COVID-19: WHO Status Reports

Introduction

This document records my efforts to work out displays for COVID-19 cases, based on status reports from the World Health Organization, as recorded in Edward Brown’s package data2019nCoV.

Installation and Setup

General tools we use:

library(tidyverse)
library(plotly)
library(glue)

Package data2019nCoV updates frequently, so you’ll want to check the Github repository for updates and re-install as needed:

devtools::install_github("eebrown/data2019nCoV")
library(data2019nCoV)

The Data

We’ll work with WHO_SR:

View(WHO_SR)
help("WHO_SR")

Each row is a WHO situation report for a large number of countries (and some other groups such as health care providers).

Comparing a Few Countries

Our goal is to write functions that permit the user to view the number of confirmed cases over time, for selected countries.

Basic selection:

select_who <- function(countries) {
  ## makes a regex like "Iran|France|Spain":
  regex <- paste0(countries, collapse = "|")
  WHO_SR %>% 
    ## selec tonly columns contaning name of
    ## one of the desired countries:
    select(Date, matches(regex)) %>% 
    ## just looking for total confirmed cases for each country
    ## not breakdown by region, etc:
    select(-matches("\\."))
}

Try it out:

countries <- c("UnitedStatesofAmerica", "Canada", "Brazil")
df <- 
  countries %>% 
  select_who()
DT::datatable(df)

Each row is a small part of a WHO status report.

We will transform the data so that it is glyph-ready for a plot of the number of confirmed cases in each country against the date of the report. tidyr::gather() is our friend, here.

df_long <- tidyr::gather(df, key, value, countries)
DT::datatable(df_long)

Note: It is very handy that gather() allows one to specify the columns to be gathered via a character-vector of the column names.

df_long %>%
  ggplot(aes(x = Date, y = value, col = key)) +
  geom_line() +
  labs(
    title = "Confirmed Cases",
    x = "Date",
    y = "Confirmed Cases"
  ) +
  theme(legend.title = element_blank())

Very nice. But can we make the graph interactive?

## make plot object
p <-
  df_long %>%
  ggplot(aes(x = Date, y = value, col = key,
             text = glue::glue(
               "{key}<br>{value}<br>{Date}"
             ))) +
  geom_line() +
  labs(
    title = "Confirmed Cases",
    x = "Date",
    y = "Confirmed Cases"
  ) +
  theme(legend.title = element_blank())

## call in plotly:

plotly::ggplotly(p, tooltip = "text") %>% 
  plotly::config(displayModeBar = FALSE)

Oops! The tooltip shows up fine, but the line plots are missing. I have googled this problem and found that (for reasons I don’t yet understand), you can get the lines if you use the group aesthetic, as follows:

## make plot object mostly as before:
p <-
  df_long %>%
  ggplot(aes(x = Date, y = value, col = key,
             group = key,  ## to satisfy ggplotly!!
             text = glue::glue(
               "{key}<br>{value}<br>{Date}"
             ))) +
  geom_line() +
  labs(
    title = "Confirmed Cases",
    x = "Date",
    y = "Confirmed Cases"
  ) +
  theme(legend.title = element_blank())

## call in plotly:

plotly::ggplotly(p, tooltip = "text") %>% 
  plotly::config(displayModeBar = FALSE)

Encapsulate in a Function

We can encapsulate our programming work into a function:

cases_plot <- function(countries) {
  p <- 
    select_who(countries) %>% 
    gather(key, value, countries) %>% 
    ggplot(aes(x = Date, 
               y = value, 
               col = key,
               group = key,
               text = glue::glue(
                 '{key}<br>{value}<br>{Date}'
               ))) +
    geom_line() +
    theme(legend.position="none") +
    labs(title = "Confirmed Cases",
         x = "Date", 
         y = "Confirmed Cases") +
    theme(legend.title = element_blank())
  ggplotly(p, tooltip = "text") %>% config(displayModeBar = FALSE)
}

Try it out:

cases_plot(countries)

This works, but note that for some reason we lost the legend! Thankfully, the presence of the tooltip prevents this from being a big deal.

Where to Go From Here

• Develop other analyses, e.g.:
  • display the data on a world map;
  • dig down into suspected cases, deaths, etc., to the extent provided by WHO_SR;
  • etc.
• Shiny-fy our analysis.
• Continue o study of plotly::ggplotly(), along with other interactive-graphics tools.

For more ideas consult the package vignettes:

• WHO SR vignette
• ON vignette