Get Social Media and Risk-Reduction Training (GET SMART) to improve infant safe sleep practices through a mobile text message-delivered video safe sleep intervention for new parents: study protocol for a type 3 hybrid implementation-effectiveness cluster randomized trial in 20 US birth hospitals

Abstract

Background

Annually, more than 3500 US infants die suddenly and unexpectedly. Many sudden unexpected infant deaths (SUID) are preventable with greater adherence to safe infant sleep guidelines. There are racial and socioeconomic disparities, both in adherence to these guidelines and in SUID rates. In the original Social Media and Risk-Reduction Training for Infant Care Practices (SMART) study in which 1600 mothers were randomized to receive short educational videos, focused on safe sleep vs. control, those receiving the safe sleep intervention had ~ 10% point higher rates of following safe sleep practices, with elimination of racial and socioeconomic disparities in safe sleep practices. GET SMART (Get Social Media and Risk-Reduction Training), a type 3 hybrid implementation-effectiveness cluster randomized trial, aims to determine optimal strategies to implement this safe sleep intervention in real-world conditions.

Methods

We have recruited 19 US birth hospitals in counties with high SUID rates. After a control phase, hospitals will complete, in a randomized, crossover design, both the “high touch” (HT) implementation strategy used in SMART (hospital staff introduced the program, obtained informed consent, and watched 2 videos with them) and a “low touch” (LT) strategy that uses a direct-to-consumer approach in which mothers use QR codes to sign up for the safe sleep video intervention. Implementation strategies and outcomes are grounded in the Proctor Conceptual Model of Implementation Research. We will compare the impact of the hospital-based HT and LT strategies with regards to differences in penetration (proportion of eligible mothers who sign up), equity of penetration according to income, race and ethnicity, and program cost per mother sign up. Effectiveness outcomes are adherence to 4 safe sleep practices (supine sleep position, sleep location [roomsharing without bedsharing], nonuse of soft bedding, pacifier use) after the 2-month intervention.

Discussion

Findings from this implementation-effectiveness study will inform strategies to broadly scale an easily replicated mobile-delivered video safe sleep intervention among groups with historically high rates of SUID.

Trial registration

This trial is registered on ClinicalTrials.gov (NCT06618586), registration date 2024-08-27.

Background

Sudden Unexpected Infant Death (SUID), which includes Sudden Infant Death Syndrome (SIDS) and unintentional injury-related infant deaths, remains a leading cause of US postneonatal mortality, with more than 3500 deaths/year [1]. Many SUIDs are preventable with greater adherence to safe infant sleep guidelines, which include supine (back) sleep position, sleeping on a firm, flat surface next to but separate from the parent’s bed (roomsharing without bedsharing), elimination of soft bedding from the sleep environment, and pacifier use [2]. Despite this, the proportion of US families who report nonsupine positioning [3, 4] and bedsharing [4–6] has increased in the last decade, and soft bedding use remains > 50% in a national sample [7].

Racial, ethnic and economic disparities in SUID and infant safe sleeping practices persist. In 2023, the SUID rates for Black and American Indian/Alaska Native (AI/AN) infants were 2.05/1000 live births and 2.21/1000 live births, respectively and were more than twice the rate for white infants (0.73/1000 live births [LB]) [1]. The prevalence of nonsupine sleep placement [3, 4, 8–11], bedsharing [4, 6, 11–13] and use of soft bedding [7, 11] is higher for Black infants and infants born to low-income families, compared to white infants and those born in higher-income families. While there is a dearth of studies in the AI/AN population, a national survey found that rates of bedsharing and soft bedding use were also higher for AI/AN infants, compared to white infants [4]. These practices persist despite adequate knowledge; e.g., most parents know about supine recommendations but still place infants prone [14–23]. These racial and economic disparities in safe sleep practices may be due to specific beliefs and attitudes (e.g., infants are more likely to choke when supine [3]) that are more prevalent among Black and low-income families [8, 17, 24]. A lack of appropriate role modeling by healthcare providers [8] among low-income families may also contribute to the disparities. Indeed, in our nationally representative survey of > 3200 mothers, only 42% of low-income mothers reported receiving advice from nurses to place infants supine [8].

We conducted the successful [25] Social Media and Risk-Reduction Training for Infant Care Practices (SMART), a cluster randomized controlled trial in which we tested the effectiveness of a safe sleep education intervention delivered in the first 60 days after birth that had 2 novel components: (1) Safe sleep educational videos, branded as TodaysBaby™ delivered by text messages and (2) Interactive text message queries that were designed to both collect data on sleep practices in near real-time and provide reinforcement of safe sleep practices. Compared to attention-matched controls, mothers who received TodaysBaby™ videos had significantly improved adherence for all 4 study outcomes (93% [vs 80%] supine sleep, 86% [vs 70%] roomsharing without bedsharing, 82% [vs 68%] no soft bedding use, and 76% [vs 60%] pacifier use). Importantly, adherence rates for those receiving the TodaysBaby™ intervention were similarly high for all race/ethnicities and among low-income mothers, and the intervention eliminated racial/ethnic and socioeconomic disparities [25].

Although the SMART intervention was successful, the original implementation model was burdensome, making it difficult to implement widely. This burden fell on the hospital staff, as they had to take time from their busy workflow to approach mothers during the postpartum hospitalization, introduce the program, instruct them how to sign up, and then watch 2 videos with them. To optimize future scalability, as reliance on significant hospital resources (i.e., cost of staff time needed for training and mother recruitment) may be a major barrier to widespread implementation, it is important to assess how this “high touch” (HT) implementation strategy compares with a “low touch” (LT; i.e., less resource-intensive) strategy that leverages social marketing principles [26] to develop a direct-to-consumer approach that attracts mothers to sign themselves up for the TodaysBaby™ program.

GET SMART (Get Social Media and Risk-Reduction Training) is a hybrid type 3 implementation-effectiveness trial with crossover design at 20 hospitals to determine optimal strategies to implement SMART in real-world conditions, with the goals of maximizing penetration (maternal sign-up), while minimizing cost and maintaining intervention effectiveness. Our aims, which are grounded in the Proctor Implementation Science Framework (Proctor framework) [27], compare the impact of the hospital-based HT and LT implementation strategies through non-inferiority tests in the following areas: (1) primary implementation outcomes of penetration (proportion of eligible mothers who sign up for TodaysBaby™_); equity of penetration according to income (Medicaid insurance vs. not) and race/ethnicity (Black, AI/AN vs. white), and program cost per mother signed up; (2) secondary implementation outcomes of feasibility, acceptability, sustainability, and fidelity to the intervention; and (3) effectiveness of the intervention, defined as parent adherence to each of 4 safe sleep practices (supine sleep position, sleep location, nonuse of soft bedding, pacifier use) after the 2-month intervention.

In this manuscript, we describe the study protocol and detail how the Proctor framework guides the study methodology and outcomes.

Methods

Study design

In this hybrid type 3 implementation-effectiveness cluster randomized trial, we will test 2 strategies for recruitment of mothers during the birth hospitalization into the 2-month-long TodaysBaby™ mobile safe sleep education program: (1) The original TodaysBaby™ implementation strategy (HT), in which hospital staff recruit mothers and view the first 2 videos with them; and (2) A new low-touch (LT) hospital implementation strategy, in which mothers are attracted to TodaysBaby™ through direct-to-consumer marketing and sign themselves up using a QR code.

Potential advantages to the HT strategy include implementation in the postpartum hospital setting, where > 98% of US mothers give birth [28] and already receive infant care education, making this an ideal location to reach mothers broadly. Mothers may be more likely to sign up for a program introduced directly by a hospital staff member because of inherent trust by many mothers of medical providers in regards to infant care practices [29]. Mothers may also be more likely to engage in the entire intervention using this HT implementation strategy, as they have seen the first 2 videos and may thus be more likely to ultimately adhere to safe sleep practices. Potential disadvantages include the staff burden for training and mother recruitment. Hospital staff time is expensive and can be limited, especially during times of higher patient volume and acuity [30]. Thus, we will test an alternative LT (i.e. less resource-intensive) implementation strategy. In this strategy, hospital staff will be familiar with TodaysBaby™ as a free educational service and will encourage participation of mothers by simply pointing out a QR code easily viewed on fliers, crib cards, etc. Mothers will be encouraged to scan the QR code, which will bring them to a web-based sign-up portal where they view the same 2 videos as they would see with the HT implementation strategy. In this approach, hospital staff will not be directly involved in the sign up for the intervention and will not watch videos with the mothers. Instead of relying on nursing staff time, this strategy will leverage social marketing principles to develop a direct-to-consumer approach to attract mothers during the postpartum hospitalization to sign themselves up for the TodaysBaby™ program.

This is a crossover study design, in which half of the birth hospitals are randomized to HT/LT and half to LT/HT. We will also recruit a control group before beginning the HT and LT phases. These mothers will not have access to the TodaysBaby™ program and will only complete postnatal surveys that assess safe sleep practices; this group will be used as a control for our effectiveness outcomes using superiority testing. We have chosen to collect control data from our study hospitals rather than other hospitals not participating in the HT/LT strategy, because it is more feasible to involve fewer hospitals within the study scope and because we can avoid confounding from introducing variation in hospital-level factors (e.g., pre-existing safe sleep education, care processes, staffing levels and types) by collecting control data from the same hospitals. We have also chosen to enroll mothers in the control group prior to the implementation phases, rather than alongside the HT and LT implementation phases because it was not feasible to randomize individual mothers to the control vs. HT/LT strategy, as the implementation strategies are targeted at the entire postpartum hospital setting. We have chosen not to perform an attention-matched control because our team has already done this in SMART and shown that the TodaysBaby™ intervention is effective.

Conceptual framework and implementation strategies

The Proctor framework (Fig. 1) posits that improvements in outcomes are dependent not only on the evidence-based intervention (the “what”), but also on the implementation strategies (the “how”) used for the intervention. The Proctor framework distinguishes between the intervention (evidence-based practice) and different levels of implementation strategies (the approach for putting the intervention into place), incorporating three outcome domains: implementation, service, and client [27, 31]. Consistent with the Proctor framework and grounded in concrete predictions from the SMART study, our past research, the Expert Recommendations for Implementing Change (ERIC) [32], and qualitative interviews with stakeholders, we employ implementation strategies at 3 levels, all within the hospital:

Fig. 1.

Adapted proctor framework for implementation, with strategies for the 3 intervention levels and the outcomes in the areas of implementation, service, and client (implementation effectiveness)

1. The “Organizational” level is comprised of birth hospitals, where recruitment for TodaysBaby™ occurs. At this level, we will identify Site Champions. These are members of the staff who are very familiar to other local staff, knowledgeable of local work flow, and willing and enthusiastic to lead the GET SMART implementation at their sites. Site champions are the points of contact to central study staff.

2. The “Group/Learning” level is comprised of hospital staff who interact with the mothers. At this level, we will employ the following strategies:

   1. Educational Outreach and Dynamic Training, conducted by webinar, to explain the rationale and logistics of the cross-over design and the roles of staff members for each approach.
   2. Audit and Feedback, consisting of monthly webinars with site champions, during which we share data on rate of maternal sign up, answer questions, share tips, and trouble-shoot any issues. All webinars will be recorded for asynchronous viewing by those who cannot attend. To minimize cross-contamination, we will plan for hospitals in the HT and LT phases meet separately.
   3. TodaysBaby^TM "Toolkits," which we have developed in partnership with hospital staff and mothers. Toolkits contain materials that can be used by hospital staff and that give easy access to QR codes. Examples of toolkit materials include laminated cards to be placed on cribs, mobile workstations, and maternal discharge packets.
3. The “Hospital Systems/Environment” level constitutes the postpartum environment in which mothers are exposed to health information to care for their infant. At this level, we will use:

   1. QR codes to enroll participants. For the HT, a QR code is used following the 1:1 conversation with bedside staff for the logistics of signup. For the LT, the QR codes are available on toolkit materials.
   2. Direct-to-consumer social marketing principles to create engaging, attractive display materials to spark enough interest (e.g., creating sense of connectedness with mothers by providing photos and/or quotes from other mothers) so that mothers will scan the QR code.

For this study, Aims 1 and 2 focus on the implementation outcomes of penetration, cost, acceptability, feasibility, sustainability, fidelity to the intervention, and the service outcome of equity. Aim 3 focuses on the client (intervention effectiveness) outcome “symptomatology” (which is our effectiveness outcome: adherence to safe sleep practices).

Study setting

Because our goal is to reach geographically diverse populations at high risk for SUID, we have recruited 20 birth hospitals (with > 500 births/ year) in US counties with SUID rates that are higher than the national rate of 0.9/1000 live births. As there is significant geographic overlap with regards to socioeconomic status [33], race/ethnicity [34], and SUID [35] (Fig. 2), this approach is a practical and efficient way to systematically reach new patients in communities with a disproportionate burden of SUID and high proportion of families who are low-income, Black, and/or AI/AN.

Fig. 2.

Maps illustrate the geographic overlap of (A) poverty [33], (B) minority groups (orange: Black; light blue: AI/AN; dark green: Hispanic; light green: Black & Hispanic) [34], and (C) SUID [35]. Panel (D) illustrates the states where GET SMART hospitals are located

To identify hospitals located in counties of interest, we cross-referenced publicly available data from the Centers for Disease Control and Prevention (CDC), which provides SUID rate by county, and the American Hospital Association, which provides birth hospitals by county. This approach yielded a list of hospitals located in US counties with high SUID rates. We then recruited 20 hospitals. Hospital locations are illustrated in panel D of Fig. 2.

Evidence-based intervention: TodaysBaby™

The TodaysBaby™ intervention is comprised of: (1) text message-delivered, short (< 3 min long) educational videos in the first 2 months after birth; (2) text queries regarding safe sleep practices, which collect data on sleep practices in near real-time and provide reinforcement of adherence; and (3) 2 surveys, conducted upon enrollment and when the infant is 2 months old.

TodaysBaby™ videos will begin immediately after the mother signs up, with each video timed to match content addressing anticipated barriers and facilitators to adhering to guidelines (See Table 1 for list of video topics and timing). For example, videos address parental concerns that the infant will choke while supine and discuss alternatives to use of blankets for parents concerned that the baby will feel cold. TodaysBaby™ was created as a clearly identified “brand” so that subsequent videos will be anticipated, recognized and viewed. Videos will continue 2–3 times/week until the infant is 2 months of age.

Table 1.

Schedule of GET SMART Video Topics

Week 1	Why the TodaysBaby program? What will I learn from TodaysBaby? Why sleep position matters Will my baby choke on the stomach? Why my baby’s sleep space is important Should I chare a bed with my baby? How to handle outside advice What is the safest mattress?
Week 2	What about bedding and bumpers? Should I feed my baby in bed? Should I give my baby a pacifier? Why smoking around my baby is not recommended What makes a baby a good sleeper?
Weeks 3–5	Are some sleep spaces more dangerous than others? A reminder about pacifiers How can I make sure my baby is comfortable? More about pacifiers I shouldn’t worry about my baby choking
Weeks 6–8	Why sleep position matters Should I share a bed with my baby? What about bedding and bumpers? Thank you

Additionally, mothers will receive 1–2 text message questions weekly that assess sleep practices in near real-time (example, “In the last 2 weeks, how have you placed [baby’s name] for sleep [day or night? ]). Text messages will be personalized with each infant’s name to increase engagement. Each query will consist of a text message with the query and instructions to reply. To encourage and reinforce the desired behavior, mothers will receive prompt feedback (e.g., “Great job, sleeping on the back is the safest for your baby”). The text queries and responses for the 4 safe sleep outcomes are listed in Table 2. Text queries will also be used to remind mothers about upcoming surveys. All queries will be implemented with an intervention delivery platform permitting efficient scheduling of text queries, monitoring message openings, individualized responses based on query response, and blast messaging to subjects meeting specified criteria. This infrastructure can be readily scaled up for tens or hundreds of thousands of mothers with relatively little additional cost.

Table 2.

GET SMART Text message queries and responses

Topic	Text queries	Responses
Sleep position	In the last 2 weeks, how have you placed [baby’s name] for sleep (day or night? ) Reply: A: Side; B: Stomach; C: Back; D: More than one	If B: Great job! You are choosing the safest position for your baby! ϑ Other responses: Most parents choose the back position. It is the safest sleep position for your baby.
Sleep location (Bed sharing)	In the last 2 weeks, have you or others slept on a bed, armchair, couch/sofa or other surface with [baby’s name] (day or night)? Reply Y = Yes; N = No (If Msg1 = Y) Where? REPLY: A: Bed, B: Armchair; C: Sofa/couch: D: Other; E: More than one	If Msg1 = N: Good decision! Babies are safest when they sleep in their own sleep space. :) Other responses: The safest place for your baby to sleep is in your room, in his/her own crib or bassinet.
Sleep location (Roomsharing)	In the last 2 weeks, where has [baby’s name] slept? Reply: A=Same room as me; B=Room with someone else; C = Own room; D= More than one	If A: Great! The safest place for your baby to sleep is in your room, in his/her own crib or bassinet. Other responses: The safest place for your baby to sleep is in your room, in his/her own crib or bassinet.
Soft bedding	In the last 2 weeks, has [baby’s name] slept anywhere with bumpers, loose blankets, pillows, or stuffed toys? REPLY: Y = Yes; N = No (If Y) Which ones? Reply: 1 = Bumpers; 2 = loose Blanket; 3 = Pillow; 4 = Stuffed toy; 5 = More than one	If Msg1 = N: Great! This is the safest way for your baby to sleep. If Msg1 = Y and Msg2 = Y or N: The safest way for your baby to sleep is alone without any soft bedding.
Pacifier Use	In the last 7 days, has [baby’s name] used a pacifier when placed to sleep? Reply: Y = Yes or N = No	If Y or N: Pacifiers help to reduce the risk of SIDS. If your baby is breastfed, wait until they are 3–4 wks old before starting the pacifier.

Concomitant care

During the study period, maternal and infant birth hospitalization care will continue as usual. Other than the study interventions, the clinical workflow and care processes for mothers and infants during birth hospitalization will remain the same at each study site.

Crossover randomized trial design

Individual participants will be those who give birth at participating hospitals. Depending on the date of birth, they will be included in a pre-intervention control group or exposed to either implementation arm (Fig. 3). Mothers in the control group will receive “usual care,” with the hospital providing any safe sleep materials that were being used before the study. Mothers who are exposed to the HT and LT strategies who sign up for the TodaysBaby™ intervention will receive safe sleep videos and text queries for the first 60 days after birth. Mothers in the 3 groups will complete a follow-up survey at 60 + days after birth.

Fig. 3.

Individual participant timeline, demonstrating the activities for participants in the control, high touch, and low touch groups

Each hospital is randomized to start with one strategy for 6 months, followed by a 1-month washout phase (during which mothers will receive usual care), and the second strategy for 6 months. (Fig. 4) Randomization is done by the study statistical staff and is conducted at the hospital level, with hospitals stratified by the number of births annually (< 1000 births/year vs. 1000 + births/year). As hospitals and not individual participants are randomized, blinding procedures are not applicable.

Fig. 4.

Hospital timeline, demonstrating the control phase, followed by randomization (in a crossover design) to high touch and low touch phases, separated by a washout phase

We chose a crossover design because statistical analysis must consider variables and differences that can be specific to each hospital, and it is difficult to perfectly match hospitals to different study arms. A crossover design eliminates these hospital-specific variables/differences and provides a stronger design with increased statistical power.

GET SMART participants

An eligible mother must be English- or Spanish-speaking, live in the US, deliver a healthy infant at term gestation (≥ 37 weeks’ gestation) at a study hospital, plan to take the baby home with her after birth, and have a device with access to text messages. Excluded are mothers who are not expected to have custody of the infant, whose infant is deceased, or whose infant has a diagnosis for which safe sleep practices are contraindicated (e.g. spina bifida). For multiple births, we will ask about the first-born infant. Family members and friends can sign up to receive videos but only the mother will receive text queries and surveys.

A participant will be discontinued from the study by participant request. The participant can text “STOP” at any time. We will confirm with the participant that they do indeed want to stop participation before ceasing the intervention.

Participant recruitment procedures

Control phase

Hospital staff will identify and approach potentially eligible mothers and ask them if they would be willing to participate in a study in which they will respond to 2 surveys (one immediately and a follow-up survey at 60 + days postpartum). Interested mothers are given a flyer with a QR code for sign up and electronic informed consent.

HT arm

Hospital staff will identify and approach potentially eligible mothers and share the details of TodaysBaby™. If the mother is eligible and interested, the staff member will help the mother scan the QR code to sign up (including electronic informed consent) and view the first 2 videos with her.

LT arm

Hospital staff will select the toolkit materials that they believe will work best for encouraging mothers to scan the TodaysBaby™ QR code for sign up and electronic informed consent. These materials will be readily visible in the hospital environment. Hospital staff will be instructed to answer any questions brought up by mothers and point out the program in an abbreviated way, similar to any other educational programming offered.

Retention procedures

Overall, we believe that the bi-directional text queries will keep participants engaged in the TodaysBaby^™ intervention (HT and LT study arms). We also will provide incentives for mothers who respond to the text queries and surveys via use of a monthly lottery for a $100 store gift card. We have found that this incentivizes participation and is feasible within our grant budget (it is likely that we may have tens of thousands of mothers participating across the U.S., and we are unable to provide individual incentives).

Our intervention delivery platform will be used as a key method of monitoring subject viewing of videos, participation in text queries and survey completion. The platform will also be used to send weekly reminders to mothers regarding the timing of the 60 + day follow-up survey. Mothers who stated a preference to complete the survey online will receive a text with a link to the study internet portal and instructions, with multiple reminders if the surveys are not completed. Mothers who express a preference for telephone interview will be contacted by telephone. If a mother does not respond to the initial telephone interview attempt, a minimum of 10 call attempts will be made on different days of the week and at different times of day. Mothers who do not respond to text or call attempts will be sent an abbreviated survey by mail covering the core infant care practice questions, along with a cover letter and pre-addressed postage-paid envelope.

For mothers in the control group, since they do not participate in text queries and thus will not be eligible for the monthly lotteries, we will provide a $20 gift card for completion of the postnatal survey.

Data sources, outcome measures, and data analysis

Table 3 provides definitions and data sources. We will use both quantitative and qualitative methods. In our comparisons of HT and LT, we will use non-inferiority tests, while we will also use superiority tests in our effectiveness analyses (when comparing to the control arm) in Aim 3.

Table 3.

Definitions and data sources for outcome measures

Outcomes	Definitions	Data Sources
Main Implementation Outcomes
Penetration	Mothers who sign up for TodaysBaby™, divided by total eligible mothers [36]	Intervention delivery platform; monthly hospital statistics
Equity of penetration	Difference in rate of sign-up according to maternal race/ethnicity (Black and AI/AN vs. white) and insurance status (Medicaid vs. not)	Intervention delivery platform; monthly hospital statistics
Program cost	Estimated monthly cost per hospital and program cost per mother who signs up	Hospital data re: salaries, time spent; cost data from intervention delivery platform
Secondary Implementation Outcomes
Feasibility	Staff perception that a given implementation strategy was successfully used or carried out; factors explaining each strategy’s success or failure [31]	Qualitative interviews with staff upon study completion
Acceptability	Staff perception that the delivery and content of each given implementation strategy was satisfactory [31]	Qualitative interviews with staff upon study completion
Sustainability	Staff perception that a given implementation strategy could continue to be delivered after the end of the study [37]; monthly penetration rate	Qualitative interviews with staff upon study completion; intervention delivery platform
Fidelity (to intervention)	Number of videos viewed by mother, divided by total videos	Intervention delivery platform
Effectiveness Outcomes
Symptomatology: Adherence to 4 safe sleep practices in the past 2 weeks	Sleep position: % reporting supine position only Sleep location: % reporting roomsharing not bedsharing only Soft bedding: % reporting no soft bedding in sleep space Pacifier: % reporting usual pacifier use when infant sleeping	Follow-up surveys after TodaysBaby™ intervention

Given our two-level sampling design, with mothers clustered within hospitals, we will use general estimating equation (GEE) regression methods to examine differences in study outcomes across study groups. GEE methods account for the inter-hospital correlation resulting from our clustered sampling design, which will be modeled through an exchangeable correlation structure. Mothers in the control group are only included in analyses of effectiveness to ensure that each intervention is effective.

Aim 1

Outcome measures and data sources

Penetration and equity of penetration

To calculate penetration (the proportion of mothers who sign up among those eligible) and equity of penetration, we will collect “denominator” data from hospitals regarding their monthly number of mothers who gave birth to infants > 37 weeks (term births), were cared for in the postpartum unit and were discharged home with their infant. Participating hospitals also will provide the breakdown in these numbers by maternal race, ethnicity, language preference, and insurance status. “Numerator” data will be captured through enrollment in TodaysBaby™ (penetration) and self-reported by mothers in the baseline survey (equity).

Program cost

We will collect data to examine the perspective of direct program costs, rather than societal costs, because our primary goal is to provide relevant data to hospital administrators and public health agencies who may consider adoption of this program. We will ask site champions to provide estimates of the cost per hour of staff time and types of staff involved, and solicit time spent for initial training and HT sign up of mothers from staff. The study team will estimate the costs of the marketing materials developed for the toolkit (HT and LT) and operational costs of the intervention delivery platform.

Data analysis

Penetration

GEE logistic regression will be used to compare the odds of signing up for TodaysBaby™ for mothers from the LT vs. HT strategies. Independent variables in the GEE logistic regression will be an indicator for strategy (LT vs. HT) and an indicator for randomization order to control for possible order effects. We will compare the penetration of the LT vs. HT strategies through a non-inferiority test based on the OR for LT vs. HT, with a non-inferiority margin OR = 0.81, which corresponds to differences in the probability of signing up of 0.55 vs. 0.60. These analyses will be based on the number of mothers giving birth at a hospital (monthly hospital statistics) and the number of mothers signing up for TodaysBaby™ (tracked through the intervention delivery platform), and so will not involve individual-level data or missing data. The cross-over design ensures that mothers from each study hospital are included under both implementation strategies, and so we do not anticipate confounding from individual-level characteristics.

Equity of penetration

(Insurance status and race/ethnicity). We will examine differences in penetration according to maternal insurance status and by race/ethnicity. GEE logistic regression will be conducted, with interaction terms between implementation strategy and insurance status and between implementation strategy and race/ethnicity.

Total program cost

Is estimated (a) per hospital-month by treatment arm and (b) cost per enrolled mother per month by treatment arm. For each hospital, we will use the fixed costs of intervention delivery, hospital-specific time accounting, and pay rate of involved hospital staff types to develop a point estimate with lower and upper bounds, based on one-way sensitivity analyses using observed (time accounting) and prescribed (e.g., pay rate +/- 10%) variability, for the average monthly cost of the LT and HT strategies. We will measure the correlation between hospital-level program costs and monthly delivery volume to assess whether program costs scale appreciably. Before doing so, we will adjust the cost data for pay differences across states and metropolitan areas using data from the Bureau of Labor Statistics Occupational Employment and Wage Statistics for nurses and other healthcare workers. We will use the hospital-month estimated program costs during the appropriate months for the LT and HT strategies along with the number of enrolled mothers, in total and among high-risk mothers, to calculate hospital-month estimated cost per enrolled mother and enrolled high-risk mother. We will test for mean differences in hospital monthly program costs, cost per enrolled mother, and cost per enrolled high-risk mother between the LT and HT strategies using a two-sided paired t-test across the hospitals with p < 0.05 as the threshold for statistical significance.

Sample size calculation for aim 1

Sample size and statistical power considerations were evaluated through simulation and focus on our primary implementation outcomes, penetration and equity of penetration. The proposed sample size gives excellent power for examining overall differences between strategies and allows for meaningful examination of effect modification. We assumed that hospitals will participate, with n = 200 eligible mothers from each hospital under each of the LT and HT strategies (n = 8000 mothers total) and an additional n = 20 per hospital for the control group (total n = 400). We assumed that, on average, 60% of mothers would sign up under the HT strategy, and that this rate will vary across hospitals according to a normal distribution with standard deviation of 5% (so that ~ 90% of hospitals have a HT sign up percentage between 50% and 70%). We evaluated power for a non-inferiority test, to show that the odds of signing up were not meaningfully lower for the LT vs. HT strategy with a non-inferiority margin odds ratio of 0.81, corresponding to a difference in the percent signing up of 55% vs. 60%. We also assumed no order effects and evaluated empirical power using 1000 simulated samples. For the non-inferiority test of the LT vs. HT strategies regarding whether mothers sign up, there is 0.99 power of showing non-inferiority if there is truly no difference between the LT vs. HT strategies (odds ratio [OR] LT vs. HT = 1.00).

For power of analyses of equity of penetration, we focus on effect modification first by insurance status, and then by race/ethnicity. Based on data in Fig. 4, we assumed that the percent of mothers having Medicaid insurance at a hospital would be 60% on average and follow a uniform distribution across hospitals from 40% to 80%. Interaction models will be used to estimate separate odds ratios for the LT vs. HT implementation strategies by insurance status, and to test for a significant interaction (i.e., effect modification). If there is truly no difference between the LT vs. HT strategies in either the Medicaid insurance or non-Medicaid insurance subpopulations, there is 0.97 and 0.88 power, respectively, of showing non-inferiority in each of these subpopulations. The study has 80% power to show a significant interaction, if the LT and HT strategies were equally effective in the non-Medicaid insurance population but the LT strategy was less effective in the Medicaid insurance population (OR = 0.76, corresponding to a difference in the probability of signing up of 60% vs. 53%).

For equity of penetration by race/ethnicity, we assumed that the percent of mothers who are Black is uniformly distributed across hospitals from 10% to 60%, with an average of 35%. Interaction models will be used to estimate separate odds ratios for the LT vs. HT implementation strategies by race/ethnicity (Black vs. not Black) and to test for a significant interaction. If there is truly no difference between the LT vs. HT strategies in either the Black or the non-Black population, there is 0.84 and 0.96 power, respectively, of showing non-inferiority in each of these subpopulations. The study has 80% power to show a significant interaction if the LT and HT strategies are equally effective in the non-Black population, but the LT strategy was less effective in the Black population (OR = 0.76, corresponding to a difference in the probability of signing up of 60% vs. 53%). AI/AN infants are also at high risk for SUID, and we will also examine penetration in this population. However, we expect a small sample of AI/AN mothers in the sample, and so these analyses must be considered exploratory.

Although not one of our primary implementation outcomes, we also considered power analyses of costs. We will compare point estimates of hospital monthly costs under the LT and HT implementation strategies. Given our cross-over design, costs are paired by hospital and will be compared through a two-sided paired t-test across the 20 hospitals. Using Cohen’s d as a standardized effect measure (calculated as the difference in mean cost divided by the common standard deviation of costs under a strategy), and assuming a moderate correlation of r = 0.50 between hospital costs under the two implementation strategies, we have 0.80 power of detecting moderate-to-large differences in monthly cost corresponding to d = 0.66.

Aim 2

Outcome measures and data sources

Feasibility

Will be defined by hospital staff perception that a given implementation strategy was successfully used or carried out. Data will be collected qualitatively through staff interviews after study completion.

Acceptability

Will be defined by staff perception that the delivery and content of each given implementation strategy was satisfactory. Data will be collected qualitatively through staff interviews after study completion.

Fidelity

Will be defined by the proportion of total videos viewed by the mother. This information will be available from the intervention delivery platform.

Sustainability

Will be defined by staff perception that a given implementation strategy could continue to be delivered after the end of the study; these data will be collected qualitatively through staff interviews after study completion. Additionally, we will look at the trend with regards to penetration rates over the course of the study.

Data analysis

Qualitative analyses for feasibility, acceptability, and sustainability

We will conduct qualitative interviews with hospital staff within 3 months of completion of study participation to gain perspectives of feasibility, acceptability and future sustainability and the extent that contextual factors contributed to the success of each implementation strategy. We will seek perspectives of nursing and physician leaders, bedside nurses, and any other staff that may have been directly involved in the conduct of the study (some of whom may have participated in toolkit development interviews). To gain a breadth of perspectives, we will recruit hospital staff from geographically diverse hospitals. We will interview approximately equal numbers of staff type. We therefore anticipate 20–25 interviews; however, we will continue interviewing until thematic sufficiency [38] is reached. To aid in recruitment, $40 store gift cards will be provided to each participant. Interviews will last 30–60 min, will be conducted virtually, and be transcribed. We will use a semi-structured approach to questions, and our interview guide will have open-ended questions to elicit responses. A trained facilitator will lead each interview based on our interview guide; however, the flow of the conversation will be largely framed and structured by the respondents. The qualitative analysis will use an iterative approach [39]. Investigators will read transcripts, decide on a uniform set of codes, and iteratively develop themes until thematic sufficiency [38] is reached. Dedoose [40] software will be used. We will similarly assess validity of our data through investigator triangulation and member checking.

Analyses for fidelity

Will use GEE Poisson regression to compare the number of videos watched (outcome measure) for mothers from the LT vs. HT strategies, controlling for order effects. These analyses will control for individual-level variables reported through the intervention delivery platform. Poisson regression describes effects through rate ratios for the probability of watching a video, and the variability assumption of Poisson regression will be checked, and negative-binomial regression will be considered as an alternative approach if appropriate. We will compare fidelity for the LT vs. HT strategies through a non-inferiority test based on the risk ratio for implementation strategy.

Analyses for sustainability

To explore the short-term sustainability of these implementation strategies over the 6-month implementation periods for the HT and LT strategies, GEE logistic regression models will be fit on monthly penetration rates, with a linear time variable, and interaction between implementation strategy and time, to examine decreasing or increasing trends in penetration.

Aim 3

Outcome measures and data sources

“Symptomatology” or adherence to safe sleep practices

Mother-completed surveys will provide data regarding adherence to 4 safe sleep practices.

Data collection procedures

Surveys

Two surveys will be collected, one upon enrollment and one after the infant is 2 months old from all participants (control, HT, and LT groups).

Once consent is obtained (and for HT and LT groups, before the first video is shown), mothers will complete the baseline survey online, which asks validated questions [25, 41] about demographics, delivery history, home environment, follow-up tracking information, and current and anticipated future infant care practices.

The 15-minute follow-up survey will be launched > 60 days after birth and asks about past, current, and anticipated future behavior regarding infant care practices.

The timing of the 2 surveys allows us to ascertain: (1) timing and characteristics of birth, and infant care practices in the immediate postnatal days, and (2) infant care practices during the 1–2 month infant age window, when the risk of SUID is the highest [42]. Mothers will be able to complete the survey online or by phone, depending on her preference.

Data analyses

Intervention effectiveness

Will be assessed for mothers from all study groups who complete the follow-up survey. Separate analyses, using GEE logistic regression, will be performed for each of the 4 safe sleep practices. Control, LT, and HT phase will be represented through indicator variables in these models, and these analyses will control for individual-level variables reported in the follow-up survey. While our primary interest is the comparison of behaviors for mothers under the LT vs. HT strategies by non-inferiority testing, in our secondary analyses, we will also compare mothers under each of these two strategies to mothers from the control phase using superiority tests, and we will use the Hochberg procedure to account for multiple comparisons across the three study phases.

Sample size calculations for aim 3

For power analysis of effectiveness outcomes, we focus on sleep position. These analyses will be based on survey data from mothers from all study groups who complete the follow-up survey. We expect approximately 60% of mothers under the LT and HT strategies to sign-up for TodaysBaby™ and for 80% of those who sign-up to complete the follow-up survey, for approximately n = 96 mothers per hospital per implementation strategy (200 × 0.6 × 0.8). We expect 80% of mothers enrolled under the Control period to complete the follow-up survey, for approximately n = 16 mothers per hospital (20 × 0.8). Based on our experience with SMART, we expect roughly 70% of mothers from the Control period and 80% of mothers under the HT strategy to follow supine sleep guidelines.

For the non-inferiority comparison of mothers under the LT vs. HT implementation strategies, with a non-inferiority margin odds ratio of 0.75 (corresponding to a difference of 75% vs. 80%), there is 96% power of showing non-inferiority if the probability of following supine sleep guidelines is the same for the LT and HT strategies.

For the superiority testing comparing mothers recruited under the HT strategy to mothers under the Control period, to achieve 80% power for detecting the assumed difference of 80% vs. 70%, assuming that 80% of mothers in each group complete follow-up, we need to enroll a total of 375 mothers during the Control period and 375 mothers during the HT strategy period (average of about 19 mothers per site in each period). We expect our actual enrollment to meet or exceed the minimum number required to achieve 80% power.

Interim analyses

The only interim analyses that we anticipate are reviews of enrollment data (either in aggregate or stratified by hospital) to provide feedback to hospitals about the number of participants who have signed up for TodaysBaby^™.

Methods in analysis to handle protocol non-adherence and any statistical methods to handle missing data

ll survey data will be analyzed on an intent-to-treat basis. Missing survey data will be handled by multiple imputation. Hospital data is expected to be complete; however, missing salary data may be supplemented with publicly available state-level aggregates and missing staffing mix data or time spent on HT and/or LT will be handled by making assumptions using data from the remaining participating hospitals.

Data management

Data from mothers

Data from bidirectional text messages and surveys will be collected from human subjects via secure websites and will be stored in secured servers. Information connecting surveys with unique participant identifiers will also be stored in a secure location, and only study investigators will have access to that information. Identifying information will be separate from data. Data will be analyzed and presented in the aggregate. No biologic materials will be taken.

Data from hospitals

Aggregate data on the number of English- and Spanish-speaking mothers discharged home from the postpartum unit with their infants will be obtained from participating hospitals monthly. We will also obtain hospital data about the breakdown of race/ethnicity and insurance status (Medicaid vs. not) for patients. These data will be sent via secure email to the research team monthly. Medical records about individual participants will not be accessed. For cost analyses, we will ask site champions to provide estimates of the cost per hour of staff time, types of staff involved, and solicit time spent for initial training and sign-up of mothers from their staff (HT only). We will also request pay rates (e.g., minimum/junior and maximum/senior hourly rate or annual salary) for each participating hospital and staff type, with state-level aggregate data used as a backup if hospitals are unwilling or unable to share salary data.

Confidentiality

All efforts will be made to protect the anonymity of participants during all queries and surveys. Using standard approaches, mothers will be reminded at the beginning of each interview that they have the right to refuse to answer any questions or to terminate the interview and withdraw from the study at any time. No identifying information other than a unique study identifier will be used in the surveys. Information used to contact participants for follow-up interviews will be recorded and stored separately from the surveys. Furthermore, to minimize breach of confidentiality, study records will be stored in secured electronic files by the study team. To further ensure data security, all data entry, coding and analyses will be conducted centrally at Slone Epidemiology Center. When analyses are actively being conducted, a copy of the data set will be kept on a password-protected machine behind the institution’s firewall; this machine will be kept in a locked office to physically limit access. Once analyses have been completed, the data will be removed from the hard drive(s) of any computer(s) used for analysis, and will be stored solely on encrypted electronic media storage devices in a secure location. All members of the study team will maintain certification with their IRB of record of their qualification to be involved in human subjects research.

Plans to give access to the full protocol, participant level-data and statistical code

Intellectual property and data generated under this project will be administered in accordance with both participating institutional and NIH policies, including the NIH Data Sharing Policy and Implementation Guidance of March 5, 2003, which was the policy guidance in effect when this project was funded. Materials generated under the project will be disseminated in accordance with participating institutional and NIH policies. Depending on such policies, materials may be transferred to others under the terms of a material transfer agreement.

We have provided the scientific community with access to data collected during our prior studies, and this will continue with GET SMART. This includes survey instruments and datasets. Researchers can contact our research staff for a request form that can be used to request access to the datasets and support for their use. This support will include statistical guidelines and programming code for SUDAAN and for STATA. Additionally, we have provided access to the TodaysBaby^™ videos to organizations who wish to use them for educational purposes under the terms of a material transfer agreement, and this will continue.

Data will typically be made available to the public approximately one year after surveys are completed.

Access to databases generated under the project will be available for educational, research and non-profit purposes. Such access will be provided using web-based applications, as appropriate. Publication of data shall occur during the project, if appropriate, or at the end of the project, consistent with normal scientific practices. The final locked dataset will be deidentified before being made available to the nonprofit community for noncommercial uses provided that a data-sharing agreement is reached and that the original creators of the dataset are able to collaborate in all secondary data analysis.

Oversight and monitoring

Composition of the coordinating center and trial steering committee

Slone Epidemiology Center at Boston University will serve as our data coordinating center. Slone houses and/or provides access to all the facilities, resources, and services necessary for the successful execution and completion of the proposed research project.

The principal investigators and project manager serve as the trial steering committee that meets biweekly and communicates as needed between meetings. The entire team also meets biweekly.

Composition of the data monitoring committee, its role and reporting structure

A Data and Safety Monitoring Committee (DSMC), comprised of 3 individuals unaffiliated with this grant, has been established to act in an advisory capacity to monitor participant safety, data quality, and study progress. The DSMC includes experts in or representatives of the fields of pediatrics, infant mortality, clinical trial methodology, and health services research. The DSMC’s objective and external role is to review recruitment, protocol deviations, protocol amendments, and all adverse events or complications related to the study; and to make recommendations for protocol changes if indicated. The DSMC meets virtually at least twice annually and as needed should participant safety questions or other unanticipated problems arise. We will compile an aggregate adverse event report from both the intervention and control groups annually; these data will be blinded in the summary reports that will be provided to the DSMC for review.

Adverse event reporting and harms

In this study, mothers in the intervention groups (HT and LT) will receive health information about infant care beginning at the time of birth and extending until the infant is 2 months of age. The risks in this study are minimal, and based on our extensive experience, it is highly unlikely that we will encounter any adverse events (AE) or serious adverse events (SAE).

An adverse event will be defined as an undesirable sign, symptom, or medical illness while participating in the study, even if not considered related to the study. Such events will be noted on an adverse event form for review by the DSMC. An adverse event may also be the unintended effect resulting from collection of private, identifiable information. Every effort will be made to keep identifiable information secure. Should there be an adverse event, this will be documented for review and reported to the IRB.

The research study team is responsible for monitoring individual events and determining whether the study protocol needs modification to minimize risk. The principal investigators are pediatricians and are well acquainted with risks to the infants of the participants. One of our co-investigators (FRH) is a family physician who cares for adults and will monitor risks to the participants themselves. The principal investigators will review, collate and evaluate AEs, both serious and nonserious, within 24 h of notification by research staff. Serious/life-threatening events will be reported to the IRB and NIH immediately; all others will be included in the annual report to the IRB and NIH.

Should any study participant experience postnatal loss, the study research team will immediately express condolences and then ensure that all study procedures and interventions are discontinued as soon as we are notified. (If women who have experienced loss wish to continue to receive videos, they will be able to do so but will be excluded from the study.) The study research team will provide information about grief and bereavement resources, if desired by the participant.

Withdrawal of enrolled participants will be assessed on a case-by-case basis, depending on the situation and whether there are any potential risks to following the study protocol. The decision to withdraw a participant from the study and/or to halt data collection for medical risk reasons will ultimately be made by the principal investigators. Should a participant develop a medical exclusionary condition, she will be given the opportunity to continue receiving the intervention messages and videos if she feels that they are useful, but further data will no longer be collected from her.

Plans for communicating important protocol amendments to relevant parties

The principal investigators are responsible for monitoring individual events and determining whether the study protocol needs modification to minimize risk. The principal investigators are well acquainted with risks to the infants of the participants. FRH is a family physician who cares for adults and will monitor risks to the participants themselves. A protocol violation will be defined as a deviation from study design procedures, local regulations, or that which was approved by the IRB. Minor protocol deviations will be documented on forms for review by the study research team and PIs. Major protocol deviations that involve departures from IRB-approved or local regulations about research will be documented and submitted to the DSMC and the IRB for review.

Dissemination plans

The project results are national in scope. We will widely disseminate the results from this project through national and international conferences and presentations. Additionally, we anticipate publishing peer-reviewed manuscripts on both the quantitative and qualitative portions of this project. We will share results with participating hospitals as well.

RYM, FRH, SH, and RFC sit on the AAP Subcommittee on SUID and therefore have multiple contacts nationally with policymakers at the NIH, AAP, and other national stakeholders in safe infant sleep efforts. MP is a member of the AAP Committee on Fetus and Newborn, a group whose explicit goal is to write and disseminate clinical and policy reports to guide clinical newborn case, based on the latest evidence. Having project investigators on these national committees will facilitate rapid policy change if warranted by the results of this project.

Discussion

Guided by the Proctor framework, GET SMART’s hybrid type 3 implementation -effectiveness trial design builds upon the known effectiveness of the SMART intervention by assessing its implementation and effectiveness when delivered through high-touch and low-touch strategies. At local, state, and national levels, tremendous resources are invested in trying to reduce rates of infant mortality, especially SUID. Despite national programs and policies to raise awareness among perinatal populations and healthcare providers regarding adherence to SUID-risk reducing practices, SUID rates have not improved in over 20 years, and racial/ethnic and socioeconomic disparities persist. More broadly, the birth hospitalization is a time during which hospital staff are often overwhelmed by the number of procedures and the amount of parent education that need to be accomplished in a short period of time, and although interventions such as SMART demonstrate the capacity to successfully improve infant safe sleep practices, implementing such interventions remains burdensome in terms of time and cost. A key research gap in this field is understanding how to optimally implement such interventions, particularly during the busy and short birth hospitalization to minimize the burdens of implementation without negatively impacting outcomes. Doing so would allow for scalability for this and other interventions.

A limitation of this study is the validity of maternal report. Reporting bias is always a possibility when there is reliance on self-report, especially when there may be a perceived negative connotation to a particular response. Objective data has demonstrated high correlation between maternal report of infant positioning and positioning detected by sensors [43]. Additionally, given the consistency of findings across studies [11, 44], the decline in SUID rates in association with changes in self-reported sleep practices [42], and our ability to demonstrate impact in the SMART intervention [25], self-report of infant care practices appears to be a reasonable surrogate for SUID risk and for changes in actual practice.

Despite this limitation, this large clinical trial may provide hospital administrators and policy-makers with the necessary data to make practical decisions about implementation of evidence-based safe sleep education interventions among populations with historically high rates of SUID. Furthermore, this first implementation study in the context of the busy and short birth hospitalization has broader implications for effectively implementing interventions in this and similar settings without increasing staff burden and the cost of staff time.

Abbreviations

AAP

American Academy of Pediatrics

AI/AN

American Indian/Alaskan Native

CDC

Centers for Disease Control and Prevention

GEE

General estimating equation

GET SMART

Get Social Media and Risk-Reduction Training

HT

High-touch

LT

Low-touch

OR

Odds ratio

QR code

Quick-response code

SIDS

Sudden infant death syndrome

SMART

Social Media and Risk-Reduction Training

SUID

Sudden unexpected infant death

Authors’ contributions

Dr. Moon made substantial contributions to the conception and design of the work and drafted the manuscript.Drs. Hwang, Hauck, Kellams, Shafer, Colvin, Karpman, Carlin, Corwin, Colson, and Parker made substantial contributions to the conception and design of the work and reviewed and substantively revised the manuscript.All authors have approved the submitted version and have agreed both to be personally accountable for their own contributions and to ensure that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriately investigated, resolved, and the resolution documented in the literature.

Funding

This research is funded by the National Institutes of Health, Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), grant #R01HD110568. The funder has no role in the conceptualization, design, data collection, analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethics approval was obtained from the Washington University Institutional Review Board (IRB). IRB # 202401227-1149.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.