Exploring a Sharp Rise in Google Searches for ‘Microclots’ between April 25, 2023 and May 1, 2023

Part 1 of a Google Trends Analysis of Trends from January 1, 2015 through May 1, 2023

1 Purpose:

This analysis is a response to an increase in Google searches for “microclots” between April 25, 2023 to May 1, 2023.

1.1 Responding to Google Search Trends Related to COVID-19 in Real Time

When trying to understand crises of global proportion like pandemics, war, climate-driven extreme weather events, and other mass-casualty incidents, it is important to recognize how quickly information relevant to the crisis emerges and changes. While examining changes to information over time retroactively can be useful, it is also important to observe these changes in-real-time given the extremely rapid pace of information discovery combined with the understanding that the discovery of new information drives changes to the “facts” of the crisis. The COVID-19 pandemic is one such crisis.

The understandings people hold about COVID-19 risk and mitigation, which I call the “social facts of COVID-19,” have remained in flux since the acknowledgement of the SARS-CoV-2 virus in early 2020. My previous research on the social facts of COVID-19 in the United States revealed an emerging and somewhat widespread belief starting as early as June 2020 that infection with and/or exposure to the SARS-CoV-2 virus was harmless for the majority of the population. Additionally, I found little evidence that this belief was becoming less dominant by the end of the study period in June 2021, instead finding a significant increase in “trusting” sentiment surrounding discussions of re-opening public venues in the United States once most people could easily access COVID-19 vaccines.
(Rosen 2023, forthcoming, “What do the People Say? Tweeting as a Tool for Establishing the Social Facts of COVID-19”)

Since June 2021, policy changes to remove COVID-19 protections combined with media rhetoric downplaying the severity of COVID-19 infection (in effect, normalizing regular periodic infection) indicate that the understanding of COVID-19 risk as negligible has persisted. This persistence is despite growing scientific evidence that COVID-19 infection, especially repeat infections, have harmful and lasting effects on the human body that are not specific to “medically vulnerable” populations (i.e. everyone is at risk). Once such effect is the risk of developing microclots, which can lead to thrombolytic events like stroke and cardiac arrest.

1.2 Tracking the COVID-19 Pandemic Response

Google trends reported that searches for “microclots” were up 23% for the period between April 25, 2023-May 1, 2023. The increase follows the release of mainstream media news reports on the long-term risks of COVID-19 infection, including that these risks increase exponentially with the third infection. To understand why the interest in “microclots” increased, it is important to understand what interest in “microclots” looked like prior to the COVID-19 pandemic, too.

2 About the Analysis

2.1 Analytic Focus

This analysis explores Google search trends for “microclots” and related queries from January 1, 2015-May 1, 2023 using tidy methods (tidyverse) and the “gtrendsR” “googleVis” and “ggplot2” packages in R.

2.2 Software and Programming

All analyses were executed in RStudio IDE for Mac (Intel, MacbookPro 2020, OS Ventura).

2.3 Examples vs. Tutorials by Medical Sociology on Wheels

The code and annotations in this document are part of my ongoing research programme focused on the social facts of COVID-19 risk and mitigation. Annotations are focused on interpreting the results of the analysis, and figures/visualizations are provided as a means of presenting important information about a trending interest related to COVID-19 infection. This document is not a tutorial on how to use gtrendsR, googleVis, ggplot2, and the tidyverse set of packages in R. The code included in this document may serve as a useful example if you have background knowledge of the packages used for the analysis. You can find tutorials with information about how to use gtrendsR and covering data visualization in R HERE.

--------------------------------------------------------------

3 Search trends for “Microclots” from January 1, 2015 through May 1, 2023

3.1 Basic Keyword Query - “microclots” only, full period

Code

library(tidyverse)
library(packrat)
library(tidytext)
library(tidyselect)
library(tidymodels)
library(httr)
library(jsonlite)
library(gtrendsR)
library(ggplot2)
library(googleVis)

Code

clots1 <- gtrends(keyword = c("microclots"),
                      time = "2015-01-01 2023-05-01")

3.1.1 Interest Over Time (iot) 1

Code

clots1_iot <- clots1$interest_over_time

3.1.1.1 Reformat ‘microclots’ Query 1 iot Dataframe for Plotting

Code

clots1_iot <- clots1_iot %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

3.1.1.2 ggplot2 Plot of iot for ‘microclots’ Query 1

Code

clots1_iot %>% 
  ggplot(aes(x = d_date,
             y = n_hits,
             color = keyword)) + 
  geom_point() +
  geom_line() +
  labs(title = "Google Search Trends for 'microclots' from Jan 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Global Search Hits",
       alt = paste("A line plot showing the proportion of Google search hits between January 1, 2017 and December 31, 2022, globally, for disability keywords: disability, chronic, illness, spoonie, and zebra. The proportion compares hits of each keyword versus the other four."))

3.1.1.3 Interactive Plot of iot for ‘microclots’ Query 1

Code

clots1_iot_gv <- gvisAnnotationChart(clots1_iot,
                                    datevar="d_date",
                                    numvar="hits",
                                    options=list(width=800, height=500, fill=10, displayExactValues=TRUE, colors="['#ff0000']"))

Code

plot(clots1_iot_gv)

View Interactive Plot for Interest Over Time

3.1.2 Interest by Country (ibc)

Code

clots1_ibc <- clots1$interest_by_country

3.1.2.1 Reformat ‘microclots’ Query 1 ibc Dataframe for Plotting

Code

clots1_ibc <- clots1_ibc %>%
  mutate(hits2 = replace(hits, hits==c(""), 0)) %>%
  filter(hits2 > 0) %>%
  mutate(hits3 = factor(hits2))

3.1.2.2 Find ‘microclots’ Query 1 Levels for Factored Hits

Code

levels(clots1_ibc$hits3)

[1] "12"  "17"  "33"  "65"  "70"  "83"  "93"  "100"

3.1.2.3 ggplot2 Plot of ibc for ‘microclots’ Query 1

Code

clots1_ibc %>%
  ggplot(aes(x = location,
             y = hits3,
             fill = location)) +
  theme(axis.text.x=element_blank(),
        axis.ticks.x=element_blank()) +
  geom_col() +
  labs(title = "Google Search Trends for 'microclots', January 1, 2015 - May 1, 2023",
       x = "Country",
       y = "Proportion of Google Search Hits",
       alt = paste("A barchart showing proportion of Google search hits for the keyword 'microclots', by country, for the entire period between January 1, 2015 and May 1, 2023")) 

3.1.2.4 Select Variables from ibc and Store in Smaller Dataframe for gviz Plotting

Code

clots1_ibc_baby <- clots1_ibc %>%
  select(location, hits)

3.1.2.5 Interactive Plot of ibc for ‘microclots’ Query 1

Code

clots1_ibc_gv <- gvisColumnChart(clots1_ibc_baby,
                              options = list(width=900, height=700, colors = "['#3cb371']"))

Code

plot(clots1_ibc_gv)

View the Interactive Plot for Interest by country

3.1.3 Preparing for Follow-Up Analyses

I can use topic modeling to analyze the related topics and related queries information. Storing information for later use in text analyses.

3.1.3.1 Extract and Store Related Topics in New Dataframe

Code

tops_rltd_to_clots <- clots1$related_topics

3.1.3.2 Extract and Store Related Queries in New Dataframe

Code

quers_rltd_to_clots <- clots1$related_queries

3.2 iot-Only Keyword Queries by Country

Examine interest over time by country. Need to pull iot by country to do this, not included with basic query.

Code

clots_Canada <- gtrends(keyword = c("microclots"),
                        time = "2015-01-01 2023-05-01",
                        geo = c("CA"),
                        onlyInterest = T)

clots_UK <- gtrends(keyword = c("microclots"),
                    time = "2015-01-01 2023-05-01",
                    geo = c("GB"),
                    onlyInterest = T)

clots_Philippines <- gtrends(keyword = c("microclots"),
                             time = "2015-01-01 2023-05-01",
                             geo = c("PH"),
                             onlyInterest = T)

clots_US <- gtrends(keyword = c("microclots"),
                    time = "2015-01-01 2023-05-01",
                    geo = c("US"),
                    onlyInterest = T)

clots_Australia <- gtrends(keyword = c("microclots"),
                        time = "2015-01-01 2023-05-01",
                        geo = c("AU"),
                        onlyInterest = T)

clots_Germany <- gtrends(keyword = c("microclots"),
                         time = "2015-01-01 2023-05-01",
                         geo = c("DE"),
                         onlyInterest = T)

clots_Netherlands <- gtrends(keyword = c("microclots"),
                             time = "2015-01-01 2023-05-01",
                             geo = c("NL"),
                             onlyInterest = T)

clots_Italy <- gtrends(keyword = c("microclots"),
                       time = "2015-01-01 2023-05-01",
                       geo = c("IT"),
                       onlyInterest = T)

3.2.1 Reformat the Data for Plotting

Code

clots_Canada <- clots_Canada$interest_over_time

clots_UK <- clots_UK$interest_over_time

clots_Philippines <- clots_Philippines$interest_over_time

clots_US <- clots_US$interest_over_time

clots_Australia <- clots_Australia$interest_over_time

clots_Germany <- clots_Germany$interest_over_time

clots_Netherlands <- clots_Netherlands$interest_over_time

clots_Italy <- clots_Italy$interest_over_time

Code

clots_Canada <- clots_Canada %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_UK <- clots_UK %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_Philippines <- clots_Philippines %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_US <- clots_US %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_Australia <- clots_Australia %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_Germany <- clots_Germany %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_Netherlands <- clots_Netherlands %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

clots_Italy <- clots_Italy %>%
  mutate_at("hits", ~ifelse(. == "<1", 0.5, .)) %>%
  mutate(n_hits = as.numeric(hits)) %>%
  mutate(d_date = lubridate::as_date(date))

3.2.2 Select Variables of Interest from Each Dataframe, Store in Smaller Dataframes

Code

clots_CAN_baby <- clots_Canada %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_UK_baby <- clots_UK %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_PH_baby <- clots_Philippines %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_US_baby <- clots_US %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_AU_baby <- clots_Australia %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_DE_baby <- clots_Germany %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_NL_baby <- clots_Netherlands %>%
  select(date = d_date, hits = n_hits, country = geo)

clots_IT_baby <- clots_Italy %>%
  select(date = d_date, hits = n_hits, country = geo)

3.2.3 Bind the Baby Frames

Code

clots2_iot <- rbind(clots_CAN_baby, clots_UK_baby)

clots2_iot <- rbind(clots2_iot, clots_PH_baby)

clots2_iot <- rbind(clots2_iot, clots_US_baby)

clots2_iot <- rbind(clots2_iot, clots_AU_baby)

clots2_iot <- rbind(clots2_iot, clots_DE_baby)

clots2_iot <- rbind(clots2_iot, clots_NL_baby)

clots2_iot <- rbind(clots2_iot, clots_IT_baby)

3.2.4 ggplot2 of Interest Over Time for “microclots” Query [Set] 2 (by Country)

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots2_iot %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("skyblue1",
                                "lightgoldenrod1",
                                "chartreuse2",
                                "turquoise1",
                                "firebrick1",
                                "orchid2",
                                "pink1",
                                "chocolate1"),
                                       breaks = c("CA",
                                                  "GB",
                                                  "PH",
                                                  "US",
                                                  "AU",
                                                  "DE",
                                                  "NL",
                                                  "IT")) +
  labs(title = "Google Search Trends for 'microclots', by Country, from January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits (within Country)") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Canada Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_CAN_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("skyblue1"),
                    breaks = c("CA")) +
  labs(title = "Google Search Trends for 'microclots', Canada, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

UK Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_UK_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("lightgoldenrod1"),
                    breaks = c("GB")) +
  labs(title = "Google Search Trends for 'microclots', United Kingdom, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Philippines Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_PH_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("chartreuse2"),
                    breaks = c("PH")) +
  labs(title = "Google Search Trends for 'microclots', Philippines, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

US Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_US_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("turquoise1"),
                    breaks = c("US")) +
  labs(title = "Google Search Trends for 'microclots', Philippines, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Australia Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_AU_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("firebrick1"),
                    breaks = c("AU")) +
  labs(title = "Google Search Trends for 'microclots', Australia, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Germany Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_DE_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("orchid2"),
                    breaks = c("DE")) +
  labs(title = "Google Search Trends for 'microclots', Germany, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Netherlands Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_NL_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("pink1"),
                    breaks = c("NL")) +
  labs(title = "Google Search Trends for 'microclots', Netherlands, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

Italy Only Plot

Code

par(mar = c(4, 4, 5, 8),
    oma = c(0.2, 0.2, 0.2, 0.2),
    cex = 0.5,
    cex.lab = 0.4)

clots_IT_baby %>% 
  ggplot(aes(x = date,
             y = hits,
             color = country)) + 
  geom_point() +
  geom_line() +
  scale_color_manual(values = c("chocolate1"),
                    breaks = c("IT")) +
  labs(title = "Google Search Trends for 'microclots', Italy, January 1, 2015 - May 1, 2023",
       x = "Date",
       y = "Proportion of Hits") +
  theme_dark(
    base_size = 10,
    base_family = "serif",
    base_line_size = 14/20,
    base_rect_size = 14/20
  )

4 Conclusions

Searches for “micoclots” from the United Kingdom, Germany, and the United States had clear positive trends, with later time points accounting for the majority of hits for all three countries. However, in 2016 when the increases begin to show, the hits accounted for a much higher proportion of German and UK searches than US searches for “microclots.” German search hits for “microclots” also reach their peak much earlier in 2021 than US search hits, which peak just prior to the start of 2022, drop drastically, then increase sharply for 2022 into May 2023. The peak interest for “microclots” in the United Kingdom occurred slightly later, well into 2022. Considering the large proportion of Uk, German, and US, search hits for “microclots” occurring between 2020-2023, and given the increase becomes more noticeable in 2021 when the science of COVID-19 risk became clearer and the understanding that COVID-19 infections are associated with long term issues with microclotting even/especially for those with “mild” acute SARS-CoV-2 infection became more widespread among scientists, the connection between COVID-19 and microclots was possibly important to the heightened search traffic for “microclots” from Germany, the United Kingdom, and the United States, during the pandemic (est. 2020).

Trends for the other countries (Canada, Philippines, Australia, Netherlands, and Italy) experienced higher peaks and lower troughs than UK, German, and US, search trends for “microclots.” Importantly, the remaining countries seemed to experience “all-or-nothing” patterns to their search hits for “microclots” where searches were either at their peak or virtually non-existent. The peaks seem to occur more frequently for some countries, like Italy, towards the end of the time period, during the COVID-19 pandemic. Most saw high search traffic for microclots as early as 2016, though. This is an important indication that information about microclots may have been more readily available to the public in Canada, the Philippines, Australia, and the Netherlands, than for people in the UK, US, Germany, or Italy, beginning at least several years prior to the COVID-19 pandemic. If this is the case, Canada, the Philippines, Australia, and the Netherlands, may be better poised to manage the long-term complications of COVID-19, particularly those related to microclotting and thrombolytic events, than the United States, United Kingdom, Germany, and Italy. Additional research into the handling of post-acute Covid sequalae should consider these pre-pandemic differences in search traffic for “microclots” when interpreting differences in pandemic-era responses to microclotting.