Perspectives surrounding robotic total hip arthroplasty: a cross-sectional analysis using natural language processing

Abstract

Background:

Robotic technology has been used in total hip arthroplasty (THA) for several years. Despite the advances in this field, perspectives surrounding robotic THA are not fully understood. This study aimed to characterize the landscape of robotic THA on social media.

Methods:

The Twitter application programming interface was queried from inception to October 2022 for keywords related to THA and robotics. Posts and accounts were extracted and classified using thematic labels. Sentiment analysis was performed on the extracted tweets.

Results:

After removal of duplicate posts and illegitimate accounts, a total of 742 tweets from 741 accounts were retrieved. Most posts pertained to raising awareness about robotic THA (n = 340, 45.8%), advertisements for THA robots (n = 204, 27.5%), and personal experiences (n = 138, 18.6%). Research was discussed in 7.0% (n = 52) of posts. Most accounts belonged to patients or caregivers (n = 177, 23.9%), followed by medical centres (n = 175, 23.6%), news outlets (n = 158, 21.3%), and physicians or researchers (n = 101, 13.6%). Most posts discussing personal experience were positive (n = 70, 50.7%) or neutral (n = 39, 28.2%). Presence of media (β = 3.3, 95% confidence interval [CI] 1.5 to 5.1) and tagging (β = 2.1, 95% CI 0.3 to 2.8) positively affected user engagement, whereas the presence of a link decreased tweet engagement count by 2.8 (95% CI −5.4 to −0.2).

Conclusion:

Topics about robotic THA were discussed in a positive tone on Twitter (rebranded to X in 2023). Posts about raising awareness and advertisements for robotic THA were most prevalent, while research-related posts were limited. Orthopedic surgeons can leverage social media to better understand patient perspectives and glean insight from the robotic surgery industry.

Contexte:

La robotique est utilisée depuis plusieurs années pour l’arthroplastie totale de la hanche (ATH). Malgré les progrès réalisés dans ce domaine, on connaît encore mal l’opinion qu’elle suscite. La présente étude visait à dresser un tableau de l’ATH robotisée telle qu’on l’aborde sur les réseaux sociaux.

Méthodes:

L’interface de programmation d’application de Twitter a été analysée depuis sa création et jusqu’à octobre 2022 à partir de mots clés reliés à l’ATH et à la robotique. Les messages et les comptes ont été extraits et classés par thèmes et une analyse des sentiments a été effectuée à partir de ces messages.

Résultats:

Après élimination des doublons et des faux comptes, nous avons retenu en tout 742 messages provenant de 741 comptes. La plupart visaient à faire connaître l’ATH assistée par robot (n = 340, 45,8 %), faisaient la publicité pour des robots d’ATH (n = 204, 27,5 %) et relataient des expériences personnelles (n = 138, 18,6 %). La recherche était abordée dans 7,0 % (n = 52) des messages. Les comptes étaient principalement ceux de patientes ou patients ou de membres du personnel soignant (n = 177, 23,9 %), suivis des centres médicaux (n = 175, 23,6 %), des médias d’information (n = 158, 21,3 %) et du milieu médical ou scientifique (n = 101, 13,6 %). La majorité des messages concernant une expérience personnelle étaient positifs (n = 70, 50,7 %) ou neutres (n = 39, 28,2 %). La présence de médias d’information (β = 3,3, intervalle de confiance [IC] de 95 % 1,5 à 5,1) et le marquage (β = 2,1, IC de 95 % 0,3 à 2,8) ont exercé une influence positive sur l’engagement des utilisateurs, tandis que la présence d’un lien a eu l’effet contraire en réduisant l’engagement de 2,8 (95 % CI −5,4 à −0,2).

Conclusion:

Les thèmes entourant l’ATH assistée par robot ont fait l’objet de discussions favorables sur Twitter (devenu X en 2023). Les messages visant à en faire connaître l’existence ou à en faire la publicité étaient les plus nombreux, tandis que les messages portant sur la recherche étaient limités. Les orthopédistes peuvent mettre les médias sociaux à profit pour mieux comprendre le point de vue des malades et prendre le pouls de l’industrie.

Total hip arthroplasty (THA) was introduced in the 1970s to surgically treat hip osteoarthritis.¹ The field of THA has evolved to improve clinical outcomes and optimize surgical technique.² In fact, THA is considered by many to be one of the most consistently effective procedures.³ In the last decade, robotic THA has emerged as a promising alternative to conventional THA, with improved component positioning and fewer intraoperative complications.^4,5 Robotic systems aim to restore joint biomechanics by optimizing the positioning of the prosthetic components and helping restore femoral offset and leg lengths. The ROBODOC (THINK surgical) was the first robotic surgical assistant for THA.^6,7 More recent systems have been developed, such as the Rio MAKO robot (Stryker) and the ROSA Hip System (Zimmer Biomet), with improved capabilities such as 3-dimensional scan reconstructions and haptic feedback.⁸

Social media platforms such as Twitter (rebranded to X in 2023), Facebook, and Instagram have become valuable tools for sharing information and updates about innovative approaches in surgery.⁹ By using social media, surgeons can connect with other medical professionals, share their experiences, and learn from colleagues at different institutions. Furthermore, social media has become a vehicle for engaging with patients and the public.¹⁰

The implications of social media in joint arthroplasty have been studied to assess the tone of patients and their dialogue with physicians.^11,12 Gleaning status reports of recovery from patients via social media as they relate to orthopedic surgery is of particular importance considering the rehabilitation that follows orthopedic procedures.^13–15 Additionally, social media can be a valuable tool for surgeons to network and collaborate with other medical professionals, sharing best practices, and discussing new techniques. Despite the advantages, few physicians regularly use social media.¹⁶ However, in recent years, Twitter (rebranded to X in 2023) has become increasingly popular among physicians.¹⁷

The long-term clinical and functional outcomes comparing conventional THA to robotic THA are similar; thus, physicians can leverage social media to help disseminate information, allowing patients to make more informed decisions about the procedures.^18–21 Therefore, we conducted a cross-sectional analysis of social media posts on Twitter related to robotic THA. The purpose of our observational study was to contribute to the social media literature by describing online themes in robotic THA and characterizing the respective accounts; determining indicators of engagement; and classifying the tone of posts as positive, negative, or neutral.

Methods

Search strategy

The Twitter API for Academic Research database was searched using the following key words independently and in combination: “total hip replacement,” “total hip arthroplasty,” “hip replacement,” “hip arthroplasty,” and “robot.” Retweet exclusion and English language filters were applied. The search spanned from inception to October 2022. The abbreviated terms “THA” and “THR” were omitted in the search because they introduced too many irrelevant posts.

Data extraction

A customized query of the Twitter API was performed to automatically extract tweet metadata. This included information about the author of the post: username, number of followers, total number of tweets by the author, author description (biography), and author location. Data pertaining to the post included the textual content of the post, date of creation, number of retweets, number of replies, number of likes, quote count, presence of link (yes/no), presence of media (photo or video, yes/no), tagging other users (yes/no), and presence of hashtags (yes/no). Retweets and duplicate tweets were excluded. Accounts with fewer than 15 followers or containing fewer than 10 tweets as well as bots and duplicates were removed.

A qualitative categorization of the posts was performed in line with a recent social media study.²² Accounts were labelled as follows based on the author description (biography): “business,” “foundation,” “journal,” “MD or researcher,” “medical centre,” “news,” “patient or caregiver,” “support group,” and “other.” The tweets were classified according to the content of the post as follows: “advertising or raising awareness,” “fundraising,” “personal experience,” and “research.” Two authors (blinded for review) independently categorized the accounts and posts, with conflicts resolved through discussion. Irrelevant tweets not pertaining to robotic THA were manually queried and removed.

Sentiment analysis

The Textblob natural language processing (NLP) library was used to analyze the content of each tweet.²³ The sentiment analysis algorithm is pre-trained on large data sets. The tweets were preprocessed by removing punctuation and special characters and were subsequently tokenized. This lexicon-based NLP technique utilizes a library that contains predefined polarity and subjectivity scores. The polarity score ranges from −1 to 1, with −1 indicating a negative sentiment, 0 indicating a neutral sentiment, and 1 indicating a positive sentiment. Words such as “disgusting” or “awful” have a polarity score closer to −1, whereas words like “excellent” or “amazing” have a polarity score near 1. The subjectivity score ranges from 0 to 1, with 0 representing objective text and 1 representing subjective text. An analysis score was calculated by averaging the polarity and subjectivity scores from the individual words in the post. Analysis scores of less than 0, 0, and greater than 0 represented an overall negative, neutral, and positive tone, respectively.

Statistical analysis

Social media metrics were summarized using descriptive statistics (mean, standard deviation [SD]), unless otherwise stated. Descriptive statistics for all social media metrics, including follower count, tweet count, tweet likes, and retweets, were calculated. Our data did not follow a normal distribution, which is consistent with results from previous social media studies.^22,25,26 As a result, we used non-parametric 2-tailed tests to evaluate variations in user numbers across different categories. Beta (β) represents the fold change in engagement for a given variable. A significance level of 0.05 was established, and we used R 4.1.3 (R Foundation for Statistical Computing) for all statistical analyses.

Ethical considerations

According to the Canadian Tri-Council Policy Statement on Ethical Conduct for Research Involving Humans, institutional research board approval was not required for this study as all data were publicly available. ²⁴ Confidentiality was maintained by omitting all usernames. The tweets analyzed in this study were collected from a publicly accessible source, the Twitter API, without any interaction with social media users, and were cross-sectional and archival in nature.

Results

Characteristics of accounts and posts

A total of 742 unique posts (from 741 accounts) were included in the analysis. Most posts fell into the category of advertising or raising awareness (n = 544, 73.3%) for a service. These posts were usually advertising a device or service in orthopedics or informing the public on robotic THA; for example, “#TotalKneeReplacement is done when all 3 parts of the knee are significantly damaged. Robot-Assisted Knee replacement can be an excellent option for this often found in mid to late-stage #osteoarthritis […]” and “[…] launched next-gen Trident II Acetabular System, new hip replacement with additive manufacturing, allows for large femoral head size options, poly thickness; potentially offers greater range of motion, joint stability, lower risk of dislocation; using Mako robot.” The second most common theme was personal experience (n = 138, 18.6%). An example of a post from this category is, “I had a hip replacement end of May and started walking really well in 5 to 6 weeks. It was a Mako robotics. I’m grateful for the pain relief and to be able to walk without a walker or a cane! I think I had a great surgeon... or was it the robot?” The categories least encountered were research (n = 52, 7.0%) and fundraising (n = 8, 1.1%). The theme count is summarized in Figure 1.

Fig. 1.

Percentage of the included posts, by theme.

The median number of followers was 1108 (interquartile range [IQR] 312–3662), and the median number of tweets from each author was 10 016 (IQR 2710–32 740). Of the 741 unique accounts, 177 (23.9%) were from a patient or caregiver, 175 (23.6%) from a medical centre, 158 (21.3%) from a news outlet, 101 (13.6%) from a physician/researcher, 88 (11.9%) from a business, 22 (3.0%) from a support group, 8 (1.1%) from a journal, 7 (0.9%) from a foundation, and 5 (0.7%) were classified as “other.”

Indicators of engagement

The mean total engagement (likes, quote tweets, retweets) per tweet was 3.93 (SD 12.8). Of the total number of posts, 258 (34.8%) posts contained media in the form of a photo or video, 220 (29.6%) contained a link to an external reference, and 267 (36.0%) contained a hashtag, such as “#hipreplacement,” “#robot,” “#tech,” and “#automation.” On multivariable linear regression analysis, the presence of media increased engagement 3.3-fold (95% CI 1.5 to 5.1). Figure 2 shows examples of media in the advertising or raising awareness and research categories. Tagging other accounts increased the average engagement (β = 2.1, 95% CI 0.3 to 3.8). Moreover, posts containing a link decreased the average engagement (β = −2.8, 95% CI −5.4 to −0.2). The presence of a hashtag did not significantly affect user engagement (β = −2.8, 95% CI −1.8 to 1.5) (p = 0.846). A summary of the tweet characteristics can be found in Table 1, and a summary of the multivariable regression model can be found in Table 2.

Fig. 2.

Sample tweets containing media and hashtags in the (A and B) advertising or raising awareness category, and a sample tweet containing media, hashtags, and a link in the (C) research category. Usernames omitted for confidentiality.

Table 1.

Characteristics of included posts

Characteristic	Overall mean ± SD*
Replies	0.21 ± 0.73
Quote count	0.09 ± 0.41
Retweets	0.74 ± 2.67
Likes	2.88 ± 10.07
Engagement	3.93 ± 12.81
Presence of media, no. (%)	258 (34.8)
Presence of tagging, no. (%)	220 (29.6)
Presence of link(s), no. (%)	611 (82.3)
Presence of hashtag(s), no. (%)	267 (36.0)

SD = standard deviation.

^*Unless indicated otherwise.

Table 2.

Multivariable linear regression model evaluating post characteristics on engagement (retweets, quotes, and likes)

Characteristic	β	95% CI	p value
Followers	0	0	< 0.001
Business	−0.8	−21.4 to 19.7	0.937
Foundation	0	−21.9 to 21.9	0.999
Journal	0.5	−21.2 to 22.3	0.961
MD/research	4.4	−16.2 to 24.9	0.677
Medical centre	2.3	−18.3 to 22.8	0.828
News	1.4	−19.2 to 21.9	0.896
Other	9.9	−12.6 to 32.3	0.389
Patient/caregiver	−1.1	−21.7 to 19.5	0.917
Support group	−1.5	−22.5 to 19.4	0.885
Media	3.3	1.5 to 5.1	< 0.001
Tagging	2.1	0.3 to 3.8	0.019
Link	−2.8	−5.4 to −0.2	0.033
Hashtag	−0.2	−1.8 to 1.5	0.846

CI = confidence interval.

^*Represents the fold change in engagement for a given variable.

Sentiment analysis

Our sentiment analysis showed that 456 (61.5%) posts were positive, 198 (26.7%) were neutral, and 88 (11.9%) were negative. Within the personal experience category, 70 (50.7%) posts were positive, 39 (28.3%) were neutral, and 29 (21.0%) were negative (Figure 3). The authors of these posts were usually patients or caregivers expressing their thoughts pre- or postoperatively. An example of a post that was labelled as positive is, “[…] My surgery will be robot assisted. I checked out a video about the process, it’s actually very cool and state-of-the-art. I’m a bit nervous, but more about surgery itself than the hip replacement. Thanks for the encouragement.” An example of a post labelled as negative is, “[…] Today, I visited the surgeon who did my hip replacement. The scar on the side of my leg is about 10–12 inches long. I have another scar from the robots alignment tool. I considered the anterior approach but was worried it would be harder to go back in.”

Fig. 3.

Sentiment analysis of patient experience with robotic total hip arthroplasty.

Discussion

To our knowledge, this is the first study to characterize discussions of robotic THA on social media. We present several novel findings, including our observation that most posts were advertisements from industry or medical centres, raising awareness for robotic THA, or posts on personal experience from patients and caregivers; that the posts were mostly published by patients, caregivers, or medical centres; that the posts were mostly positive; and that the presence of media or tagging other users was associated with increased engagement.

Social media for information dissemination

The ubiquitous use of social media has dominated many spheres of society. Recently, there has been a significant increase in the utilization of social media by health care providers and facilities.^27,28 Social media is a vehicle to express personal opinions, share insights, learn from others, advertise services or products, and more. Orthopedic surgery has a large presence on social media, notably for patients documenting their rehabilitation.²⁹ Moreover, a survey of orthopedic surgery residents and prospective applicants showed that social media is being used for educational purposes. ³⁰ Orthopedic surgeons use social media to facilitate dialogue with large and diverse patient populations.³¹ They also share research findings and publicize conferences.

To increase reach on social media platforms, users and companies leverage various techniques like including media, tagging, hashtags, and links. The 2023 social media analytics industry was valued at US$4.8 billion and is expected to grow substantially in the coming years.³² Our engagement analysis pertaining to robotic THA showed that including media or a hashtag was associated with 3.3-fold and 2.1-fold higher engagement, respectively. This association is consistent with the findings of a previous study that correlated the presence of media with higher engagement when accounting for other variables.²² The presence of a link, however, was associated with decreased engagement. When users click a hyperlink, it can either decrease interest or redirect them to other pages and thus reduce the chances of them interacting with the post.

Thematic trends

Most posts assessed in our study were either advertisements or raising awareness for robotic THA. Private surgical practices have led medical centres to implement marketing strategies to attract patients. Our study showed that both private and public medical facilities were promoting their robotic services for THA. The general public should exercise caution when relying on information shared by physicians or industry on social media, as it may not always be supported by evidence-based medicine. Although the use of robotics may appeal to some individuals, it is essential for transparent communication and raising awareness through social media to build trust with patients.

Research was discussed in only 7.0% of posts. Discussing research on social media can bring attention to publications and ongoing works in the bioengineering or clinical outcomes fields. Patients can learn about their pathology and their course of treatment. There were relatively few accounts classified as support groups (3.0%). Support groups comfort patients by alleviating emotions before surgery and providing encouragement during postoperative rehabilitation.

Sentiment analysis

Sentiment analysis is a subfield of NLP that is being widely used in medicine to review patient satisfaction.^33–35 A sentiment analysis algorithm calculates a subjectivity and polarity score to classify textual data as positive, neutral, or negative. It aggregates the sum of these scores from individual words to provide an overall score for the given text. Sentiment analysis has been previously used to analyze online patient reviews of surgeons.^36–38 This NLP technique geared the surgeons’ focus toward enhancing their bedside manners and improving their management of pain.

In orthopedic surgery, these data can provide patient perspectives on new surgical techniques and novel technologies, such as robotic THA. Our study showed that there was positive sentiment in the posts. Within the personal experience category of patients or caregivers, the predominant sentiment was also positive. In fact, a previous study found that preoperative sentiments were associated with postoperative pain following orthopedic surgery.³⁹ Thus, addressing patient concerns gleaned from social media preoperatively has the potential to improve postoperative outcomes.

Strengths and limitations

Our study has several strengths. First, we queried the entire Twitter API over an approximately 15-year period. The automated data extraction allowed us to obtain information about the authors and the posts. We also removed bots and irrelevant posts during data preprocessing. Next, we leveraged NLP using sentiment analysis to glean insights from patients and industry leaders in robotic THA. Moreover, we analyzed Twitter metrics that are associated with higher engagement.

Our study also has several limitations, including the search strategy. We excluded the keywords “THA” and “THR” because they resulted in a substantial number of irrelevant posts with similar acronyms, causing the data to become noisy. Also, we searched only the Twitter platform. Collecting data from other social media databases would be of value. A limitation of the NLP algorithm is that it does not consider context, only individual words. This means that a context-based approach would classify sentiments in the posts more accurately.

Conclusion

We present, to our knowledge, the first study illustrating the use of social media in robotic THA. Most posts raised awareness. Few posts discussed research. The predominant sentiment in the tweets was positive. The presence of media and tagging other users increased engagement. Information shared on social media should not replace the expert knowledge of physicians and researchers. As technology advances, social media can serve as a supplement to both health care and industry practices.