Background
Tardive dyskinesia (TD) is a persistent and potentially disabling movement disorder associated with prolonged exposure to dopamine receptor blocking agents such as antipsychotics [
1]. TD is characterized by abnormal and involuntary movements in one or more body regions, including the head, face, neck, trunk, hips, limbs, and/or extremities. When second-generation antipsychotics (SGAs) were developed, it was hoped that these newer drugs would reduce the risk for TD. A meta-analysis of antipsychotic studies published from 2000 to 2015 estimated that the global mean prevalence of TD was lower with SGAs (20.7% vs 30.0% for first-generation antipsychotics [FGAs]), but the authors also noted that the introduction of SGAs did not mitigate the risk for TD as previously thought and/or expected [
2]. With antipsychotics being approved for additional non-psychotic psychiatric conditions, along with increased off-label use, the risk of TD continues to be a clinical concern [
3].
TD has been shown to negatively affect daily functioning and health-related quality of life (QoL) [
4,
5]. The burden of TD may be even more pronounced in individuals who are highly functional (e.g., able to work) and aware of their movements [
6]. Effective treatments are needed for all TD patients, but they may be particularly important to individuals who are more aware of their TD symptoms and/or bothered by those symptoms. In its updated clinical practice guidelines for schizophrenia, the American Psychiatric Association (APA) recommends that reversible vesicular monoamine transporter 2 (VMAT2) inhibitors (e.g., valbenazine or deutetrabenazine) be used to treat moderate or severe TD [
7]. For milder forms of TD, the APA states that VMAT2 inhibitors can be considered based on factors such as patient preference, TD-related impairment, and impact on psychosocial functioning.
For VMAT2 inhibitors currently approved by the FDA for TD, registration trials included clinician-rated assessments such as the Abnormal Involuntary Movement Scale and patient-reported outcomes (PROs) such as the Patient Global Impression of Change (PGIC) [
8‐
11]. However, these trials and scales were not designed to capture the emotional burden of TD on individual patients. Social media listening (SML) is a method that allows for the analysis of unsolicited patient insights that are not normally captured in clinical trials. Moreover, SML analyses can represent the experiences of individuals who may be unlikely to participate in a clinical trial or patient preference study.
SML has been used to understand patient experiences in other therapeutic areas [
12,
13], and final 2020 guidelines from the FDA lists social media as an appropriate method for gathering patient input [
14]. In addition, the International Society for Pharmacoeconomics and Outcomes Research notes that collecting data from social media sources can help bring attention to the issues that are most relevant to patients and their families [
15]. Given the need for more patient-oriented research in TD, this study was conducted to analyze the emotional content of online posts from patients with presumed TD.
Methods
Study design and data sources
A comprehensive search of social media content (e.g., Twitter, blogs, online forums) was conducted using an analytics program that uses artificial intelligence and natural language processing to “understand” the sentiment of online posts and group those sentiments into themes (NetBase™; Santa Clara, California;
www.netbase.com). English-language content from the United States posted online from 01-Mar-2017 to 07-Nov-2019 was searched using the following predefined terms: “tardive dyskinesia” or spelling variations (e.g. “tardarive diskensia) (Supplementary Table S
1). Posts that were overtly political or included specific medications (brand or generic name) were excluded (Supplementary Table S
1).
All posts were collected from publicly available sites that were not password-protected. Social media platforms that do not provide public access to posts (e.g., Facebook) were excluded from the search. To protect the privacy of “users” (i.e., individuals who posted social media content), all personal information was anonymized prior to any manual review of the content. Verbatim quotes from posts are not presented to further protect the privacy of users. An ethics committee for the study sponsor, Neurocrine Biosciences, Inc. (San Diego, California), approved the study design. As a retrospective analysis based on posts from publicly available social media sources, no approval from an Institutional Review Board was required or obtained. However, the study was conducted in accordance with International Society for Pharmacoeconomics and Outcomes Research guidelines for studies pertaining to health economics and outcomes research [
15].
Selection and characterization of posts
Search results were manually reviewed by MF, EF, and LL for relevance. A unique identification number was assigned to each post included in this analysis. Multiple posts from individual users were allowed. Based on available information, whether explicitly stated or inferred from content, posts were analyzed for gender (per reviewer judgement based on screen names/pictures when available) and type of user (patient or caregiver, per reviewer judgment based on content of posts). Sources for the posts were documented (e.g., Twitter). Web addresses of posts were provided when available; screenshots were provided when links were inaccessible.
Sentiment and theme analyses
The NetBase™ analytics program was used for the following procedures: 1) develop a list of words/phrases within selected posts that described users’ experience with TD; 2) tabulate the number of times that these terms were mentioned; 3) identify the sentiment of the terms as “negative”, “positive”, or “neutral”; and 4) group the sentiments into larger themes. The analytics program was able to recognize linguistic complexities such as tone (e.g., sarcasm), cultural references, slang words, and emojis. The terms, sentiments, and themes generated by the analytics program were manually reviewed for quality and validity.
Statistical analysis
Quantitative data (e.g., demographics, user type, social media source, tabulation of terms, sentiment analysis) were aggregated and analyzed descriptively.
Discussion
SML is an analytical method that can provide insight into how patients describe their TD symptoms and their feelings about having this disorder. PROs such as the PGIC have been used in clinical trials of approved TD medications, including valbenazine and deutetrabenazine [
8‐
10,
16‐
18], and in RE-KINECT, a real-world study of possible TD in which patients were administered the EuroQoL 5-Dimension 5-Level questionnaire, the Sheehan Disability Scale, and a questionnaire about their involuntary movements and the impact of these movements on daily activities [
4]. Because the emotions and sentiments of patients (and sometimes caregivers) on social media platforms are unsolicited and unfiltered, SML methods differ from these PROs and offer a more comprehensive context for understanding how patients experience TD and the unmet needs in this therapeutic space. As indicated in the recent FDA guidance on patient engagement for drug development, social media can allow access to difficult-to-reach populations, provide accurate and automatic capture of data, and possibly result in greater self-disclosure by patients [
14].
SML is typically used in market research, but with a rigorous approach using health economics and outcomes research (HEOR) methodologies to assess the value proposition for an intervention, it can also guide patient-focused drug development and patient-centered treatment strategies [
14,
15]. For example, an SML study of chronic obstructive pulmonary disease found that cough, mucus production, and shortness of breath were the symptoms of greatest concern for patients in terms of disease management [
13]. In an SML study of dry eye, patients described the impact on work and other quality-of-life issues in addition to expressing frustration with suboptimal management of their symptoms [
12]. By evaluating patient perspectives based on their own words, without any prompting from clinicians/investigators or limitations imposed by clinical trial protocols, these studies were able to gather fresh and unbiased patient perspectives that could complement findings based on more traditional research methods. For clinicians, familiarization with the terms and language used by patients on social media platforms may help guide discussions about symptoms, burden, and treatment options.
In this study, which is the first to apply SML methods to TD, almost all posts were from patients, with a few from caregivers. All posts originated in the United States, with approximately one-third from Twitter. Some assumptions about sex, race, and age group could be made based on photos or content of online posts, but no methods were available to verify demographic or socioeconomic data. However, given the various factors which contribute to the digital divide (e.g., age/generational status, gender, race, income, education) [
19,
20], it should be noted that some key populations are probably not represented in this analysis (e.g., homeless patients, elderly patients, patients with intellectual or development disabilities). In addition, patients who are unaware of their TD are unlikely to be discussing their symptoms online and would therefore not be included in this type of study.
The results presented in this study, including how patients describe their TD symptoms and their feelings about those symptoms, are consistent with the types of data that the FDA has indicated as being appropriate for social media sources [
14]. Given that TD can be a highly distressing and potentially disabling disorder [
1,
21], it was perhaps surprising that 36% of online posts were assessed as neutral (33%) or positive (3%), although discerning the underlying reasons for neutrality or positivity was beyond the scope of this study. It is less surprising that many of the online posts analyzed in this study included negative terms and that 2 of the 3 major themes were “anger” and “insecurity”. Aspects within those themes included expressions of frustration, feelings of being unaccepted by society, and fear of being judged by others. Healthcare providers may help to offset some of these negative emotions by informing patients about the availability of approved TD treatments (e.g., valbenazine and deutetrabenazine) and by providing resources such as the “Talk About TD” website [
22] and information for local support groups. Moreover, engaging with patients through social media may foster discussions about TD that can be both educational and supportive. Most importantly, simple follow-up questions about mental well-being during regular visits, coupled with referrals for psychosocial counseling as needed, would help clinicians gain better insights into the experiences of their patients and allow patients to feel that their concerns are being addressed.
Limitations
This study had several limitations worth noting. First, a clinical diagnosis of TD in the users/patients could not be confirmed as TD was self-reported in online posts. Second, patients with TD are a heterogeneous population, but the posts in this report may not reflect the experience of all patients with TD. Additionally, the sample size was small, and as discussed earlier, certain patient populations may have been unable or unwilling to discuss their disease in online forums. Third, online posts often lack detail which may be the result of platform restrictions (e.g., character limits in Twitter). Finally, the perspective of caregivers was under-represented in this study and more research is needed to characterize the burden of TD on caregivers. More research is also needed to further understand the reasons why patients (and caregivers) choose to express their sentiments online and how healthcare providers can address these sentiments.
Acknowledgments
The authors would like to thank the entire team at Greater Than One (New York, NY) for their help in completing this project, as well as Prescott Medical Communications Group (Chicago, IL) for writing and editorial assistance. The work of both organizations was supported by Neurocrine Biosciences, Inc.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.