Engagement design in studies on pregnancy and infant health using social media: Systematic review

Highlights • Passive engagement design is prevalent in pregnancy and infant health studies using social media.• Interactive engagement design is used for recruitment and intervention for women.• Independent engagement design is not frequently used to study pregnancy and infant health.


Introduction
Social media has become the modern day information and communication platform for women of reproductive age. Recent social media utilization data indicate that of the US adults who use social media, 78% are women; over 80% are 18-40 years old; 63% have an annual income of less than $30,000; and 70% are Black or Hispanic . Previous studies indicate that women use social media to search pregnancy and parenting information, share information of their choice with others, and build social networks to strengthen social support (Asiodu et al., 2015;Gleeson et al., 2019;Holtz et al., 2015;Pretorius et al., 2019). Because of its potential to reach a large number of women, health experts and institutions have adopted social media as a platform to disseminate health information to the public or a target population (Dyson et al., 2017). Increasingly, researchers also use social media to engage study participants during various stages of research (Gruver et al., 2016).
The concept of patient engagement encourages researchers to make a paradigm shift from studying a health problem without the input of the patient to studying the problem with the patient's perspective in mind. Patient engagement, often in the form of serving on a study board or advisory council, can increase study enrollment and decrease attrition (Domecq et al., 2014). Patient engagement is gaining importance in healthcare, but there is no accepted definition (Markham et al., 2017). As described in a previous systematic review, patient engagement in eHealth is characterized by three dimensions: behavioral dimension, cognitive dimension, and emotional dimension (Barello et al., 2016). The behavioral dimension consists of engagement activities that enable patients to participate in self-care. The cognitive dimension promotes patient information-seeking and understanding. Lastly, the https://doi.org/10.1016/j.pmedr.2020.101113 Received 25 December 2019; Received in revised form 30 April 2020; Accepted 4 May 2020 emotional dimension helps patients connect physiological and emotional reactions, for example, during adjusting to a disease. These three dimensions of patient engagement, however, were used to assess studies on patients with severe medical conditions such as rental transplant, HIV, diabetes, or hypertension. Esmail and colleagues (2015) proposed three other engagement dimensions (context, process, and impact) for assessing patient and stakeholder engagement. The context dimension is the environment and conditions that support engagement. Availability of training for patients and organization support are examples of this dimension. The process dimension refers to who, when, and how engagement takes place. The impact dimension reflects intended effects, such as long-term or short-term outcomes from the engagement and qualitative or quantitative methods for evaluation. Based on 108 papers, mostly reviews and case studies, Esmail and colleagues (2015), however, found that some papers described impact dimension but none reported context or process dimension.
Lai and colleagues (2019) proposed a matrix for evaluating user engagement in health-specific social media platforms, which comprised five patient engagement categories from the US National eHealth Collaborative Framework (Healthcare Information and Management Systems Society, 2019) and three types of communications tools. The five engagement categories include inform me; engage me; empower me; partner with me; and support my eCommunity. The three communication tools include those that support one-way, two-way, and multi-way information dissemination and sharing. The matrix contains 15 cells for engagement evaluation. For instance, in one-way communication, users may passively receive health information (inform me), actively select the information they want to know (engage me), or assess risks and benefits of treatment options (empower me). Their matrix has not been used by researchers in systematic review. Neiger and colleagues (2012) developed a different set of performance indicators for assessing social media in health promotion, including insights, exposure, reach, and engagement. Insights refer to consumer feedback. Exposure is the number of times or impressions that content on social media is viewed. Reach is the number of people who have contact with the social media content. Engagement links social media to actions, which has three levels. Low engagement is merely acknowledging an agreement or preference. Medium engagement involves sharing content with others. High engagement is actual participation in off-line activities.
Extended from Neiger and colleagues' performance indicators, Platt and colleagues (2016) further developed a framework to evaluate Facebook engagement for public health communication. Their framework has non-user and user engagement factors. The non-user engagement factors include context (social, political, and cultural environment), content (quality and content of advertisements), and setup variables (advertisement type, target audience, and budget). User engagement activities are grouped into six levels. Level 1 is observation measured by the number of people reached by a health campaign. Level 2 is "likes" submitted by the Facebook audience. Level 3 is exploration measured by website clicks on content. Level 4 is connection by sharing content. Level 5 is conversation by posting comments in discussion forums. Level 6 is implementation by engaging in activities outside Facebook. Levels 1, 2 and 3 are considered lower levels of engagement; levels 4 and 5 higher engagement; and level 6 is the highest engagement.
To date, no published systematic review has reported patient engagement design in studies on pregnancy and infant health that incorporate social media. We chose pregnancy and infant health as a focus area for this systematic review for two reasons. As previously mentioned, a large number of social media users are women of reproductive age. Childbearing and childrearing are important health topics for these women (Gleeson et al., 2019;Pretorius et al., 2019). Findings of our systematic review will have potential impacts on designing social media studies focused on this large group of women. Moreover, pregnancy and infant health involves an array of interventions with some targeting general health promotion and others specific disease management. A review of literature on social media research on pregnancy and infant health can enrich our understanding of designing social media studies in response to health promotion and disease management for the pregnant and infant population. The purpose of this systematic review was to analyze participant engagement design in studies that used social media to address pregnancy and infant health issues.

Literature search and selection
We define social media as an internet-based social networking platform to facilitate communication, information sharing, and human interaction. We used several key words to search publications in EBSCO and PubMed databases, including (pregnant OR gestational OR maternal OR infant) AND (social media OR Facebook OR Twitter OR Instagram) AND (health campaign OR health promotion OR health). EBSCO consists of MEDLINE, CINAHL Complete, and PsycINFO databases. We further applied filters to narrow the search. These filters included academic journals, English language, full-text, publication years from 2010 to July 2019, and human subjects articles. We limited publication date to 10 years in order to focus on most recent articles. A previous integrative review on eHealth and patient engagement also adopted a 10-year timeframe (Barello et al., 2016). To be included in this systematic review the article had to address pregnancy, postpartum or infant health; report quantitative results from primary or secondary analysis of a study; and include a complete or partial role of social media usage in research. Excluded from our analysis were study protocols, qualitative studies, and studies that focused social media data mining without a mention of pregnancy or infant health, social media as an information-seeking source without participant engagement information, or technology aspect of social media development. We also excluded articles that reported social media in improving communication among care providers.
As shown in the PRISMA (Fig. 1), 467 articles from EBSCO and 290 articles from PubMed were found. After removal of duplicates, 235 articles remained. Reviewing abstracts against inclusion criteria further eliminated 101 articles. A member of the study team performed the literature search and abstract review. Two study team members read the text of the 134 remaining articles and used the inclusion and exclusion criteria to retain articles. Each article was labeled as "relevant" or "not relevant." Articles labeled as "not relevant" must have had a reason either of not meeting the inclusion criteria or meeting the exclusion criteria. In this stage, 103 articles were labeled as "not relevant" and were eliminated. The final analysis included 31 articles reporting 30 studies (two articles reported the same study with different study aims). Of the 30 studies, four were randomized controlled trials (RCT), four non RCT interventions, and 22 observational or descriptive studies.

Literature analysis
We developed an assessment matrix that included three engagement designs (passive, interactive, and independent) and three research stages (recruitment, data analysis, and intervention). The three engagement designs were derived from the performance indicators for social media by Neiger et al. (2012) and the Facebook-based engagement framework for public health communication by Platt et al. (2016) but were in a simpler format. We believe our matrix is less cumbersome and suitable for assessing Facebook and non Facebook social media studies relevant to health promotion issues. The matrix by Platt et al. was used for developing a newborn screen and biobank program; therefore, the assessment matrix that we developed for our analysis is likely to be appropriate for evaluating studies on pregnant and parenting women. Definitions of these designs in various research stages and positives/negatives of each design can be found in Table 1.
We examined each article and identified how engagement designs were utilized for recruitment, data analysis, and intervention. Passive engagement design is used to find, inform, and enroll participants during recruitment, access participant data from social media for data analysis, or retain study participants in intervention. Interactive engagement design facilitates study enrollment during recruitment or fosters interaction of study participants with the research team or with fellow participants during intervention. Independent engagement design encourages participants to engage in activities outside the social media platform, such as preforming additional steps to enroll in a study, self-monitoring health behavior, or going to a focus group meeting during intervention. Two team members extracted information from each article and cross-checked the accuracy of information as shown in Table 3. Inter-rate agreement was 84%. Disagreement was solved by reexamining the articles.

Results
Our analysis included 31 articles, which contained data from 30 studies (see Table 2). Among the 31 articles, 17 reported studies conducted in the United States, 4 in Canada, 3 in the United Kingdom, 3 in Australia and New Zealand, 1 in Ireland, the Netherlands, and Brazil each, and 1 unknown. Facebook was the most frequently used social media platform (n = 17) followed by Facebook combined with Google, Craigslist, university clinical trial website, local city classifieds or individual study websites (n = 8); Twitter (n = 3); and social media sites created by researchers or special interest groups (n = 3). Health topics reported in the 31 articles varied from gestational weight gain, pregnancy health, abortion, and postpartum anxiety to breastfeeding, infant vaccination, birth defects, contraception, and drinking during pregnancy. The main purpose of using social media was for recruitment (n = 16), data analysis (n = 6), intervention (n = 8), or both recruitment and intervention (n = 1).
As shown in Table 3, of the 17 articles that described using social media to recruit participants, all but two used passive and interactive engagement designs and only one article reported using independent engagement design. For articles with data analysis as a focus (n = 6), passive engagement was the primary design reported in 6 articles and interactive engagement (n = 1) and independent engagement (n = 0) designs were not common. Articles, reporting the use of social media for intervention (n = 9), included passive (n = 9), interactive (n = 8), and independent engagement designs (n = 3) with independent engagement design to a lesser degree.

Engagement design for recruitment
Passive engagement design for recruitment involved using paid and non-paid social media to reach out to potential study participants (Table 4). Researchers sent advertisements via paid services such as Facebook newsfeed and Google AdWords to potential study participants. Facebook advertisements could target a specific group of women, such as women aged 18-44 residing in the US for a study on childbirth preference among nulliparous women (Arcia, 2014). Google AdWords could also display paid advertisement text to users who happen to use similar search words. For instance, Van Gelder and colleagues (2019) chose 20 search terms related to pregnancy in Google AdWords to recruit pregnant women less than 17 weeks pregnant for a longitudinal study on prenatal health and infant development. When using paid services in passive engagement design, researchers could adopt the lowest cost bid through an automatic bidding/auction system in Facebook and Google AdWords, or a preset capped budget (Arcia, 2014;Graham et al., 2019;Laws et al., 2016).
Other researchers used non-paid advertisements, which were free but researchers may request permission to access some social network groups or compete with others who also wanted to post advertisement in the same social media site. For instance, Herbell (2019) obtained permission from 61 Facebook parent discussion private group owners and posted study flyers in those Facebook pages to recruit pregnant women. MacDonnell and colleagues (2019) used free Craigslist advertisement to post study information when recruiting women into their Contraception and Alcohol Risk Reduction Internet Intervention. Shere and colleagues (2014) used drug-sponsored sites, local classifieds, pregnancy discussion forums, and message boards to disseminate study information to recruit women not taking folic acid three months before pregnancy.
Passive engagement could be used to reconnect lost-to-follow-up study participants with a study. Stephenson and colleagues (2019) identified lost-to-follow-up participants from Facebook profiles by names, birth date, home address, email and alternate contact in friends list. These lost-to-follow people were invited back to the study.
Interactive engagement design during recruitment was to engage participants in completing enrollment process. To complete enrollment, potential participants clicked on a hyperlink in a recruitment site to go to a study website. In the study website, women could read detailed study information (Graham et al., 2019), answer questions for eligibility screens (Emery et al., 2018), give electronic consent (Admon et al., 2016;Ashford et al., 2018), send emails to the study team (Harris et al., 2015), or sign up for the study (Van Gelder et al., 2019). In six articles, completing enrollment was demonstrated by participants clicking on a link to fill out surveys or questionnaires (Admon et al., 2016;Altshuler et al., 2015;Harpel, 2018;Harris et al., 2015;Holtz et al., 2015;Moore et al., 2017).
Only one article described using independent engagement design for recruitment, in which study participants met study staff in the hospital to go over consent (Shere et al., 2014).
It was noted that when passive engagement design was used for recruitment, some researchers calculated click through rate (CTR), which is the proportion of women clicked on the advertisement in relation to the women who viewed the advertisement, to estimate their reach to potential study participants. CTR varied among studies ranging from 0.06% to 0.08% (Arcia, 2014) to 5.80% (Graham et al., 2019). Other researchers reported cost when using social media for recruitment. Arcia (2014) reported the cost per click when using Facebook advertisement as $0.63. In the studies conducted in the UK and Canada, cost per click for Facebook advertisements (€0.53 and CAD$0.35, respectively) seemed to be cheaper than those for Google AdWorks (€1.33  and CAD$0.76, respectively) (Emery et al., 2018;Graham et al., 2019). Other measures were cost per completed survey (Admon et al., 2016) and cost per eligible participant (Arcia, 2014;Harris et al., 2015;Laws et al, 2016).

Engagement design for data analysis
Six articles described using social media for data analysis ( Table 5). All of them reported passive engagement, one addressed interactive engagement, and none described independent engagement. The goal of passive engagement for data analysis was to access data posted in social media by participants. Researchers applied different methods to locate social media data, such as using an automatic classification system to identify pregnant women experiencing a birth defect from Twitter postings (Golder et al., 2019). Meaney and colleagues (2016) used hashtags to capture tweets and re-tweets related to fetal death. Marshall and colleagues (2019) adopted the Facebook application program interface at two time points to capture user postings pertaining to pregnant weight gain, food, and exercise. Perrin and colleagues (2014) classified postings into four groups based on original or reply postings related to breast milk sharing: original offer, original request, reply offer and reply request.
The one article that addressed interactive engagement design for data analysis involved asking study participants to grant data collectors access to their Facebook accounts (Marshall et al., 2019). Optimal weight gain Tested the feasibility of an intervention with mother-infant dyads to promote recommended GWG in primigravidas, optimal infant feeding, and return to pre pregnancy weight *: Articles are from the same study with different focuses.  • Facebook Ads: paid messages displayed to predefined members using an automatic bidding system to achieve the highest number of clicks and the lowest cost. Budget at Canadian $26 for Ads run continuously.
• Facebook: 772,263 impressions. The average number of times an Ad was displayed to the same individual ranged from 2.53 to 3.28.
• A user clicked on the Ad to go to campaign website which contained healthy weight gain related information, calculator to find BMI and recommended weight gain range. •

Engagement design for intervention
Nine articles including eight studies reported using social media for intervention (Table 6). Almost all of them used passive and interactive engagement designs but only three demonstrated independent engagement design. Passive engagement design for intervention mostly was to maximize the accuracy of reach to potential participants and to retain enrolled participants. Some researchers used zip codes to select women into the treatment group and to verify the accuracy of selection with the IP address and friends profile locations in Facebook (Byker et al., 2019). Other researchers assigned women into small Facebook groups to facilitate retention (Fiks et al., 2017). Passive engagement design to retain participants also included regularly posted videos, infographics, messages, and photos on websites or Facebook to update study information for participants (Cavalcanti et al., 2019;Silfee et al., 2018).
Interactive engagement design for intervention had two functions: participant interaction with the study team and interaction among fellow participants. Interaction could be done via blogs, discussion forums or chat rooms. In some studies, a facilitator would invite participants to a virtual discussion or support group, lead group discussion, and communicate one-on-one with each individual participant. In the study conducted by William and colleagues (2019), the promotora (community health worker) invited Spanish-speaking participants to a Facebook group and provided one-on-one interaction to promote optimal gestational weight gain and infant feeding. In Cavalcanti and colleagues' study (2019), women in the intervention group were invited to join a closed Facebook group called Projeto Amamenta Mamãe (Mama Breastfeeding) where women could raise doubts, make comments, and share experiences with each other. Interactive engagement • Google search to identify 8 social networking sites, 2 general pregnancy forums selected • 704 participants across both sites. These members posted 525 support-seeking messages and 1965 support-providing messages.
• Original posts and the first 10 responses for a month period were analyzed.

NR NR
Marshall 2019 • Extraction of Facebook posts by women recruited from primary clinics was performed using the Facebook application program interface (API) at two separate occasions in 2016-once upon recruitment (usually in the 1st and 2nd trimesters) and again in the 3rd trimester.
• Posts pertaining to weight gain, food and exercise were identified by finding keywords (e.g., craving, food, fat, exercise) and their morphological variants using Natural Language Processing (NLP).
• Participants granted data collectors access to their accounts by signing into Facebook during a standard clinic visit. Access was immediately terminated after the participants logged out of their accounts.
• Of all mined posts (n = 2899), 311 included information relating to health behaviors in pregnancy.

Meaney2016
• Twitter: Hashtag and terms related to "fatal failures" or fetal death were used to search posting.

NR NR
Parackal 2017 • Facebook advertising to deliver public health messages: a video and three banner advertisements  • Contraceptive information campaign carried out on social media.
• Facebook users in the treatment areas (zip codes) received 3 informational Ads on efficacy, easy-of-use, and safety.
• Facebook: residence based on the "current city" validated with IP address and friends' profile locations.
• About 80% of the targeted population saw at least one advertisement.
• 82% of the Ads appeared in the desktop right column, 12% appeared in the mobile newsfeed, and 6% appeared on third-party mobile apps and websites.
• Facebook accurately targeting these advertisements by zip codes: 87.4% of respondents from the study region self-reported zip codes in the intended treatment group.

NR
• Women went to one of the 21 Planned Parenthood of Northern New England Health Centers to receive long-acting reversal contraception.

Cavalcanti 2019
• Mothers were recruited from maternal ward of a hospital • Facebook closed group called Projeto Amamenta Mamãe (Mama Breastfeeding) for 24 weeks.
• The women were tagged in a post of the group based on a weekly topic • Each mother received a virtual invitation that she should accept to allow her inclusion in the intervention group (IG).
• The post had messages of encouragement and clarification.
• A conversation to raise doubts, comment on something, or share their experiences on the specific topic of the poster.
• Each tag or participation in the group generated an automatic and immediate notification on the personal profile of each woman, enabling communication in real time. • 1 or 2 blog posts per month: covering timely or controversial issues such as new vaccine safety research, recent vaccine-preventable disease outbreaks, or changes in state or national immunization policies.
• An "Ask a Question" portal: VSM participants could direct questions to vaccine experts (a vaccine safety researcher, a pediatric infectious diseases specialist, a general pediatrician, and a risk communication specialist).
• For any questions submitted privately through the portal or by e-mail, personalized responses were provided within 2 business days • Real time monthly online chat sessions: VSM group could converse with a team of vaccine experts.
• All interactive website and social media components were moderated to prevent bullying, abusive language, and disclosure of personal health information.
• Surveys administered online using a secure platform.
•  Shieh, et al. Preventive Medicine Reports 19 (2020) 101113 design also included real time and instant feedback. For instance, women in a Facebook study group on exclusive breastfeeding received an automatic and immediate notification on the personal profile of each woman, enabling communication in real time (Cavalcanti et al., 2019). Independent engagement for intervention aimed to engage participants in off-line activities. Ashford and colleagues (2018) asked study participants to do seven weekly practice modules to promote mental health and have 30 min of telephone coach support weekly outside social media engagement. Byker and colleagues (2019) delivered Facebook health campaign about long-term reversible contraception to women in treatment locations and then measured independent engagement in getting this contraception in clinics. Silfee and colleagues (2018) asked women before joining a Facebook group to engage in a 90-minute in-person orientation session.

Discussion
In this systematic review, we analyzed passive, interactive, and independent engagement designs for recruitment, data analysis, and intervention in 31 articles that reported pregnancy and infant health studies involving social media. Major findings from our review included (1) passive engagement design was prevalent in all studies and was utilized at various research stages, including recruitment, data analysis, and interaction; (2) interactive engagement design was common for recruitment and intervention; and (3) independent engagement design was used the least and it appeared mostly during intervention.
Although passive engagement is a term relative to the participant's perspective, it is a fundamental design initiated by researchers to access a data source, whether that is the participant or the data posted by the participant in social media, as well as to retain participants in a study. We found that researchers used passive engagement design through paid or non-paid social media platforms to reach out to potential participants. For data analysis, passive engagement design enabled researchers to access the participant's existing data already posted in social media. For intervention, passive engagement involved regularly sending participants study related information to increase retention.
A unique passive engagement design in social media research identified was the use of commercial functions in social media to recruit study participants. Paid advertisements in Facebook and Google AdWords could provide pre-selected parameters such as subject demographics, geographic locations, search words, and budget to facilitate recruitment (Emery et al. 2018;Graham et al., 2019, van Gelder et al., 2019. We also found that many researchers believed that social media platforms were a quick way to recruit participants for survey studies (Admon et al., 2016;Arcia, 2014). Shere and colleagues (2014), using interruptive time-series analysis, even identified social media as a significant reason for the increase in recruitment in their study. A disadvantage of using social media for recruitment is the potential for fraudulent application submissions by social media users (MacDonnell et al., 2019). However, researchers may avoid the problem by adopting interactive or independent engagement design. Potential participants can answer eligibility questions online or meet with study staff to complete the eligibility assessment.
Using social media to reengage lost-to-follow-up study participants is an innovative passive engagement design. Stephenson and colleagues (2019) used passive engagement design in their longitudinal study to find study participants who could not be contacted after numerous attempts by phone or email. One reason to support this approach was that participants often used Facebook URLs longer than email addresses or phone numbers. Subject attrition is a common problem especially for longitudinal studies. Future research may further explore the impact of social media on retention, such as how different social media platforms • Each message was displayed as a tweet from a Twitter account.
• Experimental group exposed to 9 multivitamins promotion messages presented in a randomized order.
• Participants re-tweeted messages. • The study employed a self-administered online survey as the data collection method.

Silfee 2018
• Recruitment from WIC program and electronic records • Videos and pictures from the in-person protocol were included in posts where applicable and supplemented by additional photos, infographics, and videos extracted from webbased sources with special attention to maintaining the original message.
• First 8 weeks: Facebook intervention posts were delivered 2 times per day, with additional posts from coaches aiming to stimulate interaction among participants or respond to participants' questions and challenges.
• For the following 8 weeks: posts were delivered once per day without additional coaching. • In order to join the private Facebook® group, participants maintained an active Facebook® user account and became Facebook® "friends" with the promotora.

NR
NR: Not relevant *: Social medial also used for recruitment **: Articles from the same study but reported different components. C. Shieh, et al. Preventive Medicine Reports 19 (2020) 101113 work to reengage lost subjects. Interactive and independent engagement designs require active participation of the participant. Our analysis found various interactive engagement designs especially for participants to complete study enrollment, fill out surveys, or participate in intervention. For instance, participants could click on a Facebook advertisement to link to a study website in order to fill out a survey (Altshuler et al., 2015) or to send a message to trigger automatic support (Emery et al., 2018). Interactive engagement also occurred between the research team and the participant or among fellow participants particularly in intervention studies. Cyber bullying may potentially be a problem during online interaction. Many researchers set up rules to prevent bullying and abusive language in posting (Daley et al., 2018). This practice should be encouraged for all studies using Facebook platform or other study-specific websites. The intention of engaging participants in online dialogue is to share personal thoughts and opinions. Negative postings, if not regulated, may result in the opposite effect.
Based on our review, independent engagement design was not frequently adopted by researchers. Only four articles described this design and three of them involved intervention. It is likely that intervention studies do need participants to engage in off-line activities in order to induce actual behavior change; therefore independent engagement design was mostly adopted in interventions. Additional measures also have to be developed in order to assess actual off-line activity attendance, which could have limited its utilization by researchers. Future research may compare health behavior outcomes of study participants that engage in off-line activities versus those without such activities. Furthermore, future research may investigate lurkers' (i.e., who read postings but do not post) reactions to off-line activities compared to those of active participants. Many studies in our review reported click through rates or cost per click. This represents another area for future research, as a range for these values, relevant to study design or sample size, has not been established.

Strengths and limitations
To our knowledge, our review is the first systematic analysis to report engagement design in social media-based studies on pregnancy and infant health. The evaluation framework used in our analysis, including passive, interactive, and independent engagement designs, is simple and easy to understand. Our framework was derived from previous studies for Facebook engagement and other social media (Neiger et al., 2012;Platt et al., 2016) but in a much simpler format for assessing Facebook and non Facebook social media. Another unique strength of our review is identifying the application of these engagement designs in the research process, which is different from previous reviews. Depending on project purpose, a researcher may adopt a suitable engagement design to enhance recruitment, analysis, or intervention. A limitation in our review, however, is that we included studies of various pregnant and infant health topics and publication years of only 2010-2019. Our findings are not generalizable to a specific health topic and are limited to findings from the past decade only. Another limitation is that we did not analyze effects associated with each engagement design. For instance, we did not compare recruitment results when using passive engagement designs with paid or non-paid social media platforms. Although we report various engagement designs in our review, we are unable to conclude which is more effective than the others. Furthermore, studies included in our review adopted Facebook and other Internet-based platforms. We did not separate engagement designs in Facebook from those in non-Facebook platforms. Social media use is influenced by availability of technology, geographical locations, and race/ethnicity. Over one half of the articles in our review reported studies conducted in the US. Our review findings therefore may not have a global implication.

Conclusions
Social media-based studies on pregnancy and infant health utilize various engagement designs during different stages of research. Regardless of study purpose, passive engagement design is commonly used for recruitment and retention of study participants, and for access to data in social media. Interactive and independent engagement designs encourage active participation of the participant. These designs are often used for interventions. Choosing a suitable design may enhance the probability of success in recruitment, data analysis, and intervention.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.