Social media data are a valuable source for rapidly exploring and understanding various real-world phenomena. Because many people share their health conditions on social media, a high volume of health-related social media data are available and the use of these data for large-scale quantitative analyses [1,2] and disease surveillance (referred to as “infoveillance”) is gaining much interest [3]. In particular, major advances have been made in the use of social media data to track the prevalence and spread of infectious diseases, including influenza and other conditions [4-13]. These studies have contributed to public health by demonstrating high correlations between fluctuations in the number of relevant social media posts and patients for a specific disease.

This study focuses on allergic rhinitis, also called hay fever, which is one of the most common allergic diseases worldwide [14]. In Japan, many people suffer from seasonal allergic rhinitis induced by Japanese cedar pollen between February and April each year. According to the results of the Japan National Epidemiological Studies in 1998, 2008, and 2019, the prevalence of allergic rhinitis in Japan has increased significantly over the past 20 years (49.2% increase in overall allergic rhinitis and 42.5% increase in cedar pollen–induced allergic rhinitis). It is now a national disease that affects the majority of the population [15].

Although allergic rhinitis is not a life-threatening disease, its main symptoms—sneezing, nasal discharge (watery), and nasal obstruction—significantly impair the quality of life (QOL) of patients, causing a major social problem [15-17]. According to the practical guideline for the management of allergic rhinitis in Japan 2020 [18], approximately 40% of people in Japan said they had allergic rhinitis, of which 30% had nasal allergies caused by pollen. In addition, since most self-medication for allergic rhinitis in Japan is for seasonal allergic rhinitis caused by cedar and cypress tree pollen, which are dispersed in the spring, the number of patients and the sales amount of over-the-counter (OTC) rhinitis drugs are greatly affected by the pollen conditions of the year or region. About three-fourths of OTC rhinitis drugs are oral medicines and one-fourth are nasal sprays; for oral medicines, more than 60% of annual sales are concentrated in the 3 main pollen dispersal months, from February to April. Retailers are required to assess the situation and make accurate predictions during the annual spring pollen season, as they are required to manage products without causing opportunity loss. Therefore, a real-time understanding of the prevalence of allergic rhinitis is important for providing necessary solutions, such as appropriate pharmaceutical distribution.

The surveillance methods currently available rely on the following three types of statistical data:

So far, correlations of hay fever–related tweets with pollen counts [12] and reported incidents of hay fever [13] have been investigated. Our previous study [20] analyzed data on pollen count, the number of hay fever–related tweets, and the number of patients during the seasonal allergic rhinitis period in Japan to explore their relationships. The results showed that increased pollen counts were associated with increased numbers of tweets and patients. In addition, increases in the number of tweets were also associated with increased numbers of patients.

This study explores the relationships between social media data related to allergic rhinitis and OTC allergic rhinitis drug sales as an outcome of consumer behavior. To the best of our knowledge, this is the first study to compare tweet trends with drug sales trends. Specifically, we investigate the correlation between the weekly number of tweets related to 3 main symptoms of allergic rhinitis—paroxysmal repetitive sneezing (sneezing), watery rhinorrhea (runny nose), and nasal obstruction (stuffy nose)—and the weekly sales volume of OTC allergic rhinitis medication (oral medicine and nasal spray).

Figure 1A shows the changes in the weekly number of tweets related to the 3 main symptoms of allergic rhinitis for the target period. Figure 1B shows the changes in weekly sales volume of OTC allergic rhinitis medication, including oral medication and nasal spray, from 2017 to 2020. The most common causative antigen of seasonal allergic rhinitis in Japan is cedar pollen, which disperses between February and April. In Figure 1B, there is a clear peak during this period each year.

Figure 2 shows the time series of the observed data and its decomposed components: trend, seasonality, and residual. Figure 3 shows heat maps of correlations of all pairs of time series data. Figure 3A shows correlations between observed time series data. Figures 3B-D show correlations between the trend, seasonality, and residual components, respectively.

Among pairs of the observed time series of tweets, the positive correlation between tweets concerning stuffy nose and tweets concerning runny nose was the highest (r=0.7349), as shown in Figure 3A. The time series of the trend components showed the highest positive correlation (r=0.9613; Figure 3B), and the time series of the seasonal components was also highly correlated (r=0.8483; Figure 3C). On the other hand, there were positive correlations with tweets about sneezing, but they were not high (r=0.3564 for tweets concerning runny nose and r=0.4382 for tweets concerning stuffy nose; Figure 3A), due to negative correlations between the time series of the trend components (r=–0.6005 for tweets concerning runny nose and r=–0.6197 for tweets concerning stuffy nose; Figure 3B). As for OTC medication sales volume, the highest positive correlation (r>0.95) was between oral medicine and nasal spray.

As for the correlations between tweets and OTC medication sales volume, the highest positive correlation (r=0.8432) was between the time series of tweets concerning stuffy nose and the time series of nose spray sales volume, as shown in Figure 3A. The correlation of the trend component was positive and not high (r=0.4088; Figure 3B). On the other hand, the correlation of the seasonal component was positive and high (r=0.9317; Figure 3C), indicating the tweet numbers of allergic rhinitis keywords and OTC drug sales have a seasonal pattern due to the seasonality of allergic rhinitis in Japan. Several Twitter-based surveillance studies [7-11] dealt with infectious diseases that demonstrate seasonality, such as influenza, and some of them effectively utilized such seasonal features.