We performed a rapid review to quickly capture the current evidence on the use of social media in interventions on diabetes. We had two research questions: (1) Is there evidence on the use of social media in interventions aimed at improving, maintaining, or promoting health among people affected with diabetes? and (2) What are the reported outcomes, for example, the effects on clinical parameters, effects on behavior, or other effects?

The rapid review method was chosen as it typically provides similar conclusions as systematic reviews, and it allows to quickly and efficiently access the current evidence on the topic [18-21]. In this rapid review, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [22] and the MeaSurement Tool to Assess systematic Reviews (AMSTAR) [23,24] guidelines. This review has been registered in PROSPERO (registration number: CRD42018088206).

To answer the research questions, we performed an electronic search on February 13, 2018. It covered published studies comprising the terms “Social media,” “Social networking,” “Facebook,” “Twitter,” or “YouTube” in combination with the term “Diabetes” included in the title or abstract and indexed in the following databases: PubMed (Medical Subject Heading terms and text word), Scopus, EMBASE, and the Cochrane Library. The search strategy was limited to studies published in English. The full search strategy is summarized in Multimedia Appendix 1.

Publications were included in the review if they (1) focused on diabetes or involved participants affected by diabetes; (2) described interventions aimed at improving, maintaining, or promoting health; (3) reported results from the intervention; and (4) used social media in the intervention. Both primary studies and reviews were considered to be of interest and were, therefore, included in this review. Papers that did not meet all four criteria were excluded from the review.

All references captured by the search engine were uploaded to EndNote X7 (Clarivate Analytics; Philadelphia, PA, USA). Duplicates were identified and removed. To assess the eligibility of the papers, two passes were done. In the first pass, all titles and abstracts were examined by one reviewer (EG). In the second pass, the full text of the studies selected on the first pass was extracted and carefully analyzed to confirm their eligibility. When it was unclear whether the studies should be included, they were discussed and agreed with a second reviewer (EÅ). The agreed upon studies were included in the quality assessment. A single reviewer (EG) extracted the data from the included studies. The following data were extracted: interventions (duration and participants), social media use (channels, use as main tool for the intervention or as supporting tool), and outcomes (effects on clinical parameters, on behavior, or other effects).

The quality of evidence and risk of bias of the studies included in this review were classified by one reviewer (EG). The quality of evidence of primary studies was assessed following the Grading of Recommendations Assessment, Development and Evaluation guidelines [25]. A second reviewer (RW) verified the assigned quality of a random sample of primary studies. The risk of systematic bias was assessed by drawing on the AMSTAR criteria [19,23,24].

A total of 1383 publications were identified, and after removing duplicates, 676 titles and abstracts were screened. The full search strategy and its results are summarized in Multimedia Appendix 1. The list of all potentially relevant studies that were read in full-text form but were excluded from the review can be found in Multimedia Appendix 2. Of these publications, 35 met the inclusion criteria [26-60]; of them, 32 were primary studies [26-32,34-38,40-59] and 3 were systematic reviews [33,39,60] (Figure 1).

Only 1 of the 35 included studies was considered to be of high quality [31]; 19 studies were considered to be of moderate quality: the 3 systematic reviews [33,39,60,61] (Multimedia Appendix 3), all 15 randomized controlled trials (RCTs) [26-30,32,34,35,42,44,45,48,54,55,58], and 1 nonrandomized intervention [40]. The remaining 15 nonrandomized studies were weighted as being of low or very low quality, and therefore, they were removed from the narrative synthesis (these studies are listed in Multimedia Appendix 4). Hence, 20 studies (1 of high quality and 19 of moderate quality) were included in this review. Multimedia Appendix 5 summarizes these 20 studies.

The PRISMA checklist of this study can be found in Multimedia Appendix 6.

A rapid review method was used to quickly capture the current evidence on the use of social media in interventions on diabetes. Following a search in 4 databases, only 20 publications considered of adequate quality were included in this review: 3 systematic reviews and 17 primary studies.

The research evidence shows that the most commonly reported outcome in intervention studies using social media is HbA_1c, followed by satisfaction with the intervention, HRQoL, and diabetes knowledge or empowerment. Most of the intervention studies using social media that evaluated HbA_1c values reported significant improvements (13/16 publications) [26-28,30,32-35,39,44,45,54,60]. Four of these publications, 2 systematic reviews and 2 RCTs, reported improvements only in intervention groups compared with that in control groups [27,33,39,44]. However, due to a heterogeneity in the methods that were used in the studies, including differences in the characteristics of participants, sample sizes, and study lengths, comparing them is difficult.

The 5 studies that measured satisfaction with the interventions where social media were used unanimously reported positive effects [26,28,30,39,60]. Two publications reported positive effects on HRQoL [33,44], and 2 others found improvements in knowledge [42,45], while 3 publications did not report any significant differences in HRQoL [27,31,48] or any improvements in knowledge [27,31].

Health authorities have recommended educating diabetic patients and their families with the aim of improving self-management, promoting a positive behavior change, and reducing the risk of complications [1-7]. Although the use of social media has not been linked to clear improvements in one meta-analysis focusing on patients with noncommunicable diseases [14], there are several other meta-analyses where some favorable effects have been found among people affected with chronic diseases [10-13,39]. The findings of the present review suggest that the use of social media in interventions for diabetes in many cases has been beneficial, and we did not find any studies that suggested worsened outcomes with this type of intervention. Studies using social media in their interventions have mostly showed superior results linked to the use of social media. Only one of the publications included in this review did not report any benefit on clinical or behavioral outcomes [30]. In this case, the researchers used social media in a 1-year intervention, and they used their own social media channel [31].

It is interesting to note that more than half of the studies used social media as a tool or resource to enhance the main intervention, and in these cases, the interventions resulted in improvements in HbA_1c values. Participants in these studies who were allocated to receive education either through Facebook chat or Skype as reinforcement of the main intervention had significant decreases in HbA_1c values [26,28,30,32,34,35,45]. Furthermore, compared with patients in the control group, improvements in HbA_1c levels were found among patients receiving health education through closed groups on Facebook and were found in one study using its own social media network [27,44]. In only 2 studies where social media was used as a supporting tool no differences in HbA_1c values were found; these 2 studies used their own social media channel [31,48].On the other hand, only 1 of the 3 studies that used social media as a main channel for delivering the intervention and measured HbA_1c values [29,42,54] reported significant improvements. This study used Viber communication for delivering doctor or peer support [54].

These findings suggest that using social media as a supporting tool for the main intervention is beneficial for improving health outcomes in T1D patients. Furthermore, it seems that the clinical improvement is most likely to happen when the chosen social media is one of the most popular networking sites.

Our review has mainly identified studies conducted with young T1D populations; however, our conclusions are comparable to those reported in a meta-analysis published in 2014 that predominantly analyzed studies involving T2D patients [39]. Therefore, we think that public health institutions, clinicians, and other stakeholders should consider the use of social media in their interventions targeting people affected with diabetes. However, special attention should be paid to the risk of misinformation or harmful health material that can coexist when carrying out interventions in open social media channels as it could lead to undesirable or unexpected effects [62-67].

Most of the included studies focused only on young people affected with T1D, probably because it is believed that these media are typically used by young people. Certainly, since the origin of social media, younger people have been the most frequent users of these channels. However, the presence of older generations on social media has increased in recent years, and about 80% of North Americans and Norwegians in their 40s report being social media users [8,68]. Hence, including older populations through social media would also make sense. This could be an especially valuable way of targeting people affected with T2D, a disease that is mostly diagnosed in adulthood and whose prevalence has dramatically increased in the last few decades [1]. Intervention studies using social media seem to improve health outcomes in T1D patients, and they could be beneficial for people with T2D as well. However, more research, using social media, on diabetes types is needed to answer this question.

In this review, we identified many abstracts presented at conferences, but there were fewer full papers reporting methods and results in detail. Knowing further details of the method used and the interventions could help identify the mechanisms or behavior techniques that work better for improving patient outcomes. So far, it seems that studies that use social media as supporting tool and where the social media is used for delivering health education report better outcomes. However, there are not enough studies where social media was used as the main channel for delivering the intervention. In future research, one should consider using different social media channels as main sources for delivering the intervention.

In research projects, it is more common to use restricted-access social media (ie, Facebook closed groups, Facebook chat, Skype, etc), which allows the researcher to have a better control of the environment and the contents and also protect the patients’ privacy. However, the use of open social media channels offers the possibility of a large-scale impact. Providing high-quality contents on diabetes through the most commonly used open social media channels and interacting with the social media users could potentially help people with diabetes. By having access to free-of-charge quality information, they could improve their knowledge, an important prerequisite for improving self-management and health behaviors. Further research should explore how to best use open social media channels for health promotion interventions in diabetes.

Our results and conclusions might be susceptible to bias as a consequence of streamlining the systematic review process. There might be a selection bias (failure to search in additional potentially relevant databases, only 1 reviewer selecting the studies) and a publication bias (we only searched in 4 databases; we did not search for gray literature; and our search was limited to the English language). Eight of the included studies conducted in Macedonia could be based on the same study, although we treated the reported results independently, as they provided different sample sizes, different age ranges of the included participants, and different intervention periods. Because many of the included publications were abstracts presented at conferences and because we did not have access to complete data, a quantitative synthesis was not possible.

There is little evidence on the use of social media in interventions aimed at people affected with diabetes. However, after weighing the existing evidence, it seems that the use of these channels is predominantly linked to beneficial patient outcomes. Public health institutions, clinicians, and other stakeholders who aim at improving diabetes patient education should consider the use of social media in their interventions.