We used the administrative claims database provided by JMDC Inc [29], which is one of the largest commercial databases in Japan [30]. Since 2005, it has held anonymized data for a population of 5,600,000, including individuals 65 years of age and younger as well as employees of large corporations and their dependents. The database consists of inpatient data, outpatient data, pharmacy data, diagnosis procedure combination data, and insured-population data. In addition, it contains anonymous data regarding patient characteristics; diagnostic codes according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10); surgery; procedures; drug prescriptions; medical institutions; costs; and prescription dates. Regarding drug data, it contains the World Health Organization Anatomical Therapeutic Chemical codes, generic drug names, product names, prescription dates, and dispensing dates. The database also includes data regarding insured people; therefore, we were able to not only obtain information about populations, but also determine if patients had changed medical institutions from hospitals to clinics.

We used pharmacy data for the period from October 2007 to September 2017, because television and newspaper coverage of moisturizer prescription increased in October 2017. Therefore, to assess the association between prescription moisturizer and internet searches, we decided to use the data from October 2007 until September 2017. Additionally, we identified heparinoid prescription—the World Health Organization Anatomical Therapeutic Chemical code: C05B01. We focused on heparinoid prescriptions only, as prescriptions of heparinoid with other drugs, such as antihistamines, might not have been affected by internet searches. We defined prescriptions from insurance identification numbers, prescription dates, and medical facilities. The study protocol was approved the ethics committee at Meiji Pharmaceutical University (approval number: 2959).

The internet search data was obtained from Google Trends, which is a product provided by Google. We included geography and time range in the Google Trends search terms. Search trend data were available for the period from 2004 to 36 hours before our search on April 23, 2018. Search results were shown on a scale ranging from 0 to 100, based on the ratio of searches on a given topic to searches on all topics. For example, imagine we were obtaining search results of the word “A” for every month for a year. If “A” was searched most frequently in April, the search results have a score of 100 for April, and the search frequency of other months are scored relative to the frequency in April. When we compare two words, “A” and “B,” the most frequently searched month is identified for both “A” and “B” [26].

We chose three search terms that we considered most closely related to heparinoid. We used the Japanese “” (Hirudoid) and “” (Besoften), which are the names of major products containing heparinoid in Japan. In addition, we entered “” (heparinoid), which is the generic drug name. We performed the search on April 23, 2018, and compared the three terms in Google Trends from October 1, 2007, to September 31, 2017, in Japan.

We performed a descriptive analysis to examine patient characteristics, monthly internet search data, and prescription trends. Patient age was stratified into three groups: 0-19 years, 20-59 years, and ≥60 years. This is because previous reports showed that moisturizer prescription increased in the 20-59-year-old age group; also, these individuals comprised a high proportion (>90%) of internet users during the past several years [6,31]. If patients’ ages crossed two age groups (ie, they moved into another age group during the course of the study), we included them in the younger group. We showed data regarding diagnoses related only to prescription heparinoid when more than 5% of patients were diagnosed with a disease of skin and subcutaneous tissue based on the ICD-10.

We adjusted the prescription time series according to the insured population’s data and showed prescriptions per 100,000 members of the population. We then selected two time series for cross-correlation. In correlating the time series, we accounted for trend and seasonality in time series data to avoid mistaken correlations. To examine the association between internet searches and moisturizer prescription, we used the Box-Jenkins approach to fit an autoregressive integrated moving average (ARIMA) model [32]. We used a seasonal ARIMA model, as all-time series data showed seasonality that increased from fall to winter and decreased from spring to summer annually. We performed the Kwiatkowski-Phillips-Schmidt-Shin test to render the series stationary and estimated the model parameters and diagnosed acceptability using the Akaike Information Criterion. We then checked the adequacy of the model, regardless of autocorrelation between residuals, using the Ljung-Box test. We performed cross-correlation analysis for each residual to assess the relationship between the two time series.

Cross-correlation analysis was performed for the following: (1) internet search trends for ”” (Hirudoid) and prescription trends (total) and (2) internet search trends for ”” (Hirudoid) and trends in heparinoid prescription in those aged 20-59 years. All trend data were calculated monthly. We used R, version 3.5.0 (The R Foundation), for statistical analysis.

The results showed that heparinoid prescriptions increased from 2012 onward and that “” (Hirudoid) was the most frequently searched of the three terms entered into Google Trends. In addition, prescription of heparinoid only had a weak positive correlation with internet search data obtained from Google Trends.

A previous study showed that drug utilization time series for antibiotics—amoxicillin, azithromycin, and cefdinir—and Google Trends were correlated [28]. This study showed similar results for heparinoid prescription. In particular, the results for the population aged 20-59 years correlated to a −1-month lag in Google Trends. This implies that there was a possibility that this population was affected by the internet and/or that they tend to seek out medical information before going a doctor. We cannot assess this relationship directly; however, we showed that there is a relationship between medical prescriptions and internet searches.

There are many reasons why people search for health-related information using the internet [33]. For instance, they may perform searches for a newly diagnosed health problem, ongoing medical condition, newly prescribed medication or treatment, or health-related knowledge. Therefore, people could search before or after undergoing medical treatment [34]. In this study, Google Trends showed that “” (Hirudoid) was the most commonly searched term related to heparinoid in patients aged 20-59 years. This result suggested that people visited a physician after obtaining information regarding heparinoid; it also suggested that internet searching affected their behavior and physicians’ decision making.

Many people were prescribed heparinoid in a clinic or hospital in this study, even though heparinoid is sold in pharmacies as an over-the-counter (OTC) medicine. There could be several reasons for this. First, people have to obtain a prescription from a physician to get Hirudoid, specifically. While people can get OTC medicines containing heparinoid, they cannot buy Hirudoid without a prescription. Results of Google Trends showed that Hirudoid was the most frequently searched of the three terms; this could be the main reason that people sought prescriptions. Second, because of the way the Japanese health care system works, some Japanese people might choose to visit clinics or hospitals for prescribed medicine in order to obtain it at a cheaper rate relative to the OTC price. In Japan, everyone has some type of insurance because of the national health insurance system. Therefore, they pay up to 30% of the total medical expenditure in hospitals, clinics, and pharmacies [35]. In a past questionnaire survey, 36.1% of patients did not consider the use of OTC drugs before visiting medical institutions. In addition, 40.5% of participants answered that visiting medical institutions was cheaper [36]. When people choose whether to visit a medical institution or buy OTC drugs, there are many reasons that affect their decision [37]. However, this could explain why people visit clinics or hospitals rather than buying OTC medicines.

Patients’ medical decision making is changed though health information, regardless of credibility [7,18]. A previous study showed no difference in the effects of high- and no-credibility sources [38]; judgment of credible information could be difficult for many consumers. Health literacy has been considered an important factor affecting health behavior and health information access [39-41]. However, levels of health literacy among the Japanese population are lower relative to those observed among the European population. They concluded that there are few reliable, understandable Japanese websites such as MedlinePlus, as well as a lack of health communication specialists and an inefficient primary health care system in Japan [42]. Moreover, some patients have explained that they do not use OTC drugs because they do not know which OTC drugs are the highest quality or most suitable for their symptoms and they cannot be trusted [36]. This may imply that people do not have the opportunity to learn how to obtain the correct information. Thus, education to enable people to achieve high internet health literacy levels and the ability to search and judge credible information is more important than ever.

On October 31, 2018, the Japanese Dermatological Association reported on medical heparinoid prescriptions for cosmetic purposes and stated that members should prescribe it appropriately; if it is prescribed for cosmetic application, patients should pay for the medication themselves [43]. In addition, the Central Social Insurance Medical Council developed two policies on February 7, 2018. The first policy states that heparinoid prescriptions shall not be covered for the purpose of cosmetic use to promote circulation and skin moisturization (ie, heparin sodium and heparinoid) [44]. The second policy states that examination and payment agencies should deal with prescriptions appropriately. However, prescriptions were not limited because of a lack of evidence regarding heparinoid prescription. Therefore, these findings provide important evidence for trends in heparinoid prescription.

This study was subject to several limitations. For example, the JMDC database contains data for employees of large corporations and their family members. However, because of a lack of data for employees of small and medium-sized businesses, public officials, and self-employed people, the results cannot be generalized to the wider population in Japan. In addition, Google and Yahoo are the main internet search engines used in Japan [45], and Google Trends does not include the entire Japanese population. However, the use of Google is increasing and it may, therefore, represent most Japanese people. Moreover, Google Trends shows data only for search queries; we did not have access to details about how research words were recognized and aggregated on Google or the information obtained using specific search terms. This study examined only the association between internet searches and prescriptions. Therefore, it did not clarify the cause of the increase in prescriptions or the number of people prescribed the moisturizer for cosmetic purposes due to a lack of detailed information about attitudes and prescription behaviors. To elucidate this relationship, another approach is needed; for example, a questionnaire survey administered to physicians or patients could assess moisturizer prescriptions for cosmetic purposes.

This was the first study to report on the association between internet searches and prescriptions in Japan. The results of the analysis suggested that internet searching for health-related information affected heparinoid prescription, although the positive correlation between these variables was weak. In particular, prescriptions for people aged 20-59 years correlated with the –1-month lag in Google Trends. Seeking internet-based information changed peoples’ behavior and physicians’ prescription habits. Also, Google Trends quantitatively demonstrated peoples’ interest and could be effective in detecting changes in drug utilization. The internet provides beneficial information to patients, medical personnel, and healthy people for decision making. However, it is not easy to distinguish between correct and incorrect information because of the volume included.