Awesome

sars2pack

Overview

The sars2pack R package provides one-line access to over 40 COVID-related datasets. Datasets are accessed in real time directly from their sources and then transformed to tidy-data form where possible and applicable. The result of each dataset accessor is a ready-to-use R dataset, often a dataframe. Documentation includes dataset descriptions, sources and references, and examples. Online documentation is available in two locations:

• The sars2pack documentation, which includes reference docs and detailed dataset descriptions.
• Extended workflows and use cases, as an online book

Questions addressed by sars2pack

• What are the current and historical total, new cases, and deaths of COVID-19 at the city, county, state, national, and international levels?
• How do changes in infection rates differ across locations?
• What are the non-pharmacological interventions in place at the local and national levels?
• In the United States, what is the geographical distribution of healthcare capacity (ICU beds, total beds, doctors, etc.)?
• What are the published values of key epidemic parameters, as curated from the literature?

Installation

# If you do not have BiocManager installed:
install.packages('BiocManager')

# Then, if sars2pack is not already installed:
BiocManager::install('seandavi/sars2pack')

After the one-time installation, load the packge to get started.

library(sars2pack)

Available datasets

Updated tracking of city, county, state, national, and international confirmed cases, deaths, and testing is critical to driving policy, implementing interventions, and measuring their effectiveness. Case tracking datasets include date, a count of cases, and usually numerous other pieces of information related to location of reporting, etc.

Accessing case-tracking datasets is typically done with one function per dataset. The example here is data from the European Centers for Disease Control, or ECDC.

ecdc = ecdc_data()

Get a quick overview of the dataset.

head(ecdc)

## # A tibble: 6 x 8
## # Groups:   location_name, subset [6]
##   date       location_name iso2c iso3c population_2019 continent subset    count
##   <date>     <chr>         <chr> <chr>           <dbl> <chr>     <chr>     <dbl>
## 1 2019-12-31 Afghanistan   AF    AFG          38041757 Asia      confirmed     0
## 2 2019-12-31 Afghanistan   AF    AFG          38041757 Asia      deaths        0
## 3 2019-12-31 Algeria       DZ    DZA          43053054 Africa    confirmed     0
## 4 2019-12-31 Algeria       DZ    DZA          43053054 Africa    deaths        0
## 5 2019-12-31 Armenia       AM    ARM           2957728 Europe    confirmed     0
## 6 2019-12-31 Armenia       AM    ARM           2957728 Europe    deaths        0

The ecdc dataset is just a data.frame (actually, a tibble), so applying standard R or tidyverse functionality can get answers to basic questions with little code. The next code block generates a top10 of countries with the most deaths recorded to date. Note that if you do this on your own computer, the data will be updated to today’s data values.

library(dplyr)
top10 = ecdc %>% filter(subset=='deaths') %>% 
    group_by(location_name) %>%
    filter(count==max(count)) %>%
    arrange(desc(count)) %>%
    head(10) %>% select(-starts_with('iso'),-continent,-subset) %>%
    mutate(rate_per_100k = 1e5*count/population_2019)

Finally, present a nice table of those countries:

knitr::kable(
    top10,
    caption = "Reported COVID-19-related deaths in ten most affected countries.",
    format = 'pandoc')

Examine the spread of the pandemic throughout the world by examining cumulative deaths reported for the top 10 countries above.

ecdc_top10 = ecdc %>% filter(location_name %in% top10$location_name & subset=='deaths')
plot_epicurve(ecdc_top10,
              filter_expression = count > 10, 
              color='location_name')

Comparing the features of disease spread is easiest if all curves are shifted to “start” at the same absolute level of infection. In this case, shift the origin for all countries to start at the first time point when more than 100 cumulative cases had been observed. Note how some curves cross others which is evidence of less infection control at the same relative time in the pandemic for that country (eg., Brazil).

ecdc_top10 %>% align_to_baseline(count>100,group_vars=c('location_name')) %>%
    plot_epicurve(date_column = 'index',color='location_name')

Contributions

Pull requests are gladly accepted on Github.

Adding new datasets

See the Adding new datasets vignette.

Similar work

• https://github.com/emanuele-guidotti/COVID19
• Top 25 R resources on Novel COVID-19 Coronavirus
• COVID-19 epidemiology with R
• https://github.com/RamiKrispin/coronavirus
• Youtube: Using R to analyze COVID-19
• DataCamp: Visualize the rise of COVID-19 cases globally with ggplot2
• MackLavielle/covidix R package