Deep brain stimulation (DBS) is a procedure that modulates neural circuits associated with certain brain disorders and pathological states.1 Most of the evidence for DBS is derived from adult populations,2 in whom this treatment has been proven to be effective for the management of several disorders, namely Parkinson’s disease.3 The application of DBS to pediatric populations is relatively novel. Lessons derived from the adult literature may not generalize to children because of limitations surrounding technical and surgical considerations,4 differences in epidemiology for certain pathologies treated with DBS (i.e., dystonia),5 and the unknown effects of DBS on the developing brain.6 Furthermore, neuromodulation in children presents several unique ethical challenges, including the protection of the child’s best interest and consideration of the child’s relative experience with their illness.7
Over the past decade, there has been increasing focus on the use of DBS to treat several neurofunctional diseases in children, most notably dystonia. Today, DBS is an accepted treatment option for children and youth with medically refractory dystonia, especially those with TOR1A-related dystonia (DYT1 dystonia) or myoclonus-dystonia.8 Furthermore, several pathologies are considered investigational and potentially amenable to DBS, including Tourette syndrome,9 epilepsy,10 and self-injurious behavior.11 Overall, the consideration of DBS in children is expected to continue to rise in the upcoming years.
Despite the increasing use of DBS for pediatric neurofunctional diseases, little is known about patients’ and caregivers’ experiences with the intervention. Further, DBS, in general, is not well understood by the public.12 A systematic review found that many adult patients have significant hopes for DBS preoperatively and experience a sense of euphoria postoperatively; despite this, the review authors concluded that there is a need for further qualitative research to better understand patient and caretaker perspectives after DBS.12 One study by Austin and colleagues sought to characterize decision-making by parents regarding DBS for pediatric dystonia. Using structured interviews, the authors analyzed responses from 8 parents of children with dystonia who had undergone DBS.13 Otherwise, little is known about parent or caretaker perspectives on DBS in children and youth.
Social media has become an active part of many patients’ and caregivers’ lives.14 In 2020, over 190 million daily active users were reported on Twitter.15 Owing in part to a large patient presence, physicians have become interested in leveraging social media to gather insights on public opinion, knowledge transfer, and patient perspectives.16 Social media analysis offers several advantages, including its widespread use and the ability to extract real-world information from users.17 Further, social media can facilitate communication between caretakers and clinicians, offering a direct channel through which they can discuss care and ask questions. Several studies in neurosurgery have analyzed social media metrics for different pathologies, including hydrocephalus,18 aneurysms,19 and epilepsy.16 Our study aims to contribute to this literature by describing the landscape of social media use surrounding DBS in children and youth. These data will be used to identify common themes in communication and further understanding of perspectives on DBS in pediatrics.
Methods
Search Strategy
We performed a comprehensive search of the Twitter Application Programming Interface (API) database for academic research to identify any tweet pertaining to the use of DBS in children and youth. A “tweet” can be defined as an online post generated by a user and published on the Twitter social media platform. The following search terms were used separately and in combination: “deep brain stimulation”, “neuromodulation”, “pediatric”, “baby”, “son”, “daughter”, “kid”, “child”, “parent”, and “guardian.” The initialism “DBS” was not included, as most tweets were linked to topics unrelated to DBS. We queried the entire Twitter database from inception to a search date in March 2022.
Social Media Metrics and Data
We extracted data about the tweet and analyzed each event separately. For accounts, we removed duplicates, as well as accounts with fewer than 10 tweets, bots, and accounts with fewer than 15 followers. The term “bot” is used to define a type of botting software that controls a Twitter account via the Twitter API; this software generally works autonomously, without continuous human input. Data extracted for analysis included originating account location, number of followers, number of tweets, and year joined. We categorized each Twitter account based on its objective or purpose, public title, and user-generated Twitter account descriptions. We created categories to describe the accounts based on an initial screening of accounts and a previous publication.16 These categories included charity or foundation, business, medical journal, patient or caregiver, support group, medical center, news outlet, medical doctor or researcher, and other.
Data on individual tweets were also extracted and included tweet date, number of likes, retweets, quotes, and full text. Duplicate tweets were removed. Each extracted tweet was independently evaluated and verified by two authors (L.M.E., J.J.L.). Only tweets directly relevant to DBS use in children and youth, as judged by a manual query of all tweets by the study authors, were included for analysis. Each tweet was assigned a tweet theme category using modified thematic analysis, with axial and open coding methods described in a previous paper.16 Two investigators (J.J.L., F.N.) analyzed all tweets and generated codes for recurring themes, with differences in categorization resolved through discussion. The final thematic categories for tweets were personal experiences, raising awareness, discussing research or publications, advertising, fundraising, and other.
Statistical Analysis
Descriptive statistics (median, interquartile range) for all social media metrics, including follower count, tweet count, tweet likes, and retweets, were calculated. These data were not normally distributed, in line with previous studies on social media.16,20 Thus, nonparametric two-tailed tests were used to assess differences in the number of users between different categories. Alpha was set at 0.05. R-4.1.3 statistical software (R Foundation for Statistical Computing) was used for all statistical analyses.
Sentiment Analysis
We used a natural language processing (NLP) Python library called textblob21 to process the tweets for sentiment analysis. This NLP library used a lexicon-based approach to score text on a polarity and subjectivity scale. The algorithm employed a predefined dictionary of labeled words based on their positivity or negativity to analyze these data semantically. The tweet’s body was preprocessed, and sentences were given a polarity score, subjectivity score, and analysis label. The polarity score was between −1 and 1, with −1 representing negative sentiments and 1 representing positive sentiments. This polarity score is based on the presence of negative or positive words. For example, words like “excellent” or “best” have a polarity score of 1, whereas words like “disgusting” or “terrible” have a polarity score of −1. The subjectivity score was between 0 and 1, with 0 representing objective information and 1 representing personal information. An analysis score of < 0, 0, and > 0 represented an overall negative, neutral, and positive analysis, respectively. Tweets were scored by pooling the average semantic scores from individual words.
Ethical Considerations
All tweets included in this study were archival and cross-sectional and were freely obtained from a publicly accessible source (Twitter API) without interaction with social media users. All usernames were omitted. Thus, the present study, per the Canadian Tri-Council Policy Statement for Research, does not require institutional research board approval since all data are publicly available.22
Results
Account Analysis
The search strategy yielded 877 tweets from 816 unique accounts that met the study inclusion criteria. The median number of followers for the accounts was 1390 (IQR 464, 3920). The median number of tweets per account was 7833 (IQR 2197, 29848). Most users (61.8%) were from the United States, the United Kingdom, or Canada. Almost 22% of users did not list a country of origin. Most tweets were from patients or caregivers (263 [32.2%]), followed by researchers (155 [19.0%]), news media outlets (141 [17.3%]), and charities or foundations (80 [9.8%]). The remaining classified accounts included businesses (39 [4.8%]), medical journals (17 [2.1%]), medical centers (71 [8.7%]), support groups (13 [1.6%]), or other (37 [4.5%]).
Tweet Analysis
Overall, 877 tweets were extracted for analysis after duplicate tweets, tweets from bots, and irrelevant tweets were removed. The time trend for tweet publication is shown in Fig. 1. There were fewer than 20 tweets per year from 2009 to 2012. Tweets then peaked (15%) in 2018 and 2019, followed by 2021 (14%) and 2020 (11%). Median total engagement (likes, quote tweets, and retweets) per tweet was 1 (IQR 0, 5). Mean total engagement per tweet was 5.8 (SD 19.2). Of all included tweets, 22.5% included media (video or image), whereas 85% included a link. Further, 38.9% of tweets tagged another account or individual, and 42.8% included a hashtag (denoted by the # symbol; used to index keywords or topics on Twitter). An overview of tweet characteristics can be found in Table 1.
Number of tweets meeting inclusion criteria per calendar year, from 2008 to 2022.
Summary of characteristics among 877 tweets
Factor | Value |
---|---|
Median no. of followers (IQR) | 1390 (464, 3920) |
Median no. of tweets/account (IQR) | 7833 (2197, 29848) |
Thematic category, no. (%) | |
Advertising | 56 (6.4) |
Awareness | 146 (16.6) |
Experience | 240 (27.4) |
Fundraising | 3 (0.3) |
Other | 36 (4.1) |
Research | 396 (45.2) |
Median total engagement/tweet (IQR) | 1 (0, 5) |
Mean total engagement/tweet (SD) | 5.8 (19.2) |
Includes media, no. (%) | |
No | 680 (77.5) |
Yes | 197 (22.5) |
Includes tagging, no. (%) | |
No | 536 (61.1) |
Yes | 341 (38.9) |
Includes link, no. (%) | |
No | 128 (14.6) |
Yes | 749 (85.4) |
Includes hashtag, no. (%) | |
No | 502 (57.2) |
Yes | 375 (42.8) |
The most encountered tweet theme was research, seen in 45.2% of tweets. These tweets usually discussed novel therapeutic strategies or scientific findings from peer-reviewed journals. For example, one journal tweeted “#Online First: Frameless robot-assisted pallidal deep brain stimulation surgery in pediatric patients with movement disorders: precision and short-term clinical results.” The next most common theme was personal experiences, seen in 27.4% of cases. Many of these tweets focused on the direct operative or postoperative experiences of children after DBS, as described by parents or guardians, for example, “My [child’s age] had deep brain stimulation at [treating hospital] a year ago for dystonia. It changed his life and our world. We will never stop fighting. We are in this fight with you.” Spreading awareness was the third most common theme, seen in 16.6% of tweets. These tweets were generally targeted at creating optimism or uplifting other caretaker outlooks toward DBS or discussed and presented news articles related to DBS use in pediatrics, for example, “The incredible team of Canada’s first pediatric deep brain stimulation clinic @SickKidsNews.” Other themes found in our analysis included advertising, usually for conferences, journals, or medical centers, and fundraising. There was no discernable trend in the proportion of tweets categorized as personal experiences from 2009 to 2022.
On multivariable regression analysis, tweets from the personal experience category, tweets containing media, and tweets tagging other accounts were all associated with an increase in engagement metrics (retweets, quotes, and likes). The most significant factor, by far, was the presence of media embedded in the tweet, defined as either a video or an image. The presence of media (on average) increased the tweet engagement count by 10.5 (95% CI 7.3–13.6), when controlling for all other variables. The full multivariable model results can be found in Table 2.
Results of multiple linear regression model evaluating the effect of tweet characteristics on total engagement (retweets, quotes, and likes)
Characteristic | Beta | 95% CI | p Value |
---|---|---|---|
Thematic category | |||
Advertising | 0.1 | −6 to 6.1 | 0.982 |
Awareness | 0.8 | −5 to 6.6 | 0.788 |
Experience | 6.4 | 0.9 to 11.9 | 0.022 |
Fundraising | 21.4 | −0.1 to 42.9 | 0.052 |
Other | −1.1 | −9 to 6.8 | 0.788 |
Research | 1.2 | −4.1 to 6.4 | 0.663 |
No. of followers | 0 | 0 | 0.101 |
Includes media | 10.5 | 7.3 to 13.6 | <0.001 |
Includes tagging | 3.2 | 0.6 to 5.8 | 0.015 |
Includes link | 0 | −3.7 to 3.7 | 0.993 |
Includes hashtag | −1.2 | −3.7 to 1.4 | 0.383 |
Boldface type indicates statistical significance.
Tagging and the presence of media were also identified as predictors of higher engagement when analyzing a subgroup of tweets belonging to the research category. Specifically, the presence of media in tweets categorized as research increased the average engagement count by 4.9 (95% CI 3.2–6.7), whereas the presence of tagging increased the engagement count by 2.8 (95% CI 1.3–4.4) when controlling for other variables.
Overall, a total of 478 (54.5%) tweets were classified as positive, 308 (35.1%) as neutral, and 91 (10.4%) as negative per sentiment analysis (Fig. 2). Examples of tweets categorized as having a positive sentiment include the following: “[Location] boy can now walk thanks to deep brain stimulation here at the Children’s Medical Center! Way to go, [Name]!” or “Beautiful story: Teen undergoes deep brain stimulation at [Hospital name and link].” Examples of tweets categorized as negative include the following: “Watching his son go through deep brain stimulation is so sad [link]” and “Family hopes deep brain stimulation at [Hospital] can stop child’s seizures [link].”
Percentage of tweets categorized as negative, neutral, and positive according to sentiment analysis.
Discussion
This is the first study to provide a comprehensive overview of Twitter use related to DBS in children and youth. Several novel findings are presented: 1) most tweets are from patients, caretakers, or researchers; 2) most tweets discuss novel research findings, treatment strategies, or personal patient or caregiver experiences; 3) an overwhelming majority of tweets present positive sentiments; and 4) several tweet variables, including tagging, tweets about personal experiences, and especially tweets containing media, are associated with higher engagement metrics.
Social media is demonstrably important in present-day communication for patients. Studies have shown that patients turn to social media to receive support and further their understanding of their disease, both of which can help to improve their autonomy and sense of self-empowerment.23 Further, social media can offer direct lines of communication between clinicians and patients and among patients and caregivers themselves; thus, social media can facilitate real-world communication in ways that traditional media cannot. A 2016 systematic review of social media use in medicine revealed that social media often affects patients by improving their subjective and psychological well-being and their sense of self-management and control.23 However, patients have raised several concerns about social media use in medicine, including security, patient confidentiality, quality and accuracy of shared information, and author credibility.24
Tweet Thematics
The most frequently encountered tweet themes identified in this study were research and personal experiences, representing almost 73% of all tweets. Social media analysis has been done for multiple neurosurgical conditions, including traumatic brain injury (TBI), aneurysms, hydrocephalus, and rhizotomy for patients with cerebral palsy.25 Many of these studies use similar methods to categorize tweets but often identify themes different from those found in our study. For example, a study on tweets related to TBI identified subjective opinions, instances of injury, education, policy and rules, and medical as tweet theme categories.26 Another study on selective dorsal rhizotomy (SDR) identified emotional support, sharing information and advice, appreciation and success, advertising, challenges, inequities in treatment, and social media as a second opinion as the main themes.27 While not completely identical, the studies on TBI and SDR and the current study revealed that personal experiences account for a significant portion of social media content. In studies focusing on social media and TBI and SDR, these experiences were classified as “subjective opinions” and “emotional support/appreciation and success.”
The finding that many tweets on pediatric DBS discuss personal experiences demonstrates the potentially important role that social media has in certain patient’s or caretaker’s lives. These patients may turn to social media to vent frustrations, express gratitude to healthcare workers, or communicate the outcomes of DBS. Further, several of the tweets in the personal experience category with the highest engagement metrics came from physicians or news media outlets discussing DBS as a novel treatment strategy in children. For example, one tweet read, “[City] girl first child in [country] to undergo deep brain stimulation for epilepsy [news.link],” whereas another read, “A 2-year-old girl has become the youngest child in the world to have deep brain stimulation to treat severe dystonia. The operation was carried out by a team from @KingsCollegeNHS & @EvelinaLondon. Tune into @Channel4News tonight & find out more about [child’s name] and her operation.”
Sentiment Analysis
Sentiment analysis aims to analyze an individual’s sentiment and derive opinions about or feelings toward certain topics, issues, or events. Sentiment analysis can be considered a branch of machine learning; its use is common in social media studies given the suitability of social media data for data mining and because people often post their direct thoughts onto platforms.28 Sentiment analysis has been widely used in studies analyzing Twitter data for commercial, political,29 and health research–related purposes. Despite its prevalence, sentiment analysis has been infrequently used in studies analyzing social media in neurosurgery. Many studies have given examples of tweets with positive or negative connotations16,27 but have not provided formal analysis using recognized NLP methods. In the present study, we found that most tweets (54.5%) discussing DBS in children and youth were positive, with only 10% being negative. It is also possible that negative sentiments predominate in tweets made preoperatively, perhaps in anticipation of the surgical procedure. Future social media studies could categorize posts pre- and postoperatively and compare sentiment analysis between these groups.
Engagement Metrics
Understanding what content drives social media engagement metrics is of value to clinicians, researchers, and medical facilities. There is currently a thriving social media analytics industry, worth over $3.5 billion in 2020 and forecasted to reach $15.6 billion in 2025.30 While social media analytics are very popular in the commercial industry, they are used much less frequently in medical research. Our multivariable regression model identified tagging, a personal experience theme, and the presence of media (video or photograph) as predictors of higher Twitter engagement metrics, specifically likes, quotes, and retweets. The presence of media in a tweet was the strongest predictor, increasing tweet engagement by an average of 10.5, when accounting for other factors.
Study Strengths and Limitations
This study has several strengths. It is the first study to comprehensively review Twitter use related to DBS in children and youth. We used several methods to analyze data, including qualitative and quantitative methods and sentiment analysis. We captured all tweets from Twitter inception to the present date. Further, we offered an analysis of variables associated with increased engagement metrics, which is of value to clinicians, researchers, and medical centers.
Our study also suffers from several limitations. Most notably, we were unable to include data from several other social media platforms, including Facebook, YouTube, Snapchat, and Instagram. Many of these platforms are media-based communication networks, and a future study analyzing posts from these platforms may be of interest. It is also possible that certain patients or caregivers prefer to share personal experiences following DBS with family and friends, perhaps through Facebook, Instagram, or WhatsApp, rather than on public forums like Twitter. Thus, analysis of these platforms in future studies is warranted. Further, an analysis of posts (in our case, tweets) is usually the best way to derive insights from individuals discussing a specific topic. It is also likely that some tweets related to DBS use in pediatrics were missed if they did not include certain keywords used in the search strategy. For example, some parents may not explicitly mention DBS in their tweets and may refer to the intervention using simpler terms, such as “surgery” or “operation.” However, we wished to maximize the specificity of included tweets and therefore opted to only include posts with a direct mention of DBS.
Conclusions
This study is the first to analyze social media use related to DBS in children and youth. Most tweets are from patients or caregivers, researchers, and news media outlets and discuss research or patient personal experiences. The sentiment in tweets is mostly positive. Tweets discussing personal experiences, tweets tagging other individuals, and especially tweets containing pictures or videos are associated with higher engagement metrics.
Disclosures
The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.
Author Contributions
Conception and design: Ibrahim, Elkaim, NM Alotaibi. Acquisition of data: Elkaim, Niazi, Levett. Analysis and interpretation of data: Ibrahim, Elkaim. Drafting the article: Elkaim, Fallah. Critically revising the article: Bokhari, Breitbart, F Alotaibi, Alluhaybi. Reviewed submitted version of manuscript: Gorodetsky, F Alotaibi, Weil. Administrative/technical/material support: Levett. Study supervision: Ibrahim, Elkaim, NM Alotaibi.
References
- 1↑
Lozano AM, Lipsman N, Bergman H, et al. Deep brain stimulation: current challenges and future directions. Nat Rev Neurol. 2019;15(3):148–160.
- 2↑
Lozano AM, Lipsman N. Probing and regulating dysfunctional circuits using deep brain stimulation. Neuron. 2013;77(3):406–424.
- 3↑
Follett KA, Weaver FM, Stern M, et al. Pallidal versus subthalamic deep-brain stimulation for Parkinson’s disease. N Engl J Med. 2010;362(22):2077–2091.
- 4↑
de Vloo P, Ibrahim G, Stone S, Kalia S. Deep brain stimulation for metabolic movement disorders. In: Ebrahimi-Fakhari D, Pearl PL, eds. Movement Disorders and Inherited Metabolic Disorders: Recognition, Understanding, Improving Outcomes. Vol 1. Cambridge University Press; 2020.
- 5↑
Lin JP, Lumsden DE, Gimeno H, Kaminska M. The impact and prognosis for dystonia in childhood including dystonic cerebral palsy: a clinical and demographic tertiary cohort study. J Neurol Neurosurg Psychiatry. 2014;85(11):1239–1244.
- 6↑
Elkaim LM, De Vloo P, Kalia SK, Lozano AM, Ibrahim GM. Deep brain stimulation for childhood dystonia: current evidence and emerging practice. Expert Rev Neurother. 2018;18(10):773–784.
- 7↑
Davidson B, Elkaim LM, Lipsman N, Ibrahim GM. Editorial. An ethical framework for deep brain stimulation in children. Neurosurg Focus. 2018;45(3):E11.
- 8↑
Elkaim LM, Alotaibi NM, Sigal A, et al. Deep brain stimulation for pediatric dystonia: a meta-analysis with individual participant data. Dev Med Child Neurol. 2019;61(1):49–56.
- 9↑
Coulombe MA, Elkaim LM, Alotaibi NM, et al. Deep brain stimulation for Gilles de la Tourette syndrome in children and youth: a meta-analysis with individual participant data. J Neurosurg Pediatr. 2018;23(2):236–246.
- 10↑
Yan H, Toyota E, Anderson M, et al. A systematic review of deep brain stimulation for the treatment of drug-resistant epilepsy in childhood. J Neurosurg Pediatr. 2018;23(3):274–284.
- 11↑
Yan H, Elkaim L, Loh A, et al. Lesions causing self-injurious behavior engage putative networks modulated by deep brain stimulation. Brain Stimul. 2021;14(2):273–276.
- 12↑
Mathers J, Rick C, Jenkinson C, et al. Patients’ experiences of deep brain stimulation for Parkinson’s disease: a qualitative systematic review and synthesis. BMJ Open. 2016;6(6):e011525.
- 13↑
Austin A, Lin JP, Selway R, Ashkan K, Owen T. What parents think and feel about deep brain stimulation in paediatric secondary dystonia including cerebral palsy: a qualitative study of parental decision-making. Eur J Paediatr Neurol. 2017;21(1):185–192.
- 14↑
Househ M, Borycki E, Kushniruk A. Empowering patients through social media: the benefits and challenges. Health Informatics J. 2014;20(1):50–58.
- 16↑
Meng Y, Elkaim L, Wang J, et al. Social media in epilepsy: a quantitative and qualitative analysis. Epilepsy Behav. 2017;71(Pt A):79–84.
- 17↑
Harber P, Leroy G. Social media use for occupational lung disease. Curr Opin Allergy Clin Immunol. 2017;17(2):72–77.
- 18↑
Elkarim GA, Alotaibi NM, Samuel N, et al. Social media networking in pediatric hydrocephalus: a point-prevalence analysis of utilization. J Neurosurg Pediatr. 2017;20(2):119–124.
- 19↑
Alotaibi NM, Samuel N, Wang J, et al. The use of social media communications in brain aneurysms and subarachnoid hemorrhage: a mixed-method analysis. World Neurosurg. 2017;98:456–462.
- 20↑
Eysenbach G. Can tweets predict citations? Metrics of social impact based on Twitter and correlation with traditional metrics of scientific impact. J Med Internet Res. 2011;13(4):e123.
- 22↑
Canadian Institutes of Health Research. Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans—TCPS 2 (2018). Accessed August 18, 2022. https://ethics.gc.ca/eng/policy-politique_tcps2-eptc2_2018.html
- 23↑
Smailhodzic E, Hooijsma W, Boonstra A, Langley DJ. Social media use in healthcare: a systematic review of effects on patients and on their relationship with healthcare professionals. BMC Health Serv Res. 2016;16(1):442.
- 24↑
Hoffman-Goetz L, Donelle L, Thomson MD. Clinical guidelines about diabetes and the accuracy of peer information in an unmoderated online health forum for retired persons. Inform Health Soc Care. 2009;34(2):91–99.
- 25↑
Shlobin NA, Hoffman SC, Clark JR, Hopkins BS, Kesavabhotla K, Dahdaleh NS. Social media in neurosurgery: a systematic review. World Neurosurg. 2021;149:38–50.
- 26↑
Workewych AM, Ciuffetelli Muzzi M, Jing R, Zhang S, Topolovec-Vranic J, Cusimano MD. Twitter and traumatic brain injury: a content and sentiment analysis of tweets pertaining to sport-related brain injury. SAGE Open Med. 2017;5:2050312117720057.
- 27↑
Canty MJ, Breitbart S, Siegel L, et al. The role of social media in selective dorsal rhizotomy for children: information sharing and social support. Childs Nerv Syst. 2019;35(11):2179–2185.
- 28↑
Yue L, Chen W, Li X, Zuo W, Yin M. A survey of sentiment analysis in social media. Knowl Inf Syst. 2019;60(2):617–663.
- 29↑
Drus Z, Khalid H. Sentiment analysis in social media and its application: systematic literature review. Proced Comput Sci. 2019;161:707–714.
- 30↑
Social media analytics market with COVID-19 impact analysis, by component, analytics type, application (sales and marketing management, and competitive intelligence), deployment mode, organization size, vertical, and region—global forecast to 2026. Markets and Markets. Accessed August 18, 2022. https://www.marketsandmarkets.com/discountreports.asp?id=96768946