes in attitudes and perceived social norms mediated the effect of the mHealth intervention. On the basis of the Baron and Kenny10 principles, these results suggest that the information provided by the mHealth intervention was successful in promoting SS behaviors by positively affecting both maternal attitudes and perceived social norms in favor of supine positioning and room-sharing without bed-sharing. Given that one is more likely to behave in a specific manner if one has positive attitudes about the behavior, if one perceives that the social norm is to behave in that way, and if one perceives that one has control over what the behavior is, the positive effects on maternal attitudes and perceived social norms are important.

Many of the mHealth videos were focused on changing persistent negative attitudes about recommended sleep position and location. For instance, there is a widespread perception that infants are more likely to choke or aspirate when they are supine, and this fear is 1 of the most commonly stated reasons for placing the infant nonsupine.18,22 Thus, in 1 of the early videos, it was specifically explained in concrete terms why choking and aspiration are not more likely while the infant is supine. Interventions in which the concern about choking and aspiration has been specifically addressed have resulted in decreased prone placement.23 Addressing this issue directly as part of the TodaysBaby campaign may have contributed to the impact the campaign had on attitudes about supine sleep. Health care provider discussion about parental concerns, such as choking and aspiration or increased infant arousability in the supine position,18,22 may also impact positively on parental attitudes.

Mothers were also more likely to use recommended infant-sleep practices if there was a positive social norm. Social norms are the practices perceived to be prevalent and acceptable. Receipt of consistent health messages increases the perception that recommended practices are the norm24; indeed, the likelihood of a mother using recommended infant-sleep practices is directly correlated with the number of people who have told her to do so.20 It is thus important that health messages from all health care providers be consistent.

Mothers in our study exhibited high levels of perceived control of infant-sleep practices. Among control mothers, 84% reported control over sleep position, and 73% reported control over sleep location; in our intervention, we did not specifically target perceived control. For both these reasons, it is not surprising that the mHealth intervention had little to no effect on mothers’ perceptions that they had control over how or where their infant slept. Other studies have revealed that infant comfort (ie, the infant sleeps longer or better) is an important reason for infant sleep position and location.18,22 Indeed, mothers have reported that the infant is the primary decision-maker regarding sleep position and location21; if the infant does not sleep well or cries while in a specific position or location, this can lead to a perception that the parent does not have control over these decisions. Future interventions to improve parental perceptions of control may be helpful in further improving infant-sleep practices.

We found no significant pathways from the QI intervention to supine sleep position or room-sharing without bed-sharing. The QI intervention had a synergistic effect with the mHealth intervention in improving supine sleep position rates,8 suggesting that teaching and role modeling by health care professionals in the birth hospital is important in imparting information and emphasizing the importance of SS recommendations. However, it did not result in any changes in maternal attitudes, perceived social norms, or perceived maternal control regarding infant-sleep practices. It is possible that the influence of nursing staff was not significant because the period of exposure was short compared with the more sustained exposure from the mHealth intervention.

Changes in societal attitudes and norms about specific health behaviors have resulted in documented behavior change. For instance, over the last 5 decades, attitudes and social norms regarding smoking have become more negative because of a combination of recognition of the dangers of smoking and secondhand smoke exposure; antismoking advertising; and increased cigarette taxes, laws and regulations regarding smoking in public spaces, with resultant declines in smoking rates.25,26 However, changing attitudes and norms can be challenging and can take decades and significant investment (eg, advertising and legislation) to accomplish. Use of mHealth should be considered as a potentially powerful tool to effectively change attitudes and perceived social norms regarding a variety of health issues.

Although our loss to follow-up was only 21%, nonrespondents were predominantly younger, African American, single, and less well educated, factors typically associated with higher rates of nonadherence to SS recommendations; we acknowledge that it might have been more difficult to impact attitudes and social norms in this group. However, although the baseline levels of SS adherence were lower in African Americans and subgroups of low socioeconomic status, the SS mHealth intervention improved these rates to levels comparable to those of other groups, (ie, effect sizes were stronger for supine position and room-sharing without bed-sharing in these subgroups). This suggests that we were able to impact attitudes and social norms in these subgroups as well. Nonetheless, it will be important to expand studies like this to populations with high rates of nonadherence to determine if these types of interventions can impact attitudes and social norms specifically in these groups. In addition, because enrollment was limited to English speakers, the results cannot be generalized to recent immigrant and other non–English-speaking populations. In addition, we acknowledge limitations inherent in self-reporting. It is possible that mothers who received the SS interventions were more reluctant to report practices that were inconsistent with the messages received in the interventions. However, our prestudy levels for each behavior are comparable to those of other studies.3,19,27 We also asked about infant-care practices other than sleep and feeding (eg, immunizations) in our survey so that outcome measures were less obvious. In addition, to minimize contamination of the various intervention groups, we used a cluster-randomization scheme. We also acknowledge limitations to the causal mediation analysis. The intervention may affect an intermediate confounder that is associated with both the mediator and outcome of interest. If this is the case, the estimated effect of our mediator could be biased, and we could be misattributing the causal effect. Similarly, there may be latent confounding in which there are multiple intermediate variables (mediators) that reflect a single latent mediator. In our case, attitudes and perceived social norms might both reflect a single latent construct, and we may be misattributing a causal pathway by treating these as separate mediators. Finally, because maternal attitudes toward health, comfort, safety, and choking risk were highly correlated, we could not differentiate whether 1 particular attitude was more important than another in leading to improved SS behavior.

The TodaysBaby mHealth intervention achieved success in improving adherence to the American Academy of Pediatrics SS recommendations by changing maternal attitudes and perceived social norms about supine sleeping and room-sharing without bed-sharing. Recognition that these attitudes and social norms may be the main drivers of mothers’ choices regarding infant-sleep practices should inform health messaging strategies, including the use of mHealth, to promote SS.

Dr Moon conceptualized and designed the project, participated in the analysis of the data, drafted the initial manuscript, and reviewed and revised the manuscript; Drs Corwin, Colson, Kellams, Drake, and Hauck and Ms Tanabe conceptualized and designed the project, assisted in the analysis of the data, and reviewed and revised the manuscript; Mr Kerr and Dr Heeren conducted the initial analyses and reviewed and revised the manuscript; Ms Geller helped conceptualize and design the project, participated in site recruitment and enrollment, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

This trial has been registered at www.clinicaltrials.gov (identifier NCT01713868).

FUNDING: Funded by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (1R01HD072815-01) and the CJ Foundation for sudden infant death syndrome. Neither funder had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the article; and decision to submit the article for publication. Funded by the National Institutes of Health (NIH).

*TodaysBaby materials are available on request.

We thank the staff at Slone Epidemiology Center, including Lisa Crowell, RN; Taylor Platt, BA; and Dena Margolis, BA, who assisted with study implementation and data collection. We also acknowledge the study staff at the 16 participating hospitals for their role in recruitment and data collection: Baystate Medical Center in Massachusetts, Bethesda Memorial Hospital in Florida, Brookdale Hospital and Medical Center in New York, Delaware County Memorial Hospital in Pennsylvania, Jersey Shore University Medical Center in New Jersey, Johns Hopkins Hospital and Medical Center in Maryland, Kaweah Delta Health Care District in California, Medical Center of Arlington, in Texas, Moreno Valley Community Hospital in California, Mount Carmel in Ohio, Riverside County Regional Medical Center in California, Riverside Regional Medical Center in Virginia, Rush-Copley Medical Center in Illinois, Saint Francis Hospital and Medical Center in Connecticut, Saint Joseph Hospital in California, and Texas Health Presbyterian Hospital in Plano, Texas. Drs Heeren and Corwin (Slone Epidemiology Center) had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Dr Heeren (Department of Biostatistics, School of Public Health, Boston University) and Mr Kerr (Slone Epidemiology Center, Boston University) conducted and are responsible for the data analysis.

aOR

adjusted odds ratio

CI

confidence interval

GEE

generalized estimating equation

mHealth

mobile health

QI

quality improvement

SS

safe sleep

TPB

theory of planned behavior

1
Moon
RY
;
Task Force on Sudden Infant Death Syndrome
.
SIDS and other sleep-related infant deaths: expansion of recommendations for a safe infant sleeping environment.
Pediatrics
.
2011
;
128
(
5
). Available at: www.pediatrics.org/cgi/content/full/128/5/e1341
[PubMed]
2
National Infant Sleep Position study. Available at: http://slone-web2.bu.edu/ChimeNisp/Main_Nisp.asp. Accessed March 12, 2018
3
Colson
ER
,
Willinger
M
,
Rybin
D
, et al
.
Trends and factors associated with infant bed sharing, 1993-2010: the National Infant Sleep Position Study.
JAMA Pediatr
.
2013
;
167
(
11
):
1032
1037
[PubMed]
4
Shapiro-Mendoza
CK
,
Colson
ER
,
Willinger
M
,
Rybin
DV
,
Camperlengo
L
,
Corwin
MJ
.
Trends in infant bedding use: National Infant Sleep Position study, 1993-2010.
Pediatrics
.
2015
;
135
(
1
):
10
17
[PubMed]
5
Ajzen
I
.
The theory of planned behavior.
Organ Behav Hum Decis Process
.
1991
;
50
(
2
):
179
211
6
Colson
ER
,
Geller
NL
,
Heeren
T
,
Corwin
MJ
.
Factors associated with choice of infant sleep position.
Pediatrics
.
2017
;
140
(
3
):
e20170596
[PubMed]
7
Fadel
CW
,
Colson
ER
,
Corwin
MJ
, et al;
Study of Attitudes and Factors Effecting Infant Care (SAFE) Study
.
Maternal attitudes and other factors associated with infant vaccination status in the United States, 2011-2014.
J Pediatr
.
2017
;
185
:
136
142.e1
[PubMed]
8
Moon
RY
,
Hauck
FR
,
Colson
ER
, et al
.
The effect of nursing quality improvement and mobile health interventions on infant sleep practices: a randomized clinical trial.
JAMA
.
2017
;
318
(
4
):
351
359
[PubMed]
9
Task Force on Sudden Infant Death Syndrome
.
SIDS and other sleep-related infant deaths: updated 2016 recommendations for a safe infant sleeping environment.
Pediatrics
.
2016
;
138
(
5
):
e20162938
[PubMed]
10
Baron
RM
,
Kenny
DA
.
The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations.
J Pers Soc Psychol
.
1986
;
51
(
6
):
1173
1182
[PubMed]
11
Hwang
SS
,
Rybin
DV
,
Heeren
TC
,
Colson
ER
,
Corwin
MJ
.
Trust in sources of advice about infant care practices: the SAFE study.
Matern Child Health J
.
2016
;
20
(
9
):
1956
1964
[PubMed]
12
Hwang
SS
,
Rybin
DV
,
Kerr
SM
,
Heeren
TC
,
Colson
ER
,
Corwin
MJ
.
Predictors of maternal trust in doctors about advice on infant care practices: the SAFE study.
Acad Pediatr
.
2017
;
17
(
7
):
762
769
13
Parker
MGK
,
Colson
ER
,
Provini
L
, et al
.
Variation in safe sleep and breastfeeding practices among non-hispanic black mothers in the United States according to birth country.
Acad Pediatr
.
2017
;
17
(
8
):
887
892
[PubMed]
14
Rao
JNK
,
Scott
AJ
.
On simple adjustments to chi-square tests with sample survey data.
Ann Stat
.
1987
;
15
(
1
):
385
397
15
Muthén
B
,
Asparouhov
T
.
Causal effects in mediation modeling: an introduction with applications to latent variables.
Struct Equ Modeling
.
2015
;
22
(
1
):
12
23
16
SAS Institute, Inc
.
SAS/STAT User’s Guide, Version 9.4
.
Cary, NC
:
SAS Institute, Inc
;
2013
17
Muthén
LK
,
Muthén
BO.
Mplus User’s Guide
. 8th ed.
Los Angeles, CA
:
Muthén & Muthén
;
1998–2017
18
Colson
ER
,
Levenson
S
,
Rybin
D
, et al
.
Barriers to following the supine sleep recommendation among mothers at four centers for the Women, Infants, and Children Program.
Pediatrics
.
2006
;
118
(
2
). Available at: www.pediatrics.org/cgi/content/full/118/2/e243
[PubMed]
19
Colson
ER
,
Rybin
D
,
Smith
LA
,
Colton
T
,
Lister
G
,
Corwin
MJ
.
Trends and factors associated with infant sleeping position: the national infant sleep position study, 1993-2007.
Arch Pediatr Adolesc Med
.
2009
;
163
(
12
):
1122
1128
[PubMed]
20
Von Kohorn
I
,
Corwin
MJ
,
Rybin
DV
,
Heeren
TC
,
Lister
G
,
Colson
ER
.
Influence of prior advice and beliefs of mothers on infant sleep position.
Arch Pediatr Adolesc Med
.
2010
;
164
(
4
):
363
369
[PubMed]
21
Oden
RP
,
Joyner
BL
,
Ajao
TI
,
Moon
RY
.
Factors influencing African American mothers’ decisions about sleep position: a qualitative study.
J Natl Med Assoc
.
2010
;
102
(
10
):
870
872, 875–880
[PubMed]
22
Robida
D
,
Moon
RY
.
Factors influencing infant sleep position: decisions do not differ by SES in African-American families.
Arch Dis Child
.
2012
;
97
(
10
):
900
905
[PubMed]
23
Moon
RY
,
Oden
RP
,
Grady
KC
.
Back to Sleep: an educational intervention with women, infants, and children program clients.
Pediatrics
.
2004
;
113
(
3, pt 1
):
542
547
[PubMed]
24
Rainey
DY
,
Lawless
MR
.
Infant positioning and SIDS. Acceptance of the nonprone position among clinic mothers.
Clin Pediatr (Phila)
.
1994
;
33
(
6
):
322
324
[PubMed]
25
Burns
D
.
How far we have come in the last 50 years in smoking attitudes and actions.
Ann Am Thorac Soc
.
2014
;
11
(
2
):
224
226
[PubMed]
26
Cummings
KM
,
Proctor
RN
.
The changing public image of smoking in the United States: 1964-2014.
Cancer Epidemiol Biomarkers Prev
.
2014
;
23
(
1
):
32
36
[PubMed]
27
Degan
VV
,
Puppin-Rontani
RM
.
Prevalence of pacifier-sucking habits and successful methods to eliminate them–a preliminary study.
J Dent Child (Chic)
.
2004
;
71
(
2
):
148
151
[PubMed]

Competing Interests

POTENTIAL CONFLICT OF INTEREST: In 2016, Dr Moon testified as a paid expert witness in a case of sleep-related infant death; the other authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data