Table 2 General Characteristics of Eligible Studies
First Author and Year | Country of Authors | Study Setting | Article Type | Objective | Population and Sample Size | CAM Discussed/Used | Social Media Discussed/Used | Primary Outcomes | How Primary Outcomes Were Measured |
|---|---|---|---|---|---|---|---|---|---|
Al-Samaray et al. 2020 [44] | Iraq and Jordan | Iraq | Qualitative study | To evaluate the impact of social media on consumers' decisions to use dental products made primarily from herbal ingredients rather than well-known chemical formulas | n = 300 (n = 100 pharmacists, n = 100 dentists and n = 100 consumers) | Herbal dental products | Social media (not specified), but Facebook and Twitter mentioned in results | Consumers' interview answers | Face to face interviews |
Barnes et al. 2020 [45] | Australia | Australia | Case study | To describe how Facebook was used to recruit a targeted sample of expectant and breastfeeding mothers to participate in a national online survey | 1418 participants enrolled in the study, 810 completed the 20-min (70 question) survey | CAM | Recruitment post success and response rates | Reach, impressions, link clicks, post clicks, and click through rates were used to determine the success of the post, as well as the cost per day and cost per engagement with the post for boosted posts. Data from the Facebook ‘Insights' page was used to calculate response rates in a variety of ways | |
Chan et al. 2020 [46] | United States | International | Content analysis | To contrast and compare the nutrition advice given by registered dieticians (RDs) and non-RD bloggers on blogs | n = 400 blog posts | Nutrition | Blogs | Codes and themes from each blog post's coding; Readability of RD and non-RD blog posts using the Simple Measure of Gobbledygook (SMOG) | To minimise error, all posts in the sample were printed and given identification codes for each researcher to use in their analysis. After a lengthy analysis process, each post was subjected to a text-based content analysis using open coding to identify trends in the data and develop themes. A Simple Measure of Gobbledygook (SMOG) readability measure was used to assess the differences in readability between RD and non-RD blogs based on themes found in RD and non-RD blogs, respectively |
Gierth et al. 2020 [47] | Germany | Germany | Exploratory experimental study design | To look into the effects of science-critical user comments attacking Facebook posts that made scientific claims | There were 2 smaller studies in the larger study: 1. 144 participants (98 were female), median age was 22 years, 84% were students 2. 160 participants (114 of which were female), median age was 23 years, 80% were students | Homeopathy | Participants' attitudes toward and self-reported knowledge and involvement in the topics discussed in the Facebook posts; their agreement with the research claims, perceived study credibility, agreement with the user comment with one item each, and source trustworthiness judgement | Questionnaires with five 5-point Likert-type scale items for each topic to determine participants' attitudes toward the topics were used. Single items to assess self-reported knowledge and involvement in each subject were also used. The Muenster Epistemic Trustworthiness Inventory was used which assesses three aspects of a scientist's trustworthiness: expertise, benevolence, and integrity. It has a total of fourteen 7-point semantic differentials, six of which make up the expertise scale and four make up the benevolence and honesty scales, respectively | |
Kawchuk et al. 2020 [48] | Canada, Denmark and Switzerland | International (all of Twitter) | Content analysis | To detail how Twitter activity can be used to discuss misinformation claiming a link between spinal manipulation therapy and increased immunity | 1118 Tweets were generated from search one, 778 of which were not relevant, 187 were promoting SMT for immunity, 141 refuting and 12 neutral, 809 tweets didn't mention a professional, 7 mentioning a professional irrelevant to the study | Spinal manipulative therapy (SMT) | Tweet mentions (tweets with search terms) over time, tweet coding and sentiment (emotional expression), themes and word frequency, profession coding (professions mentioned), influencers (tweet authors with an engagement score greater than zero), and tweet author demographics | The primary search (Search #1) was built using Boolean syntax and consisted of three main components. Four evaluators scored this sample of 86 tweets independently to calibrate their use of the Twitter Tone Index (TTI). The TTI was then used by these same evaluators to independently assess each tweet resulting from Search #1. Tweets with less than three evaluators in agreement were discussed and a fifth evaluator broke any unresolved ties. Importantly, only whether tweets mentioned a profession could be coded; it was not possible to determine whether or how the author was associated with a specific profession | |
Merten et al. 2020 [49] | United States | International (Pinterest) | Content analysis | To see how cannabidiol (CBD) products were presented on Pinterest | 226 pins of the 1280 cannabidiol and CBD related pins were analyzed | Marijuana (CBD) | Pin coding and engagement | Using code categories from previous health-related Pinterest studies, a codebook for CBD products was developed, tested, and used for this research. A pilot set of 58 Pinterest pins was used to test the codebook. Researchers independently coded the pins before meeting to address coding inconsistencies and challenges. Engagement was measured using previous studies' methods, which included the amount of saves (formerly known as repins) and comments | |
Allem et al. 2019 [50] | United States | International (Twitter) | Content analysis | To quickly capture and describe the public's recent experiences with cannabis using publicly accessible data from Twitter | n = 60 861 tweets | Cannabidiol (CBD) | Topic classification of the tweets | Based on the presence of at least one topic-related pattern, each tweet was assigned to one or more topics. This was done by examining each tweet for the presence of a specified set of patterns representing a topic using a rule-based classification script written in Python | |
Aubrey et al. 2019 [51] | United States | United States | Experimental study | To see how self-objectification, appearance anxiety, and product preferences are affected by how health advice for adolescent girls is framed | 154 girls | Yoga | YouTube | Girls' self-stated thoughts on self-objectification, appearance anxiety, and preference for appearance-enhancing products | Self-objectification and having participants describe themselves using 10 words of a 40-word bank. Attempted to balance negative and positive descriptors in each category. Used the Trait Self-Objectification Questionnaire (described in the “Trait Self-Objectification” section, r = .25, p = .002). Study also looked at appearance anxiety and preference for appearance-enhancing products |
Cano-Orón 2019 [52] | Spain | Spain | Content analysis, qualitative discourse analysis and in-depth interviews | To investigate the discourse of Spanish sceptics on Twitter and understand the mobilisation against these therapies by the sceptical movement | 6252 tweets from the #StopPseudociencias campaign were identified and 912 of these were used as the basis of the content analysis | CAM | Analysis of tweets and discourse related to CT skepticism. Variables were created ad hoc using the casuistry of the campaign as a basis and looking into the standard of the other research | The different campaign hashtags’ tweets were collected with the data scraping tool data-miner.io. For the content analysis, the three variables tweets were analyzed on were (a) positioning (whether the tweet was sceptical, neutral or favourable to CTs, or whether it was irrelevant to the topic, (b) type of information sources contained in the messages and (c) use of mentions. Following this, a content analysis methodology was applied to the description that they put in their bios and the variables analyzed were (a) the account type (anonymous/corporate/personal), and, for those personal accounts, (b) the profession and (c) whether they were specialised in biomedicine. Manual content analysis and a qualitative discourse analysis of the tweets was also performed using Nvivo, in order to categorise the arguments raised by the users | |
Hasan et al. 2019 [53] | United Kingdom | Did not specify | Experimental, quantitative study | To evaluate the effectiveness of hypnotherapy on skype for IBS patients and draw a comparison of it with face-to-face treatment | n = 20 | Hypnotherapy | Skype | IBS Symptom Severity | A minimum 50-point reduction in the IBS Symptom Severity Score (IBS-SSS) |
Hu et al. 2019 [54] | International (China, United Kingdom, Australia, United States) | Did not specify (Online) | Content analysis/Comparative analysis | To look at the consistency of adverse events (AEs) and adverse drug reactions (ADRs) documented in the literature, as well as in monitoring and social media data | 660 posts (351 Baidu post bar themes and 309 Sina micro-blogs) | Chinese patent medicine | Baidu post bar and Sina micro-blog | Safety related information about Chinese patent medicine on Baidu post bar and Sina | The contents of the post bar and blogs were screened by two reviewers separately, and agreements were reached. The reporter abstracted information relevant to protection, including manifestations, severity, length, effects, combined interventions, and causal-relation judgments. All of the information gathered was categorised as AEs. The ADR words are based on the following guidelines: 1) Names of AE/ADRs found in various resources that were compatible with WHO-ART words were specifically used; 2) For records where the AE/ADR names did not fit WHO-ART, the names were coded according to the thorough definition of the AE/ADRs; 3) for records where the names and descriptions of AE/ADRs did not match, descriptions were used as the key proof of coding; 4) for records with ambiguous names and descriptions where coding was difficult, the records were labelled as "unable to code" The frequencies and relative frequencies of AEs, ADRs, and the systems-organs groups they affect were counted. Each individual symptom or abnormality was counted when more than one AEs or ADRs occurred in the same patient. When a single AE or ADR was graded into two or more WHO-ART system-organ groups, it was counted multiple times. The top five AEs, ADRs, and affected system-organ groups were then ranked |
Lacasse et al. 2019 [55] | Canada | Did not specify if they focused on posts of a specific country | Text, content, and image analysis | To investigate the themes and patterns that prominently emerge within the engagement (comments, conversations) from posts tagged #yoga on Instagram | Content: n = 35,000—Analysis of 100 unique authors and images from the unique data set | Yoga | Netlytic categories; coding of themes by 2 researchers independently | To identify the most popular words in the dataset, Netlytic and content category analysis was used to determine the themes within the discussion. After identifying the terms, the prevalence of the use of the most popular words in the dataset was determined and recorded using a table. A pie chart was created to represent the four content categories and the prevalence of the words relating to these categories | |
Lognos et al. 2019 [56] | France | The posts were from a French database. Geographical information on whether the authors were French or Francophonie was not obtained | Exploratory study | By exploring posts published in health forums and French-language social media groups, to see whether patients with breast cancer were interested in CAM | n = 264,249 posts | CAM | Inpatients and Breast Cancer social forums; Facebook | CAM-related terms used in posts by cancer patients on health forums and social media groups, and how they fit into the five categories of NPIs: visual, nutritional, psychological, physical, and other | We used a database collected from internet-based French-language forums and discussion groups of patients treated or followed for breast cancer to perform a retrospective frequency analysis of the terms used in NPIs. With the consent of the French non-profit breast cancer patient association, the 264,249 posts published in these forums and Facebook pages (without additional details for each post, such as the number of views, comments, shares, or likes) were gathered and anonymized. In the compiled database of 264,249 articles, all NPI words were searched. We performed two descriptive frequency analyses: (1) an examination of the occurrences of NPI groups and their synonyms, and (2) a subcategory examination of NPI words, synonyms, and related terms (e.g., ingredient, technique, method, and profession) |
Majmundar et al. 2019 [57] | United States | International (Instagram) | Content analysis | To identify the themes, user profiles, and references to different types of e-liquid solutions used with KandyPens from Instagram posts about KandyPens | n = 1775 Instagram posts from 546 unique accounts | Cannabis aromatherapy | Thematically coded categories and references to e-liquid solutions of the posted images and captions; Instagram user profile categories associated with each post | Investigators collaborated to familiarise themselves with the data, then created a coding frame and identified common categories. The posts' themes were identified. Each Instagram user profile associated with each post was assigned to one of several categories | |
Plachkinova et al. 2019 [58] | United States | Not specified | Content analysis/mixed methods | To respond to the question, “Can a social media forum be used to boost CAM communication?” and explain the development of an online community to link patients with CAM and western biomedicine providers | Semi-structured interviews and user feedback: 12 key stakeholders from the CAM, biomedicine, and patient populations; Content analysis: based on content of 11 active users of the tool; Usability analysis: 8 graduate students | CAM | Social media tool created by the researchers | Usability, key concept codes from the semi-structured interviews, google analytics data, content analysis codes | We performed usability tests with eight subjects to assess the mobile application's technological effectiveness. Following the interviews, the participants were sent a link to an online survey in which they were asked to rate the tool's usability. The System Usability Scale was used (SUS). Semi-structured interviews were conducted with a total of twelve participants. Three people were patients, six were CAM or Integrative Medicine (IM) practitioners, and three were MDs or Physician Assistants working in a typical medical environment who conducted interviews |
Rizvi et al. 2019 [59] | United States | International (Yahoo! Answers) | Content analysis | To use topic modelling to better understand consumer awareness needs for dietary supplements (DS) and to assess the accuracy of correctly identifying topics from social media | 16,095 unique alternative medicine questions and 438 unique DS Ingredients | Dietary supplements | Yahoo! Answers | 38 health-related categories and 12 higher level groups based on the Correlation Explanation topic modeling results | To uncover the different topics in which consumers are most interested, we used Correlation Explanation (CorEx), an unsupervised topic modelling tool. All 200 topics produced by CorEx had their keywords manually checked and assigned to 38 health-related categories, which corresponded to 12 higher-level classes |
Webb et al. 2018 [60] | United States | International (Instagram) | Content analysis | To explore the shared and unique approaches used to encourage fitness and health on Instagram | n = 300 | Yoga | Coded attributes of the images taken from Instagram. The distribution of the coded attributes tagged from photos with #curvyyoga | Systematically coded images for factors such as: sociodemographic (e.g., race/ethnicity, gender), body-as-process (e.g., physically active portrayals), and body-as-object (e.g., weight loss, body modification) attributes | |
Yin et al. 2018 [61] | United States | International (YouTube) | Content analysis | To look at the different types of user-generated content on YouTube about cupping therapy for pain management | n = 100 videos | Cupping therapy | YouTube | Content and sources of the videos | Coded the 100 most widely viewed cupping videos for content and sources. Used logic regression to obtain the associations |
Dai and Hao 2017 [62] | United States | All of Twitter (International) 78% of the tweets with an identifiable location were from the United States of America | Content analysis | To evaluate factors that could potentially impact the public's attitudes to PTSD related marijuana use on Twitter | n = 1,253,872 tweets | Marijuana (medicinal) | Text mining, the opinions represented by the tweets measured through human judgement and the Naïve Bayes model, self reported location and the state regulation and socio-economic information associated with that location | Text mining was used to identify PTSD tweets related to marijuana use. Using a combination of human judgement and the Naïve Bayes model, the opinions represented by the tweets were measured (whether they support or are neutral/against marijuana use for PTSD). Geographical, state regulation, and socio-economic factor data was also collected. To obtain the location information from a large number of tweets, the following was extracted: geographic locations based on the self-reported location, time zone and longitude/latitude information. When determining state regulation, states were classified based on the following 3 groups: 1) legalized marijuana use for both medical and recreational use 2) legalized medical but not recreational use 3) no laws legalizing marijuana. The 2014 American Community Survey (ACS) was used to find socio-economic information. Then, the Spearman correlation was used to assess how state regulation and socio-economic factors correlate with the attitudes present in marijuana-PTSD tweets | |
Marcon et al. 2017 [63] | Canada | International (YouTube) | Content analysis | To explore debates or disputes about chiropractic in the comments section of popular YouTube chiropractic videos | n = 4 videos | Chiropractic manipulation | YouTube | Coding categories and analysis of the comments | Categories were assigned to the various comment characteristics and comments were coded from each top-10 list (likes and replies) for each of the four videos. Then, for the top-10 discussions in each of the four videos, we used coding analysis to determine the types of discussions that were taking place and whether debates were taking place. This method was guided by a classification scheme for YouTube comment content review and a YouTube interaction guide |
Bitcon et al. 2016 [64] | Australia | International | Thematic analysis | To look for evidence of herbalism discussed in the blogosphere | n = 9 blogs | Herbalism | Blogs | Coded broad-themes and sub-themes | Over a four-week period, the post history of each of the final samples of blogs was collected. To identify bloggers' gender, age, location, level of education, and occupation, demographic data was gathered from their biographies and blog content. Next, a data analysis was carried out. The content of the blog was openly coded until the themes were exhausted. During the authors' regular meetings, implicit and explicit descriptions of stakeholder experience were extracted, and the coding was organized into broad themes and sub-themes |
Guo et al. 2016 [65] | Taiwan | Taiwan | Program evaluation | To measure the effectiveness of a mindful yoga program that is delivered on social media to women that are pregnant | 50 pregnant women | Mindful yoga | Comments received via telephone from study participants to determine usage of Facebook and/or DVD program frequency and preference | A phone call was made to validate the participant's use of Facebook and/or the DVD software, as well as their opinions on it | |
Krampe et al. 2016 [66] | United States | United States | Program evaluation | To evaluate Fuze and compare it with Skype as a platform for dance-based therapy with older adults and nursing students | The participants from the assisted living facilitiy (ALF) were 14 residents (10 completed evaluations, with a ratio of 60% female to 40% male), with an average age of 77 years (range, 59–97 years), and with a diversified racial population (white, 60%; African American, 40%) Traditional sophomore baccalaureate nursing students (n = 14) from a Midwest university enrolled in a Health Promotion Across the Lifespan course participated in the Fuze dance therapy Four students were on site at the ALF; the other 10 were with a nursing instructor at the school of nursing (SON) | Dance-based therapy | Fuze (Skype was mentioned in discussion) | Program evaluation survey | The following major areas were evaluated through participant completion of a post-intervention form: (1) feasibility, including visual and audio quality, (2) engagement between the older adults, and (3) nursing students’ overall satisfaction |
Marcon et al. 2016 [67] | Canada | International (Twitter) | Case study analysis | To see if there were any criticisms or debates about SMT's effectiveness and risks on Twitter | Sample (n = 1267) from corpus of a total of 20,695 tweets | Spinal manipulative therapy (SMT) | Text content of the tweets and number of mentions or hashtags in each critical or skeptical tweet | Each critical or cynical tweet was counted for the number of mentions and hashtags. The prevalence of sceptical or critical tweets was then calculated by compiling lists of the top ten retweets and liked tweets from each corpus. Retweets and likes were used to determine how much coverage a tweet had received and which information was most commonly disseminated using Twitter tools | |
Robertson 2016 [68] | United States | United States | Quantitative study (experimental, longitudinal) | To determine the impact of a music therapy blog on Newborn intensive care unit (NICU) music therapy resources awareness, and education | 12 participants: 5 neonatologists, 6 nurse practitioners, 1 physician assistant | Music therapy | Blog created by the researcher | (1) how informed the participants are about NICU music therapy therapy services; (2) how confident they are when ordering NICU music therapy services for patients; (3) how helpful they believe music therapy is for NICU patients; (4) how comfortable they are communicating about music therapy with other NICU staff; and (5) how useful they perceive the MT blog to be for obtaining information | The survey had five questions that measured the primary outcomes. All questions were rated on a 5-point Likert scale with descriptors such as "No Knowledge, Not Comfortable," "Moderate Knowledge, Moderately Confident," "Very Knowledgeable, Very Confident," and "Very Knowledgeable, Very Confident,". Using QualtricsTM, the researcher developed an online version of the questionnaire. Due to the primary care providers' time constraints, a 5-point Likert-type scale was used for a fast and effective response experience. The sensitivity index for this survey ranges from 5 to 25, with higher scores suggesting greater knowledge of NICU music therapy services. Before, during (after two weeks), and after the study finished, survey results were recorded |
Keim-Malpass et al. 2013 [69] | United States | United States | Observation/thematic analysis | To look at descriptions of complementary therapy used by young women diagnosed with cancer who kept an online cancer blog | n = 16 blogs | CAM | Blogs | Descriptions of CAM used by young women diagnosed with cancer between the ages of 20 and 39 who kept a cancer blog online | Analysis of textual data and reflective journal entries and field notes |
Krampe et al. 2013 [70] | United States | United States | Case study analysis | To describe how video call technology was used to introduce nursing students to a group of older adults during a dance-based therapy session | 120 students but 8 were the focus (2 men and 6 women), and 5–6 older adults | Dance based therapy | Skype | Feedback about the process from all parties involved | An analysis of phone conference debriefings |
Walden 2013 [71] | United States | Did not specify | In-dept interviews with bloggers and textual analysis of blogs | To determine the reasons for why naturopathic physicians utilize blogs | 18 interviewees (1 naturopathic medical student and 17 naturopathic physicians) and the blogs of 14 of them | Naturopathy/Naturopathic medicine | Blogs | The reasons for why naturopathic physicians utilize blogs | Data analysis, coding of interviews, blog analysis (check of blog characteristics and a deep textual analysis of blog themes) |
Gregory et al. 2012 [72] | United States | International (YouTube) | Descriptive Analysis | To see if music therapy documentary videos on the internet effectively and accurately represent the profession | n = 27 documentaries | Music therapy | YouTube | An examination of the rubric, which includes documentaries chosen by their creators, video-specific information, and therapy-specific information | Researcher-created rubric nearly identical to the rubric created for the earlier investigation of YouTube music therapy session videos |