A Guide to my Covid-19 Charts
There are many excellent chart resources around the world, I’ve been developing my own since the start of March with a focus on Ontario and Canada. As the pandemic has progressed, so have my data viz. Most are interactive Tableau’s, some are automated R-charts. Hopefully this blog helps you navigate and find what you need. Because my Tableau’s are always in development, please contact me if anything isn’t working right. NEW: Vaccine charts can be found in tabs within the Canada Covid-19 Dashboard and Ontario Key Indicators viz below.
My Tableau viz are all found at bit.ly/Billius27. The R charts case and death trend catalogue for every country, province, and state can be found at bit.ly/billius27drive.
1. Canada Covid-19 Dashboard
When?
As of January 15, 2021 this dashboard is now updated each morning, typically before 9am. Canadian Public Health Units report daily and this data is compiled by the Canada Covid-19 Open Data Working Group each evening.
Plans are underway to use covid19tracker.ca as the source for vaccine data (1/19/21).
How to Use and Key Information
Here’s the details provided by the main dashboard Info button.
Why?
This viz provides the most timely, detailed and complete chart-based summary of the Canadian covid situation. Because it uses PHU reports, it is more timely than next day official provincial reports and allows ranking and trend comparisons of regions across the country.
What’s it look like?
Because it is interactive, there are several stats options and hundreds of possible charts, from Nunavut daily cases to Canada death rates per 100K. Scroll, hover, select, highlight, or choose tabs across the top. There is also a special multi-chart dashboard for the largest province, Ontario.
NEW JAN 2021! A Vaccines tab provides scrollable charts on Canada’s vaccine delivery, for Canada and by province. Sample:
Here’s some samples from the scrollable Canada main dashboard, using weekly cases per 100K, a favourite statistic of mine. I like it because it smoothes over daily reporting issues and adjusts for population. Great way to compare regions or provinces.
The main advantage of the viz is that you can create a dashboard and custom charts for your own region. Plus everything has an additional layer so you can “hover to uncover” stats and dates beneath that mapped region, bar chart, or trend curve. It even includes an option for you to look at your region’s data over time, so you can create your own chart if you like.
2. Ontario Regions and Demos
When?
Each morning the provincial government reports its official numbers in a daily snapshot at 1030am. I access the provincial datasets directly around 1015am to update this viz then compare them to previous days to calculate some of the stats. Provincial numbers extract regional data at various times during the previous day then do certain cleaning and corrections, eg removing probable cases. So they will never match what regions report on their websites. The province also uses standard definitions for resolved and active that differ from regions. The Info button provides help.
Why?
While PHU reports are available the previous evening, this dashboard focuses on the official provincial numbers, the ones that drive policy. It also includes additional insights, including active case levels, test positivity and demographics.
What’s it look like?
The main dashboard packs 5 key indicators for each region as well as age and infection source. If you scroll down, you will find a map and an interactive chart to study trends by symptom onset. Because the chart is packed, there are also fuller-sized charts under the tabs at the top.
Let’s call it the “Sardine can” but I designed this to provide more perspective than the standard daily case reports. Weekly cases per 100K with the percent change from last week and test positivity are critical metrics in monitoring a region’s status. New cases and the 7 day average let you understand the absolute counts and help see if the day is out of pattern. Active cases let you get an idea of how much load your public health system is under, whether it’s test, trace, isolate or hospitals and long term care.
You can also see the age and infection source stats for active cases. Plus you can switch case types, eg from “Active cases” to “Fatal”.
Weekly case rate colour scheme:
Like my Canada PHU viz, I use a standard weekly case rate colour scheme for every region and province. Red is the range I arrived at by studying countries and regions over the summer. It means “proactive action needed”. Blue is a stage that can be described as “losing control”. Purple means the virus is winning. These ranges are absolute and independent of what governments and officials choose to display and so not concede ground because other countries or provinces are now doing worse. They are meant to be easily understood. If you have over 42 cases in the last seven days for every 100,000 people in your region, your region has left more moderate levels behind and needs to take stronger actions. Checking the direction arrow lets you know if things are worsening.
You can scroll down to a map. Same colour scheme. Here the bubbles are also sized by weekly case rates and centred on the PHU cities to get a unique view of the regions.
Because this viz has underlying data at the individual case level it supports deeper exploration. Here is the “epicurve” that is available for every region, age group, and infection source. Be careful to not be misled by drop offs in the latest 14 days, they are affected by people with recent symptoms not being reported yet.
From the top tabs, you can find this age heat map, describing the progression of cases from younger age groups to older groups over the Fall. It’s lags because of the symptom onset reporting issue I mentioned.
3. Ontario Key Indicators Dashboard
Like the “Regions and Demos” viz, this one is updated after 10am each day, tied to official provincial reports. To create this dashboard, I access a wide variety of online datasets from the Ontario government and elsewhere and also add stats manually for data they do not provide, like Rt and new hospitalizations.
It’s important to understand that the official provincial numbers will not match the ones you see reported regionally on websites for today or yesterday. That’s because they are based on different timing and use different criteria that are consistent across all regions, eg removing probable cases.
The purpose is to add a more complete picture to the daily ‘official’ government numbers by showing changes, trends, and information not normally announced, especially those related to outcomes like hospitalizations and deaths. The general theme is to show trends that help let you understand where the numbers are heading despite daily spikes and drops. Again, I use a method called “loess” to smooth and detect some of these trends.
The Info button provides important tips and information.
When?
Between 10 and 1030 am each day. Some stats may be updated during the day or lagged, eg school data is not available until mid-day and new hospitalizations and new ICU patients are lagged a day or two.
Why?
I have not seen a comprehensive or timely dashboard for Ontario, so this is my attempt to close the gap and present multiple key indicators in one place.
What’s new?
A “Vaccinations Board” with multiple scrollable charts is now available via a tab at the top.
What’s does the viz look like?
The main dashboard is constantly evolving. It’s scrollable and includes a banner with key stats at the top and then three sections of four trend charts each. You can link to the “Regions and Demos” dashboard from the upper left too.
Head Banner as of January 16th.
Over ten trend charts are organized from top to bottom by priority, starting with new cases and vaccinations. Several include additional information and day/day changes in the titles.
The default trend smoothing uses a mathematical method (loess) which can better adapt to trend changes and erratic daily reporting. Sometimes a 7-day moving average is used and that is noted.
The weekly test positivity I show takes the daily reported positivity from the labs and then averages them across the last 7 days weighted by the completed tests each day. This then becomes a more accurate and stable estimate of total test positivity over the past week for all of the tests completed. (I’ve found that the daily positivity is subject to a repeatable weekly pattern of testing and backlogs, leading to misunderstood peaks and valleys.) It’s important that we also appreciate how higher and widened testing levels can increase new cases while lowering positivity and that’s why this chart is in the top 4. A positivity above 2.5% is generally considered high enough to warrant public health actions. A level above 5% is considered “a LOT of Covid” by the ex-CDC director, Tom Frieden.
Other charts focus on addition measures of spread. Long term care and school outbreaks/cases are covered here. School data is underreported due to a government decision to not conduct surveillance testing among a younger population that can often be asymptomatic and a lack of parental off-work support programs to encourage testing. The term “school-related” cases is used to recognize that we don’t know where the transmission occurred. Cases can generated from community or household transmission as well. However, the actual number of school infections is unknown given the low level of test surveillance of kids who can often be asymptomatic.Rt is supplied by Ryan Imgrund. I’ve simply converted it so that 1 is 0 on the chart, allowing better graphics. An Rt above 1 (red in chart) signals an exponential growth in infections. Likewise, the speed signal indicates increasing spread when it is above zero. It’s derived from the slope of the new cases curve, credit for the concept goes to Jonathan Wang.
Other important charts include the cumulative wave 2 deaths and LTC deaths.
New hospitalizations are derived from the daily reports of “Ever Hospitalized” reported in the Ontario daily status report. Loess smoothing is needed to help decipher the noisy reporting of this indicator.
There are many additional specialized charts in this viz, I will highlight two more. Another tab labelled “Severity” at the top contains trend analyses for current hospitalizations, current ICU cases, and deaths. A combination of loess and 14 day moving averages are shown as well as the speed signal underneath. Note that a moving average will generally be lower than a current rising trend because it averages over the past whereas a smoothing like loess looks at mathematically fitting and smoothing the trend at every point.
The speed signal is an important indicator because it helps “zoom in” on the changes in the hospitalization curve. Consistent red bars represent a strong and persistent rise in hospitalizations.
The “Metrics” tab may be the biggest secret on my Tableau site. It allows the exploration of over 60 indicators, with numerous trends on case, deaths, test, and hospitalization data. It also includes trends on topics you may not see elsewhere like licensed child care.
4. Other Tableau Dataviz
I encourage you to check out my other Covid-19 Tableau’s at my bit.ly/Billius27 site. They go on to chart PHU country comparisons and Ontario-focused charts on test positivity by age, outbreak cases by category, regional testing trends, case rates by PHU and age, as well as and a greater geographical breakdown of hospitalizations.
5. R Charts
Though not interactive, these charts at my googledrive cover all countries, Canadian provinces, and US states. Each group can be found in a downloadable pdf book with download, print and zoom options. The trend charts include both 7-day moving averages and loess smoothing. Johns Hopkins, Our World in Data, and the Canada Covid-19 Open Data group are all used as sources.
Sample: Daily Canada case growth summary.
Summarizing the trend for the four western provinces.
Example of a US state summary chart.
There are charts for every country in the world, showing daily values as well as 7 day averages and loess trends.
These charts can help keep you up-to-date on Covid-19 trends. They are in constant development as more data sources become available and I find time to improve them.
#flattenthecurve
- Bill Comeau December 12, 2020
