MSOA-level COVID-19 case count time series data

These data are weekly totals for new COVID-19 cases published daily for England by MSOA (Middle Level Super Output Area).

Data in this repository start from the 2–8 October period.

These data were originally pulled from the ArcGIS API, but from the 16th October are now pulled from the endpoint used by the new interactive map.

Why?

I originally couldn't find historical granular data anywhere, and wanted to be able to track numbers in my local area.

I've now found that PHE/NHSX publish historical weekly data but this is still not the entire dataset. That data has one value per week, whereas this is collecting daily rolling sums for week-long periods. This data will also track revisions to previously-released data.

Data files

Some days are currently missing data as I forgot to run my scripts on those dates. These should be obvious from missing entries in data/observations.csv. These data are expected to be filled in on subsequent weekiversaries (i.e. when fetching 7, 14, 21, etc days later).

The data/ directory contains:

data/observations.csv

A CSV file of all MSOA-level data points with the following columns:

• msoa — The MSOA code
• source — Either arcgis or api, corresponding to which API the data was pulled from. arcgis is from when the data were published in an ArcGIS dashboard, and api from when the cases map was moved to coronavirus.data.gov.uk.
• specimenDate — The end date of the period for which tests are counted in this observation.
• observationDate — The date reported by the API for when the tests were counted. It's possible to have multiple rows for a given specimen date and different observation dates if test results for the same day are reported at different times.
• fetchDate — When the data were fetched from the API
• rollingSum — The number of positive tests for this MSOA in the 7 day period up to and including the specimenDate. i.e. if specimenDate is 2020-10-20, then the rolling sum covers the period from the 14th to the 20th of October.

data/latest.csv

This file contains the latest data for each MSOA and specimen date in an enormous table, with MSOAs on the x axis, and specimen dates on the y axis. The cell values are again the 7-day rolling sums. You could open this in your favourite spreadsheet program, select the first column and the columns for the MSOAs your interested in, and create a graph of the rolling sums over time.

Be warned that this has ~6700 columns.

data/totals.csv

The total 7-day rolling sum for each day across England.

data/{date}.json

These are pretty much verbatim files from the ArcGIS API published on the date in the file name, with geometries removed. There is a wk{week}_{day} field analogous to the ISO week date format which contains the data for the 7 days up to and including the day in the field name. i.e. wk41_3 covers the period from 1 October to 7 October.

Note that the publication date is currently about four days behind the data. Use the field date, not the file date.

These files are kept in order to reconstruct other data files if later required. They are no longer created as the data is no longer maintained in the ArcGIS API.

data/{date}.csv

The same data as above as a CSV table.

data/geometries.json

The geometry data pulled from the ArcGIS API. These are all rough polygons, and probably not in a standard format. It would be trivial to convert them to GeoJSON though, or for users of this data to match up the MSOA codes to geometries from elsewhere (e.g. from MapIt).

Updating

These data are currently fetched, committed and pushed manually, The module for fetching new data is england_covid_msoa.update_observations_from_api. Data updates generally happen late afternoon.

I'm doing it manually so I can check that things are working properly each time, as I fully expect to have not anticipated some aspect of the data or API, or for the API to change from under me at some point.

If you run this API

I couldn't see how to get batch data out of any API or download on api.coronavirus.data.gov.uk at the MSOA level for individual days, so I've resorted to making a request for every single MSOA, with a 0.2s hard-coded interval between requests. If my script is causing you trouble, please get in touch and let me know how I can do things better for you.