This systematic review was conducted according to the Cochrane recommendations and was reported in accordance with the updated guidelines of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [17,18]. The review protocol was registered in the International Prospective Register of Systematic Review (PROSPERO; February 27, 2023, registration number CRD42023400177) [19].

This review included all types of study designs published in the English language, which are original research reported in peer-reviewed journals.

Studies were selected according to PICO (Population, Intervention, Comparison, Outcome) criteria [18]. The selection of studies is outlined in Table 1.

The population included social media users of all age groups who assessed the social media posts that were delivered on existing, commercial social media platforms such as Facebook and Instagram (Meta Platforms). Social media posts were aimed at reducing any of the 4 key lifestyle risk factors, and the posts were created by HCPs. In our study, HCPs represented part of the health workforce in various health settings such as hospitals, clinics, community health centers, research institutions, academic institutions, and health organizations. The social media posts in the included studies were examined for elements that were linked to user engagement metrics. The term “elements” in this review refers to all the components found in social media posts that could be deduced either from the outlook of the post itself (eg, image and poll) or from its content (eg, informative post). User engagement metrics included direct interactions performed by social media users toward the posts that were reported numerically [20,21]. User engagement metrics were restricted to quantifiable postlevel interactions as they represent the most objective and interpretable measures to compare study outcomes over time across published studies [22].

A search strategy comprising controlled vocabulary (eg, MeSH [Medical Subject Headings]) and free text terms informed by previous literature [21,23,24] was developed and reviewed by 2 authors (YYY and WWC). The search strategy was structured into 4 concept headings which are elements in social media posts and their derivative terms, social media platforms, lifestyle risk factors, and outcome measures.

A literature search was conducted in 4 electronic health databases (MEDLINE or OVID, Scopus, Web of Science, and CINAHL) using the designated search strategy to identify relevant studies for inclusion. Multimedia Appendix 1 outlines the search strategies for all 4 databases. Given the focus on peer-reviewed studies, a gray literature search was not conducted. The search was restricted to studies published in English, between January 2006 (the year when X (previously known as Twitter; Twitter, Inc) and Facebook were publicly accessible, according to the review by Chen and Wang [23] until June 2023. Additionally, OVID auto alerts were used to monitor and include any newly published papers until March 31, 2024. The references from included studies and related systematic reviews were also screened to identify further eligible studies.

The initial systematic literature search involved a single researcher (YYY) who screened and reviewed the titles and abstracts. Full-text papers from potentially relevant studies were retrieved and assessed for eligibility based on the inclusion criteria. Subsequently, another researcher (WWC) reviewed all included studies. Any discrepancies or disagreements were resolved through collaborative discussions among the study authors. The title, abstract, and full-text screening were completed on Rayyan (Rayyan Systems, Inc) [25].

A standardized Microsoft Excel sheet was developed for data extraction. Extracted data included (1) publication details (author, publication year, and country); (2) study design; (3) target population; (4) sample size; (5) description of posts (lifestyle risk factor, social media and post creator, and number and duration of posts); (6) a brief description of methods involving delivery of posts; (7) investigated elements and user engagement metrics; and (8) elements that influenced user engagement metrics.

A single researcher (YYY) extracted data from the included studies. Accuracy was ensured by cross-checking the extracted data with another researcher (WWC). Any discrepancies were discussed and resolved through consensus among the study authors.

The investigated elements encompassed all elements in included studies that examined social media posts, that may or may not have influenced user engagement. Due to the variability of the investigated elements, a narrative synthesis was performed to meticulously synthesize the findings from the included studies. All investigated elements were categorized based on their descriptions as reported in the studies. The initial categorization was carried out by the first researcher (YYY) and was further refined by a second researcher (WWC). The finalized categories were collectively reviewed and assessed by all authors, with any discrepancies resolved through consensus.

The included studies were then reviewed to identify the elements in social media posts that influenced user engagement metrics. The identified elements that influenced user engagement were determined based on 2 criteria—those demonstrating the highest measured user engagement and those for which user engagement was reported as significant during univariate or multivariate analysis.

As the studies varied in research design, methodological quality assessment was conducted using appraisal tools according to each research design. Joanna Briggs Institute (JBI) critical appraisal tools were used for cross-sectional studies, randomized trials, and quasi-experimental studies [26] whereas, the Mixed Methods Assessment Tool (MMAT) was used for mixed methods studies [27].

The methodological quality of included studies was tabulated with responses for each item assigned (yes, no, unclear, or not applicable for JBI critical appraisal tools and yes, no, or cannot tell for MMAT). Quality assessment was completed by the first researcher and was checked by a second researcher.

A total of 2458 studies were identified from the database search, OVID auto alerts and citation search. The full text of 242 papers was assessed for eligibility. A total of 19 studies met the inclusion criteria and were included in the systematic review. Figure 1 outlines the selection of eligible studies.

The reviewed studies are summarized in Multimedia Appendix 2. The study designs for included studies were cross-sectional (n=10) [14,28-36], mixed methods (n=3) [15,37,38], quasi-experimental (n=2) [39,40], and randomized trials (n=4) [11,41-43]. A total of 3 out of 4 randomized trials [11,41,43] were subgroup analyses of previously published trials [44-46]. Studies included were from North America (n=10) [11,15,31,34,35,37,39,40,42,43], Australia (n=3) [14,29,41], Europe (n=3) [28,33,36], Great Britain (n=1) [38], South America (n=1) [32], and Asia (n=1) [30]. All 19 studies depicted posts promoting healthy lifestyle behaviors in reducing risk factors for NCDs. In total, 18 studies targeted healthy populations and 1 study aimed at reducing NCD-related complications among diabetic patients [28]. The examined topics included tobacco-related posts (n=11) [29-36,39,40,42], behaviors to reduce obesity (n=4) [11,14,37,43], posts promoting physical activities (n=2) [28,41], dietary habits (n=1) [38] and multiple lifestyle behaviors (n=1) [15]. The social media platforms used were Facebook (n=15) [11,14,15,29,31-33,35-37,39-43], Weibo (Sina Corporation; n=1) [30], and Instagram (n=1) [38] with 2 studies using multiple platforms of Facebook, Instagram, and Twitter [28,34]. The average number of posts examined was 710 (range 2-3515). In terms of the delivery of posts, 10 studies had posts with elements categorized before being posted on social media [11,15,29,34,37,38,40-43], whereas 7 studies collected existing social media posts and subsequently categorized the elements in the posts [14,28,30,31,33,35,36]. Two studies had posts with elements categorized before being posted as paid Facebook ads [32,39].

Quality assessment for included studies is outlined in Multimedia Appendix 3. All studies had high or moderate quality, having at least 50% of items labeled as “Yes” (4 out of 8, 50% items for cross-sectional studies; 7 out of 13, 54% items for randomized trials; 5 out of 9, 56% items for quasi-experimental studies; and 9 out of 17, 43% items for mixed methods studies). All cross-sectional studies had valid and reliable measures of exposures and outcomes, with appropriate statistical analysis used. A total of 4 cross-sectional studies addressed confounders through the conduct of multivariate regression analyses [14,29,31,34]. Almost all randomized trials did not blind their participants, investigators, and outcome accessors, with the exception of Tomayko et al [43] that blinded its participants. Both quasi-experimental studies did not have any control groups present [39,40]. Among the mixed methods studies, only 1 study addressed confounders [37].

Table 2 describes the categories of elements in social media posts and user engagement metrics. The elements were grouped into 9 categories based on their characteristics. These categories were communication using supportive or emotive elements, post appearance, communication toward behavioral changes, post topics, requests for direct interaction with the post, tailoring of post content toward the targeted audience, source of post content, social media platform, and day and time of post. A total of 6 types of user engagement metrics were identified, which are likes, comments, shares, emojis, clicks, and votes.

Multimedia Appendix 4 shows the categories of elements and user engagement metrics tabulated according to the study. A total of 14 studies [11,14,15,28-31,33,36-39,41,43] reported likes as outcomes of user engagement metrics, with the other metrics involved being comments (n=13) [11,14,15,28,30,31,33,35,37,39-42], shares (n=8) [14,28-31,33,37,39], emojis (n=4) [14,29,36,39], clicks (n=4) [14,32,34,39], and votes (n=3) [11,37,41] (see Multimedia Appendix 4).

The breakdown of investigated elements for each study with the subsequent elements that influenced user engagement metrics is outlined in Multimedia Appendix 5. The PRISMA 2020 checklist for the systematic review is provided in Multimedia Appendix 6. A total of 35 elements from the 9 categories were found to influence user engagement metrics. Table 3 summarizes the definitions of all 35 elements that influenced user engagement metrics.

Communication using supportive or emotive elements and the appearance of posts were the 2 most featured categories with elements influencing user engagement, each appearing in 8 studies [11,14,15,28-31,35,37-39,41,42]. Among the elements under supportive or emotive communication, the informative post was the most popular element (n=5) [15,30,31,35,38]. Networking with other users was also an effective way of delivering support (n=4) [28,30,31,38], as was providing tangible and intangible assistance (n=2) [28,42]. Regarding the appearance of posts, polls (n=4) [11,15,37,41] and videos (n=4) [15,29,31,39] were the 2 most featured elements (n=4), followed by images (n=2) [14,37].

Communication elements leading to behavioral changes influenced user engagement in 6 studies [14,30,32,33,40,42]. The elements were predominantly based on various behavioral models, including the Health Belief Model (HBM) [30], the Theory of Planned Behavior (TPB) [30], and the Transtheoretical Model (TTM) [40,42]. Additionally, 3 studies used non-model elements such as call-to-action, message framing toward loss of outcomes, and person-centered motivational interviewing approaches [14,32,33].

The results of 5 studies show that topic-based elements focusing on positive lifestyle behaviors for risk reduction can increase user engagement [11,14,28,38,43]. These topics include encouraging dietary habits, such as drinking water (n=3) [11,14,43], promoting physical activity through exercise (n=3) [28,38,43], and general well-being topics (n=1) [38].

Social media posts that requested users to interact directly with the post influenced user engagement in 4 studies [11,15,35,36]. Posts that asked questions and encouraged discussion were found to promote engagement in 3 studies (n=3) [11,15,35]. However, engagement baits such as “like this post!” were found to decrease engagement in 1 study [36].

The tailoring of post content to targeted audience groups was shown to influence user engagement in 3 studies [14,29,34]. According to Hefler et al [29], posts that incorporated local branding to generate trust among the local community resulted in increased user engagement. However, the use of hashtags marked by “#” to enhance post searchability showed an unusual trend of decreasing user engagement. On the other hand, amplifying post reaches through payment, as demonstrated by Kite et al [14] led to increased user engagement. In Reuter et al [34], unpaid, organic posts were preferred by users.

The source of post content has been found to impact user engagement in 2 studies [29,30]. According to Jiang and Beaudoin [30], original and unpublished content has a positive influence on user engagement. Hefler et al [29] observed positive user engagement in posts containing externally sourced content from previously published material, whether presented unchanged or with minor modifications.

The selection of social media platforms was found to increase user engagement in 2 recent studies examining similar posts on multiple platforms [28,34]. Facebook and Instagram were found to be the preferred platforms in these studies. Additionally, the timing of postings was shown to have an impact on user engagement. Kite et al [14] found that posts made on Fridays between 8 AM and 5 PM received higher user engagement, whereas posts made on Mondays generated lower user engagement.

To the best of our knowledge, this is the first systematic review to identify elements in social media posts that influenced user engagement metrics. This study addressed the current knowledge gaps related to how HCPs can create posts for optimal user engagement. Our analysis has identified 35 elements across 9 categories that influence user engagement metrics. Communication elements with supportive and emotive elements that encourage behavioral changes, as well as the appearance of posts were the dominant categories that have a positive impact on user engagement. By prioritizing these elements, we can potentially maximize the effects of health promotion by HCPs through social media. However, the categories of source of post content, social media platforms, and post timings had less than 3 studies showing elements that affect user engagement. Therefore, more studies are needed to confirm the findings in relation to these elements.

It is worth noting that at least three-quarters of the studies on social media posts were conducted in high-income countries, which is not surprising since these countries have more developed digital information infrastructure [48]. Our findings mirrored the review by Elaheebocus et al [49] that focused on targeted behaviors on social media, where approximately half of the studies examined tobacco-related behaviors, while the other half focused on physical activity promotion, healthy dietary habits, and weight control. Interestingly, there were no studies on alcohol-related posts, despite the fact that marketing brands tend to promote alcohol more frequently on social media [50]. This is concerning because HCPs are already sharing limited health-related information on social media platforms [51]. It may be important to increase health promotion pertaining to reducing and abstaining from alcohol consumption on social media platforms.

Communication using supportive or emotive elements was one of the categories with the most studies on elements that affect user engagement. In our review, most of the supportive elements that led to high user engagement were provided through informative posts, audience support, and assistance such as live counseling support [42] and diabetic health tools [28]. Tagging other health organizations in posts increases users’ trust and allows them to reach out for external assistance. Emotive elements such as humorous posts that adopted contemporary memes also yielded higher user engagement. However, they need to be used alongside informative elements [38]. In addition, posts that evoke negative or unpleasant feelings were also found to increase user engagement. This is supported by evidence in experimental psychology, whereby the presence of negativity bias leads to greater user responsiveness toward negative stimuli [52] resulting in higher postlevel interactions. The context in these posts may still be aligned with positive lifestyle behaviors. For example, the post “Banning vape leads to black market sales which is worrying” shows that the user is worried despite agreeing with the vape ban [31].

The appearance of posts is the second most studied factor influencing user engagement. Polls, videos, and images were the 3 elements grouped under this category. The conduct of polls positively influenced user engagement as users can view the immediate results and participate in discussions under the poll post, leading to higher comments. Photos are static visuals with straightforward messages that draw users’ immediate attention and positively influence user engagement [39]. Despite videos requiring more attentive processing, they still have a positive influence on user engagement [14,29,31,39]. With the introduction of shorter videos that require less attention span, it is expected that videos will continue to increase user engagement.

Elements of communication toward behavioral changes that influenced user engagement were mostly based on existing behavioral models. The 3 models studied included the HBM, the TPB, and the TTM [30,40,42]. The health behavioral models were used in the creation of posts on smoking prevention and cessation, with 2 studies using the TTM [40,42]. In the TTM, decisional balance focuses on the advantages and disadvantages of behavioral changes. Decisional balance is consistent with motivational interviewing techniques, whereby mixed feelings are acknowledged before users are guided toward the advantages of behavioral changes [33,40]. Such elements are effective among younger smokers with lower motivation to quit [40,53]. Posts based on consciousness raising in the TTM were more effective among smokers in the midst of quitting and who required additional health information in quitting [40]. Although TTM elements of dramatic relief (ie, eliciting negative emotional responses to old behaviors of smoking and positive emotional responses to new behaviors of quitting smoking) and self-liberation (ie, firm commitments to quitting) reduced engagement levels, authors in the study were optimistic that reduced engagement could be reversed by eliciting positive emotional responses to new behaviors instead of negative emotional responses and restrategizing self-liberation elements into more organized, incremental methods [40].

Other elements of communication toward behavioral changes that positively influenced user engagement included call-to-action and loss-framed posts [14,32]. Findings of loss-framed posts influencing user engagement by mentioning the losses of a behavioral outcome were supported by Graham et al’s [54] study assessing smoking cessation advertisements delivered through websites. However, gain-framed messages were still effective in studies exploring messages that convinced active smokers to quit smoking [55,56]. The effects of framed messages should, therefore, be explored further.

In several studies, 3 categories of elements—post topics, requests to interact directly with the post and targeted content have been found to influence user engagement. First, among the post topics, instructional posts delivered directly (eg, exercise more) or indirectly (eg, recipe ideas toward a healthy diet) have been shown to increase user engagement [11,14,28,38,43]. Instructional posts simplify users’ comprehension of what actions are required and how to perform them [57]. Second, questions asking for suggestions or discussion act as a cue for users to directly comment on the post [11,15,35]. Users are also inclined to further engage with the post by liking or sharing them. In contrast, “engagement bait” strategies deployed on Facebook using phrases such as “like this!” have been found to reduce user engagement. The authors hypothesized that lower engagement was due to Facebook demoting such posts from users’ newsfeeds, causing them to appear less frequently on users’ social media timelines [36,58]. Third, tailoring content toward targeted audience groups has also been found to increase user engagement. Information delivery to a specific community should include a personalized touch by incorporating local elements, such as community logos or local dialects [29]. The inclusion of hashtags in posts can increase their discoverability, making it easier for users to search for them. However, Hefler et al [29] observed that adding hashtags to posts shared from other pages or profiles might lead to lower user engagement. This is because hashtags on shared posts may increase the length of a post, which might reduce its legibility and aesthetic appeal. As both organic and paid posts positively impacted user engagement [14,34], selection for payment should be based on budget availability and target audience group.

The source of post content and social media platforms were found by 2 studies to influence user engagement metrics. The adoption of post content from previously published sources would mean that a meticulous selection of high-quality content has been made, which generated higher user engagement [29]. A study by Waters and Jamal [59] has shown a higher reliance for government and nonprofit organizations to adopt content from external sources, as they are usually less creative in generating their own content as compared to corporate and consumer-driven companies. This postulation is applicable to posts by health organizations which are usually either government or nonprofit-based. In contrast to the study by Hefler et al [29], Jiang and Beaudoin [30] likened an increase in engagement with original, unpublished content created entirely by the health care individual or team involved in posting the content. The preference for original content is due to the novelty of the posts as they have not been viewed by users previously. Despite the potential impact of both original and adopted content on user engagement, original content is preferred when a creative team workforce is available.

Regarding social media platforms, Facebook and Instagram were both favored due to their function as networking sites, which are ideal for sharing ideas. Other microblog-based platforms, such as Twitter were less preferred as only short updates were allowed with limitations in post characters [60,61]. As for the timing of social media posts, only Kite et al [14] found elements that affected user engagement metrics. According to their study, postings made on Fridays generated more user engagement than those made on Mondays, possibly due to users spending more time browsing social media toward the end of the workweek. This finding is consistent with previous studies [62,63]. Interestingly, Kite et al [14] found that users preferred to view social media postings during work hours (between 8 AM and 5 PM), which contradicts previous research suggesting that users are more active on social media during the night [63]. Kite et al [14] suggested that users may feel more comfortable browsing health-related information during work hours.

This systematic review is the first of its kind to focus on social media posts that were delivered by HCPs to identify the elements that influenced user engagement metrics. However, some limitations should be considered. First, there was a variability of study designs for included studies. Thus, a meta-analysis was not conducted. Nevertheless, quality assessment was conducted using critical appraisal tools according to the study design to evaluate the methodological quality of all included studies.

Second, the decision to only include studies that reported user engagement metrics as outcome measures, in the form of a direct action toward a post was made to facilitate objective deduction of measures and allowed comparisons to be made across published studies. We acknowledged that our exclusion of studies with outcome measures reliant on subjective assessments, such as psychometric scales might have resulted in overlooking certain findings. However, such subjective outcomes may vary according to individualized studies [64], and the data are prone to response bias [22].

Third, the information pertaining to the elements and the post creators for social media posts was based on the information provided in the papers. Some elements in the social media posts may not have been sufficiently explained, which may have caused limitations when elements were grouped into categories. For example, the element “informative post” categorized under communication using supportive or emotive elements might also have underlying elements of communication toward behavioral changes. The categorization of elements was refined through a collective discussion with all authors.

Fourth, social media posts from Facebook groups may only be accessed by the social media users who are subscribed to the groups. Despite the difference, studies that incorporated such posts were included in the review as the functionalities resemble those of Facebook pages, allowing users to interact directly with the posts.

The review focused on studies examining social media posts on the reduction of lifestyle risk factors that were created by HCPs. Identifying elements that influence user engagement metrics would allow HCPs to have a greater understanding of the post features that are potentially favored by users. Implementation of such elements into future social media posts would empower the delivery of public health messages by HCPs.

Further research could be proposed to help strengthen the interpretations of elements that influence user engagement. Findings from this review mostly came from countries with highly developed digital media infrastructure. We may want to conduct similar experimental studies in less developed countries, to examine if similar elements would affect user engagement metrics. Furthermore, we recommend conducting more studies in areas that are still underresearched, based on the findings from this review. These areas include the source of post content, the choice of social media platform, and the timing of posts.

The systematic review outlined the prospects of effective health promotion by HCPs on social media in future postings. This is made possible by incorporating elements that have a positive impact on user engagement metrics. Positive user engagement metrics serve as an indication of a favorable response from users to the posts made by HCPs. Communication techniques that either used supportive or emotive elements, or techniques that emphasized behavior changes were 2 of the most dominant categories of elements that could potentially maximize postlevel interactions. These communication elements should also be supported by paying attention to the appearance of each post. As social media continues to evolve, the elements in social media posts should be continuously evaluated, providing adjustments when required.