Longitudinal analysis of twitter trends for sinusitis

In 2018, the FDA encouraged the use of social media to shed light on patients’ perspectives regarding symptoms and impacts of disease.1 Researchers have used this form of data analysis to study how patients use social media when searching for information on new therapies, improving quality of life (QOL), finding patients with similar diseases.2-7 At present, there is a dearth of data on social media activity in the field of rhinology. Sinusitis is a common condition that has a significant impact on patients’ QOL.8 In this study, we aimed to use a large database of tweets to describe the qualitative and quantitative features of sinusitis-related tweets and compare them between the general public and healthcare professionals. Symplur Signals9 is an independent social media analytics company. From this database, 1.1 million tweets about “sinusitis” from March 1, 2012 to May 31, 2020 were included in this analysis. Data were limited to English posts, and retweets were excluded from the primary analysis. Tweets were categorized into those posted by “General Public” accounts (GPA) and “Healthcare” accounts (HA). We performed the following analyses: User Analysis, Tweet Metrics Summary, Tweet Activity, Words, Hashtags, Retweets, Trending Terms, Links. A hashtag is a word or phrase preceded by the symbol “#” that categorizes the accompanying text. Table 1 demonstrates an overview of all tweet characteristics. After cleaning the data, the total number of sinusitis tweets was 314,637, with 276,487 (87.9%) of those being posted by GPA. Of the GPA tweets, 11.7% were replies, 31.5% included mentions of other users, 5.7% had media attached, and 25.7% included a link. In contrast, HA tweets had a larger proportion of links (71.2%) and media (18.1%) and a lower proportion of mentions (18.9%) and replies (7.1%). GPA had a higher total number of users than HA (181,449 vs. 8,636), but HA had a higher average number of tweets per user (4.42 vs. 1.52). The yearly tweet characteristics are presented in Table 2. Tweets with both media and links increased until 2017, but those with media plateaued into 2020, while tweets with links had a steep decline until 2019 (82.8%−49.7%). For GPA, the highest tweet activity in most years was from January to March, followed by a dip during the summer and a subsequent rise from August to December. For HA, there is a less apparent seasonal tweet activity pattern. For GPA, the most common categories of links were homeopathic medical treatments (32 links) with 289 shares, sinusitis informational articles (10 links) with 99 shares, sites that sell antibiotics or other pharmaceuticals (four links) with 50 shares, healthcare provider offices (five links) with 40 shares, and otolaryngology journals or conferences (six links) with 30 shares. For HA tweets, the most common categories of links were homeopathic medical treatments (nine links) with 848 shares, healthcare provider offices (nine links) with 622 shares, medical devices (nine links) with 567 shares, sinusitis informational/teaching articles (19 links) with 185 shares, and otolaryngology journals or conferences (one link) with 18 shares. In this study, we used a database of 1.1 million tweets containing the keyword “sinusitis” to analyze tweets posted by GPA and HA. In contrast to GPA, the number of HA users was significantly lower (8636), but these accounts had much higher tweet activity per user (4.42). GPA had more words and hashtags related to the personal experience of illness than they did about treatment. This may explain the seasonal trend of GPA tweet activity to be highest in winter, during peak times of sinusitis symptoms. “Healthcare” accounts used more words and hashtags related to treatment strategies and other healthcare-specific terms. Unlike GPA, HA did not show seasonal changes in tweet activity, but rather it remained steady throughout the year—a logical finding given HA tweet about treatment strategies or devices, whereas GPA generally tweet about sinusitis when they experience the symptoms. GPA seasonal variation suggests that at certain times of the year, the activity of HA users may be more efficacious at providing patient education, particularly to counter misinformation spread among GPA users during peak sinusitis season. Notably, the quantity of hashtags used by HA was nearly as many as GPA, even though the number of HA tweets was significantly less, which may be related to HA users applying hashtags to garner exposure. Another observation from the trending terms of GPA tweets showed a higher number of layman's terms for common conditions associated with sinusitis such as “ear infection” and “hay fever,” whereas HA used medical terminology such as “otitis media” and “allergic rhinitis.”10 The percentage of positive tweets within each year was often much higher for HA than GPA. Patients are more likely to speak about negative experiences with symptoms and QOL, whereas HA generally speak about treatment and symptom improvement. In terms of procedures, GPA tweets had an overall positive sentiment due to the high success rate and included negative terms regarding symptoms or unfavorable outcomes. In GPA, links about homeopathy were the most shared, and while some of the suggestions for symptomatic relief held medical evidence, others made false claims. Links for sinusitis articles, otolaryngology journals, and conferences had the lowest number of shares. This propensity of GPA to favor sharing articles not backed by medical evidence is concerning for the efforts of physicians who aim to disseminate information to their patients. Lydia Weykamp: . William Swift: . Andrew Strumpf: None.