Eysenbach [1] in 2002 defined Infodemiology as the “science of distribution and determinants of information in an electronic medium, specifically the Internet, or in a population, with the ultimate aim to inform public health and public policy.” Since then, web-based sources have been suggested to be valuable in various academic fields for understanding timely and regional differences in the public interest in certain terms and concepts [2].

Especially in health and medical research, infodemiological studies using Google Trends have increased over the last decade [3]. Recently, several studies have used Google Trends data to elucidate the progression of the COVID-19 pandemic. For example, Husain et al [4] used Google Trends to track public interest in COVID-19 [4] and reported that in the United States, when containment and mitigation strategies were first implemented, public interest was very low. However, the search volume of COVID-19–related terms increased and correlated with the increase in positive findings on COVID-19 tests nationwide. Furthermore, Walker et al [5] correlated the search activity for the term “loss-of-smell,” an early symptom of COVID-19, to the number of daily confirmed cases and COVID-19 deaths. Similarly, Jimenez et al [6] correlated searches for different symptoms of COVID-19 with daily incidences. Another study used Google Trends to identify mental health consequences related to physical distancing during the lockdown. Notably, Knipe et al [7] reported a reduction in search activity related to suicide and depression after the announcement of the pandemic.

The use of internet searches to gather information about patients experiencing pain has been insufficiently studied. Of 2 studies on populations of people with chronic pain, 1 reported that 24% of the population searched the internet for pain-related information, and the other reported that 39% of the population performed internet searches [8,9]. However, both studies are more than 10 years old. Considering the recent growth of the usage of online media, it is reasonable to suppose that the proportion of individuals searching for pain-related information on the internet increased during the last decade.

The COVID-19 pandemic has affected the health condition and quality of life of people with chronic pain in 2 ways. First, several changes have led to increased susceptibility to and a higher risk of pain exacerbation. Specifically, the availability of health care was significantly reduced and an atmosphere of fear and isolation was developed [10]. Previous experiences with epidemics have shown a latent, but long-term, deterioration of psychosocial problems [11]. Therefore, it has also been postulated that chronic pain, as a biopsychosocial phenomenon, exacerbates in response to the COVID-19 situation [12]. Second, acute stress can alleviate chronic pain [13,14]. Indeed, acute stress in response to disasters alleviates pain [15]. One could therefore hypothesize that the COVID-19 situation alleviates chronic pain, at least at the onset of the pandemic. To date, no studies have examined changes in the experience of chronic pain after the onset of the COVID-19 pandemic in large patient populations. To bridge this knowledge gap, search activity can be used as a surrogate for the public interest in pain. This does not necessarily reflect the incidence or the burden of pain in the general population; however, acute changes can be interpreted as a momentary shift of attention and thus highlight the importance of these symptoms.

This study aimed to assess the Google search activity of pain-related expressions in the most populated regions worldwide and to use them as markers of the social importance of pain before and during the first wave of the COVID-19 pandemic.

This infodemiological study was designed and carried out at Graz Medical University in cooperation with the Graz University of Technology and Fraunhofer Austria. The workflow of this study is presented in Figure 1. The temporary popularity of pain-related search terms was assessed using Google Trends [16]. Google Trends delivers data on web-based search queries for specified timeframes and countries. Google Trends offers 2 analysis modes: one with “terms” and the other with “topics.” In contrast to the analysis with “terms,” Google Trends analysis with “topics” corresponds to a fuzzy search. The difference is that “topics” include all search terms related to a particular aspect, whereas “terms” are specific, and the results will only show the relative volume of the corresponding term. Since different types of pain are related to one another—for example, neck pain and shoulder pain or dysmenorrhea and pelvic pain—a fuzzy search that includes related terms is not adequately discriminative. Furthermore, it is unclear which terms are summarized in a specific “topic.” Hence, smaller, discriminative trend analysis with “terms” was preferred. Data are presented as relative popularity, normalized to the region and period and scaled from 0 to 100. Subsequently, the data were exported for further calculations [17]. The study followed the methodology framework developed by Mavragani and Ochoa [18].

The search terms used in this study were based on a recent study of global internet searches associated with pain [19]. In total, 24 expression entities based on anatomical regions were used: “abdominal pain,” “back pain,” “breast pain,” “chest pain,” “dysmenorrhea,” “dyspareunia,” “ear pain,” “epigastric pain,” “eye pain,” “groin pain,” “headache,” “knee pain,” “low back pain,” “neck pain,” “odynophagia,” “pelvic pain,” “penile pain,” “podalgia,” “rectal pain,” “shoulder pain,” “sore throat,” “testicular pain,” “toothache,” and “wrist pain.” The results for 3 of these terms (“penile pain,” “podalgia,” and “rectal pain”) were excluded from further analysis because of insufficient data. Weekly data sets were downloaded from January 1, 2018, to December 31, 2020. Google Trends analysis was carried out in February 2021.

Our intention was to provide an overview of the development of the aforementioned pain-related search parameters in the most populated regions worldwide. The most populous areas worldwide were extracted from the recommendations of the United Nations and a web-based population database [20,21]. Europe, consisting of individual European countries, is defined as a single region in this study. Figure 1 shows all the selected regions and the individual European countries. We identified the following as the most populated regions: Europe, the United States, China, India, Pakistan, Brazil, and Indonesia. Search terms were extracted in the respective language of each country. The data set of Europe is accumulated; it consists of all available nationwide data sets of the European countries weighted by their population. Details including the complete list of countries, their population sizes, the corresponding weighting factors, and the date of the first officially reported COVID-19 case of the country were provided by the European Centre for Disease Prevention and Control. Using this list of European countries, we retrieved a data vector for each country and pain type in the respective language of each country. Thereafter, the individual vectors were weighted with the country’s population and accumulated to a European vector (Multimedia Appendix 1). Since the weighted summation with sum 1 is an affine map, the result is also normalized. Calculations with the European vector thus correspond to stratified sampling; that is, individual characteristics of the target group deemed relevant for the study are transferred to the sample in approximately the same proportion as those in the population.

Data from China could not be extracted adequately, and data extraction yielded numerous error parameters. Therefore, China had to be excluded from our analysis.

We assumed that the occurrence of biological and psychosocial effects of the pandemic correlated with the first appearance of the virus in a certain region. Accordingly, we defined the beginning of the “COVID-19 period” for each geographic region with the first officially confirmed case for each country on the basis of reports from the European Centre for Disease Prevention and Control [22]. The date of the first officially reported COVID-19 case for each country was used as the independent variable for our calculations.

For data extraction, we used Python scripts. The Google Trends analysis data were obtained with the software libraries pytrends and pandas [23,24] using Python. The number of COVID-19 cases were obtained from the European Centre for Disease Prevention and Control on February 9, 2021. Results from countries with erroneous data or insufficient cases and subsequently inconsistent data are excluded from the study (Figure 1).

IBM SPSS Statistics (version 26, IBM Corp), Microsoft Office 365, and LibreOffice 7.1 were used for the statistical calculations. Explorative data analysis and nonparametric Mann–Whitney U tests were performed. The level of significance was not corrected for multiple comparisons. These corrections are necessary for multiple tests carried out with the same samples: the greater the number of hypotheses tested on 1 data set, the higher the probability that 1 of them is (incorrectly) assumed to be true. As the data were collected separately by country and pain type, the precondition for test corrections (all tests on 1 data set) was not met; that is, Bonferroni corrections or similar methods were not required and hence not applied.

Additionally, we performed a visual analysis to illustrate the difference between significant and nonsignificant changes in the time series data (Figures 2 and 3) and to plot a side-by-side comparison of the progression of pain-related search criteria with the so-called waves of COVID-19 outbreaks (Figure 4). This analysis has been performed for “headache,” as this query term increased significantly in all geographic regions, and for “wrist pain,” as this query term did not show any significant changes.

For both terms, the analysis involved a seasonal autoregressive model whose parameters have been determined using Google Trends vectors retrieved for 2018 and 2019. Using these models, weekly differences between the model values for 2018-2019 and 2020 were calculated and plotted (Figures 2 and 4).

Our results demonstrate significant differences in pain-related search queries by comparing quantities before and after the first confirmed COVID-19 case. We observed a peak in the incidence of pain-related search parameters in March and April 2020 (Figure 5). Abdominal pain, dysmenorrhea, dyspareunia, groin pain, eye pain, knee pain, low back pain, and pelvic pain were the only pain types with a significantly decreased frequency of search relevance in some of the most populous geographic regions since the COVID-19 outbreak (Table 1). In contrast, the frequency of searches related to back pain, breast pain, chest pain, ear pain, headache, odynophagia, neck pain, shoulder pain, sore throat, testicular pain, toothache, and wrist pain significantly increased after the first confirmed case of COVID-19 was reported (Table 1, Figure 4).

Our data show that in Europe, Brazil, Pakistan, and India, all pain entities listed above increased significantly or remained unchanged after the first confirmed COVID-19 case was reported. In contrast, in the United States, we observed an inhomogeneous trend, with the frequency of some pain-related search terms decreasing significantly after the first confirmed COVID-19 case was reported, whereas that of the others increased or remained unchanged (Table 1).

A comparison among individual European countries revealed that the most frequently observed significant differences among countries were noted for headache, chest pain, sore throat, abdominal pain, and back pain since the COVID-19 outbreak. With respect to individual European countries, Spain displayed the most frequent significant pre– and post–COVID-19 discrepancies in the frequency of pain-related search queries (Figures 3 and 6).

The analysis of the difference plots for the term “headache” revealed that the weekly differences between the 2020 and the 2018-2019 models were more often significantly positive than negative; in other words, the number of queries for “headache” increased. As the Google Trends data are normalized with a range of 0 to 100, the 2018-2019 model was normalized accordingly (Figure 2).

If the search frequency of a term did not increase significantly, the differences would be expected to be as often positive as negative. Consequently, the difference plots would resemble random noise; for example, the search term “wrist pain” displayed no significant change in most geographic regions and its plots illustrate random noise around 0.

Furthermore, on assessing country-specific developments in pain-related search parameters, which have significantly increased in relevance since the COVID-19 pandemic outbreak, it becomes apparent that the relevance of pain-related search parameters increased in almost all countries before the peak of the first wave of the COVID-19 pandemic. “Headache” was consistently high in relevance, especially in Belgium, Brazil, Germany, Indonesia, Italy, Russia, Turkey, and the United States. The relevance of “chest pain” and “sore throat” seemed to resonate with the first wave of the COVID-19 pandemic, as seen in several countries including Belgium, Croatia, France, Italy, the Netherlands, and the United Kingdom. In contrast, the frequency of “abdominal pain” appeared highly dynamic compared to that of COVID-19 cases (Figure 4).