The Science of Diabetes Self-Management and Care2023, Vol. 49(2) 101 –111© The Author(s) 2023
Article reuse guidelines:
sagepub.com/journals-permissionsDOI: 10.1177/26350106231158828journals.sagepub.com/home/tde
AbstractPurpose: The purpose of this study is to use text-mining methods to examine the dominant sources of online information and content about continuous glucose monitors (CGMs). Because the internet is the most popular source for health information, it is important to understand what is being said about CGMs in online sources of information.
Methods: A text miner, algorithmic-driven statistical program was used to identify the main sources of online information and topics on CGMs. Content was limited to English and was posted from August 1, 2020, to August 4, 2022. Using Brandwatch software, 17 940 messages were identified. After cleaning, there were 10 677 messages in final analyses conducted using SAS Text Miner V.12.1 software.
Results: The analysis identified 20 topics that formed 7 themes. Results show that most online information comes from news sources and focuses on the general benefits of CGM use. Beneficial aspects ranged from improvements in self-management behaviors, cost, and glucose levels. None of the themes mentioned changes to practice, research, or policies related to CGM.
Conclusions: To improve diffusion of information and innovations going forward, novel ways of information sharing should be explored, such as diabetes specialist, provider, and researcher engagement in social media and digital storytelling.
Availability of continuous glucose monitors (CGMs) for people living with type 2 diabetes (T2DM) has become more ubiquitous. Originally used predominantly in people with type 1 diabetes (T1DM), recent changes to CGM policies and guidelines have made CGMs more widely available to people with other types of diabetes. In the annually released Standards of Medical Care in Diabetes—2020, recommendations stated that time in range provided by CGMs could serve as an acceptable endpoint for glucose assessment in clinical trials instead of the traditionally used standard of A1C.1 This impacted the landscape of clinical diabetes research with studies expanding to include and/or focus solely on new target populations, such as people with T2DM2,3 and pregnant people with gestational diabetes.4-6 With the increase in evidence, recommendations for CGM use were broadened to include people with diabetes on multiple daily injections and continuous subcutaneous insulin infusions not defined by diabetes type.7 Following this, in 2021, the Centers for Medicare and Medicaid Services (CMS) approved reimbursement for CGMs for CMS recipients with diabetes that have multiple daily injections of insulin with no limitations based on diabetes type.8
Despite these recently expanded standards and policies based on strong evidence of CGMs as a powerful clinical tool, adoption of new tools and technologies takes time. One relevant theory that seeks to explain how, why, and the rate which a new idea or technology spread is the diffusion of innovations (DOI) theory. According to this theory, the following 5 elements influence the spread of innovations: (1) the innovation itself, (2) adopters, (3) communication channels, (4) time, and (5) a social system.9 Research has explored what aspects related to the innovation itself adopters assess and identified factors such as ease of use, relative advantage, trialability, and observed effects.10-12 Given that CGMs are not newly released, there is sufficient information available regarding the innovation itself that has shown positive results.13-16 However, it remains unclear what information about CGMs is readily available and being presented to new groups of potential adopters, such as people with T2DM.
Currently, the internet is the most popular source for health information, with studies suggesting that more than half of US adults use the internet as their primary source for health information.17 Unfortunately, this does not mean that identifying accurate and high-quality health information is easy. Studies from the COVID-19 pandemic illustrated how quickly misinformation can spread on the internet given that people are unlikely to fact check what they see18 and rely on social media for their information.19-22 To understand the state of diffusion of CGM devices to newly eligible groups of people with diabetes, it is important to assess the third element that influences the spread of innovations—communication channels. Thus, the purpose of this study is to use text mining to examine the dominant sources of public information and content on the internet about CGMs since the 2020 changes to care guidelines and policies. To achieve this, the following specific aims were addressed: (1) to identify the frequency and types of public sources of online information dominating the discourse on CGMs and (2) to assess the content of public sources of online information and social media conversations surrounding CGMs for prevalent themes.
This retrospective study of publicly available online information used text-analytic methods to identify topics and extract meaning contained in unstructured textual data. It followed the same analytic process used in multiple textmining studies examining public reactions to health and other issues across social media.23-26
A broad search query was created using the phrase “continuous glucose monitor” and the abbreviation “CGM.” All mentions were collected using Brandwatch software because it captures a wide array of public online information sources that range from formal (eg, professional news/journalism) to informal (eg, blogs, reviews) to social media (eg, forums, Reddit, Tumblr, Twitter, and YouTube). This software includes the following categories of content sources: blogs, forums, news, Reddit, reviews, Tumblr, Twitter, and YouTube. For the category of news, Brandwatch states that the majority of popular news sites along with numerous smaller local news sites are included in its search query with examples such as CNN, ABC, The Washington Post, Google News, BBC, Forbes, USA Today, Fox News, ESPN, etc.27 All mentions were captured during the 2-year period of August 1, 2020, through August 4, 2022. The search terms returned 17 940 sources of information and accompanying content. A Python script written to remove retweets and robotic messages was used and resulted in a total of 10 677 messages that were included in the final text analysis.28 These messages were analyzed separately using text-mining software, SAS Text Miner 15.1.29
Analysis of the textual content of the information sources was conducted using SAS Text Miner 12.1 that is an algorithmic-driven statistical software used to understand textual data. This software provides the ability to parse and extract information from text, filter and store the information, and assemble messages into related topics for researchers to inspect to identify insights from the unstructured data.29 The initial step in Text Miner is to extract, clean, and create a dictionary of words using natural language processing. Next, each message was divided into individual words, which were listed in a frequency matrix. Words that contributed little to the understanding of the topic, such as auxiliary verbs and conjunctions, were excluded from analysis. Following, a Text Filter node was used to exclude words that appeared in less than 4 messages to reduce noise. Words that are not essential (eg, “of,” “and,” “but”) were removed. A researcher then reviewed the output and removed unrecognizable characters and/or strings of letters. A single author followed a systematic process to maintain objectivity.
With inclusion criteria set, the Text Topic node was used to combine terms into 10-, 15-, and 20-topic solution groups. This clustering divided the document collection into groups based on the presence of similar themes using expectation maximization clustering. SAS employs a series of algorithms that select words that are used together frequently to build the topic groups. Two authors evaluated the results by comparing the different topic solutions and selecting the optimal solution after review of the output. After visually examining each of the topics, the 20-topic solution most clearly illustrated the main themes. Authors then grouped each of the topics together to form the larger themes. Lastly, the team reviewed all primary data (eg, original messages/content) within the 20 topics to confirm, refine, and finalize thematic results. Table 1 reports the 20 original topic clusters and shows which topics contribute to each main theme.
Table 2 reports the frequency of each online information source included in the final sample of messages (N = 10 677), and Figure 1 illustrates these results. Of the 9 types, most online information came from formal news sources (n = 7460, 70%). All social media sites combined accounted for 25% of the information sources, with most of the information coming from Twitter (n = 1425, 13%) and Reddit (n = 743, 7%) and the least from Tumblr (n = 422, 4%) and YouTube (n = 41, 0%).
Textual analysis of the mutually exclusive 20-topic solution yielded 7 discrete themes. Each of the themes and their respective sources of online information are described in further detail in the following. Table 3 provides exemplar primary data by theme.
Theme 1: information on specific CGM devices and supplies. This content provided information about CGM devices and supplies that varied depending on the source. News reported general information, and articles rarely mentioned specific CGM systems or companies. These articles primarily included information that is similar across various brands, such as the basics of proper use and how CGMs work. Content in blogs and Twitter focused on sharing brand-specific information, such as unique features and benefits of a particular CGM device or system. Frequently highlighted benefits included aspects of the sensor (eg, size, longevity) and the size of the transmitter. The cost-effectiveness of specific brands was frequently mentioned, with many making price comparisons to other well-known brands. Only one of the sources mentioned insurance and/or discount programs, and none mentioned changes to reimbursement policies.
Theme 2: using CGM with other devices and/or education to increase benefit. This theme identified how use of a CGM paired with other devices or education has the potential to exponentially increase health benefits. Information sources were diverse, with most of the content coming from news and social media sites. No differences in content by source were apparent. Insulin pumps and artificial pancreases were the most common diabetes devices mentioned. Physical activity trackers (PATs) were frequently mentioned in conjunction with CGMs, with content describing how use of PATs and CGMs together can allow for better management of blood glucose (BG) values with exercise. One of the topics contributing to this theme focused on specific brands of CGMs and PATs that were designed together so users only need 1 app to access both streams of data. Finally, content from 1 topic described how use of CGM is embedded into the curriculum of a specific digital diabetes educational program.
Theme 3: financial information on CGM companies. Financial information on the companies that manufacture CGM devices was reported primarily in the news but also by blogs. News focused on stock market share prices and earnings, fluctuations in current values, and future potential values of multiple brands and companies. All financial content was positive, with reports of higher than expected increases that exceeded analyst estimates. Citing the financial reporting from news sources, content from blogs identified good CGM companies for investment and outlined future potential innovations that might lead to increases in profits for particular companies.
Theme 4: CGM can identify and forecast hypoglycemia, giving users peace. Coming from a mix of sources, this content focused on how CGM can assist with forecasting hypoglycemia. News articles included the following content related to hypoglycemia (defined as glucose <70 mg/dL1 ): the definition and a description of the prevalence, incidence, risk factors, and impact of complications (eg, significant association with higher mortality).30,31 News also reported how CGMs facilitate identification of hypoglycemia and reported the ability for CGMs to predict future BG values using machine learning algorithms. Blog and Tumblr content was generated by CGM users or their family members and consisted of personal stories or questions about CGM use. Content frequently articulated the peace of mind that forecasts provide and engaged multiple people on 1 post. The phenomenon of hypoglycemia unawareness, which is the lack of symptoms during episodes of hypoglycemia, was touched on by all information sources.
Theme 5: short- and long-term eating choices are impacted by knowing blood glucose values. Comprised of blog posts, this theme described how continuous data allow users to make knowledgeable decisions regarding nutrition. For short-term eating choices, content described the new ability to look at BG values in the moment and use it to direct eating choices, with examples ranging from the type of food to the portion sizes. Many posts emphasized the utility of this for eating that is atypical (eg, celebratory) or not prepared by them (eg, at a restaurant). For long-term food choices, content focused on how seeing unique BG patterns after consuming particular foods gives people specific information that enables them to make food choices that they are confident will work for them.
Theme 6: CGM gives daily blood glucose measurement and trend data. This theme was comprised of news articles that described multiple brands of CGMs and their ability to provide vast amounts of real-time and trend data. Content stated that having this data would allow for better diabetes management and could help those with prediabetes to prevent progression of the condition. None of the included sources gave any concrete or actionable examples of ways that access to more BG data from a CGM could be used to improve management.
Theme 7: the long-term cost effectiveness of CGM. Content focused on 2 research articles that each examined different types of CGM effectiveness—the long-term cost effectiveness and the accuracy and utilization rates of 2 different CGMs. The study examining cost-effectiveness was published by Roze et al32 and found that for people with T1DM, CGMs are cost-effective relative to fingerstick checking over a lifetime horizon, with an incremental cost-effectiveness ratio of 16 931 (Canadian dollars) per quality-adjusted life year gained. For the study comparing 2 CGMs for accuracy and effectiveness, authors found that they had nearly equivalent, high accuracy, and realworld utilization rates.33 Both studies were described in news articles sharing the results generally and to a lay audience.
This study examined publicly available online information sources to understand what content is dominating the discourse about CGMs since recent changes to practice and policy. Overall, results show that most information comes from the news and focuses on the general benefits of use. The focus on benefits of use maps all study findings to Element 1 from the DOI theory (eg, the innovation itself) by speaking to characteristics of CGM/innovation such as ease of use, relative advantage, trialability, and observed effects. These results align with prior literature that demonstrates improvements to glycemic outcomes34,35 and behavior modification36-39 for people with diabetes using CGM. However, findings also show that most content on CGMs does not capture innovations or changes, given that none of the themes included the updates to care guidelines, changes to use in research, or the expansion of CMS CGM reimbursement policies. None of the content reported the significant improvements in outcomes seen for people with T2DM who use CGM but are on nonintensive insulin regimes40,41 or are not on insulin therapies at all.42 By not highlighting new advances and better tailoring content to specific groups particularly to potential adopters (represented by Element 2 of the DOI theory), online information remains general and potentially reduces broader interest and ultimately uptake. Future research should further investigate the efficacy of using CGM more broadly, assess how to generate more online information that shares innovations, and identify effective ways to better target online health information to specific groups.
Because the internet is the most popular source for health information,17,43 finding ways to increase content on changes to care is critical. One novel approach to impacting the type, quantity, and quality of online health information is to engage clinicians, diabetes care and education specialists, and researchers in creating and sharing health information content on social media. An article by Sultan et al44 outlines how clinicians can generate their own content and highlights the agility of this approach to report breaking news like nationwide drug shortages or changes to guidelines. However, prior research has shown that using social media as a means of communication is not popular among health care providers,45 with estimates of providers willing to create content ranging from 24% to 34%.46 Results also indicated that physicians felt uncertainty related to the boundaries or best practices for health communication on social media.47 To better understand the experiences of “early adopter” physicians using social media as a professional platform for health communication, a qualitative study by Campbell et al47 found that few providers were using platforms to their full potential because many did not engage with the more advanced, interactive features. The need for guidance has led to many professional organizations, such as the American Medical Association and the Association of Diabetes Care and Education Specialists, to create policies and ethical codes for social media professionalism and for research to examine best practices of professional use.48-50
Training and empowering clinicians and diabetes care and education specialists to share information with the public on social media might be a good approach to fill this need, but there is a dearth of evidence that describes the types of information or attributes of topics that this is best suited for. The issues with misinformation related to the COVID-19 pandemic serve as a powerful example of the difficulties with ensuring accuracy of health information on the internet. Ample evidence has described the medical misinformation and conspiracies about COVID-19 that have been shared on social media platforms.51,52 Although research has identified key individual characteristics that correspond to beliefs in misinformation such as distrust of scientists,53 it is unclear whether there are any predictors, moderators, or mediators that could inform the selection of future health topics that are appropriate for social media information sharing.
Findings from this analysis show no evidence of misinformation on CGMs. Contrary to what has been seen on social media related to COVID-19, the results from this study illustrate how some social media platforms captured the benefits of CGMs in ways that news articles could not. These sites provided real people with a way to share powerful and nuanced stories that spoke to many of the emotional elements that are a part of living with or supporting a loved one with diabetes. Current research is investigating new ways to incorporate storytelling into digital spaces to share information.54 A recent systematic review by Rieger et al55 (2018) examined digital storytelling and found it to be effective at giving voice to people to share knowledge and evoke change. Furthermore, Briant et al56 found digital storytelling to be a highly effective communication strategy for public health. Future research should explore the application of digital storytelling methods and evaluate which health topics are appropriate for this mode of communication.
One health topic that would benefit from changes to current communication practices is health policy. Absent from this study’s findings were changes to CMS reimbursement for CGMs. Literature on traditional news coverage has demonstrated the importance of how news stories are framed and reported,57,58 with research illustrating how health journalism’s focus on a tight narrative often conveys an incomplete or inaccurate story.59,60 Studies have shown increased news coverage to impact public perceptions on health behaviors, like changes in harm perceptions related to e-cigarette use.61 Recent research is indicating that reporting of health policy has been framed to focus on politics and partisanship instead of policies or their potential impacts. For example, a study on news coverage of the Affordable Care Act (ACA) found that the most common sources of news provided little publichealth-relevant substance about the ACA even after its implementation. The major policy aspects included in the ACA, such as Medicaid expansion, were mentioned in less than 10% of new stories, and less than 4% cited research in their reporting.62 Further exacerbating these issues are news sources that research is identifying as including nonfactual information. A 2017 study by Bard62 examined coverage of the ACA from Fox News and found evidence of mostly nonfactual information, reinforcing prior work by Conway et al.63 This analysis found that almost three-quarters of the current information on CGMs is generated from news, making it critical that future research examine how the news covers health policy.
Although these findings are useful for understanding the main sources of online information on CGMs and the larger themes that represent this content, this study does have limitations. The data acquisition software, Brandwatch, that queried the publicly available online information is comprehensive and highly representative of the information on the internet, but it is not exhaustive. Furthermore, although all attempts to create a broad and comprehensive search were made, our query could have missed key terms that limited content retrieval. Finally, results were limited by the lack of specificity in attribution related to news publications because no names or identifiers of the information sources were provided (eg, no publication titles were provided).
This study found the publicly available online information related to CGMs is generated mostly by formal news sources. Content focused on beneficial aspects of this technology, including improvements in self-management behaviors, cost, and glucose levels. None of the results identified changes to practice, research, or policy on CGMs. To improve diffusion of information and innovations going forward, novel ways of information sharing should be explored. Health care provider, diabetes care and education specialists, and researcher engagement in social media regarding CGM updates could improve the quality of online information and potentially impact the public’s understanding of health care policy.
The authors declare that there is no conflict of interest.
St David’s Center for Health Promotion and Disease Prevention Research in Underserved Populations Pilot Grant.
Elizabeth M. Heitkemper https://orcid.org/0000-0001-9537-5195
Julie Zuñiga https://orcid.org/0000-0003-3778-8148
Heather Cuevas https://orcid.org/0000-0003-4314-6686
From School of Nursing, The University of Texas at Austin, Austin, Texas (Dr Heitkemper, Dr Zuñiga, Dr. Kim, Dr Cuevas); Center for Health Communication, Moody College of Communication, The University of Texas at Austin, Austin, Texas (Dr Wilcox); and Stan Richards School of Advertising and Public Relations, Moody College of Communication, The University of Texas at Austin, Austin, Texas (Dr Wilcox).
Corresponding Author:Elizabeth M. Heitkemper, PhD, RN, School of Nursing, University of Texas at Austin, 1710 Red River, Austin, Texas 78712.Email: e.heit@utexas.edu
