Visualisation of local statistics
Covid-19 weekly incidence rate in the UK
I have been fascinated with data visualisation for a long time. Not only can visualisations summarise complex information very effectively, they also combine statistics with design to make the message more intuitive and therefore easier to grasp. Some even become works of art.*
The Covid-19 pandemic presented a challenge to finally get started with some more involved (from an economist’s perspective) visualisations myself. We have seen a lot of Covid-19 related data visualisations over the last few months. And many of us probably have a rough idea of how many new lab confirmed cases there are in the UK every day. However, as countrywide restrictions ease and as we adjust to a new normal, the risk in my area (and the areas I might travel to) instead of the total across the UK became more relevant to me.
While we start hearing about local outbreaks and lockdowns in the general press, I find that systematic information on new incidences on a local level is still underreported in the UK as of this date. Data issues might be somewhat responsible for this. For example, for a long while, Public Health England (PHE) only published results for an increasingly smaller subset (pillar 1) of total tests. This has been resolved beginning of July and the data now contains results for both PHE/NHS labs (pillar 1) and commercial labs (pillar 2).** Other issues remain. While Scottish data is by test result date, data for England is reported by specimen date. According to PHE it might take up to 5 days (or more) for results to come in. The most recent data for England is therefore likely not complete (and omitted from my maps) and the Scottish data might not depict the most recent activity.*** This restricts the usefulness of these maps somewhat, as outbreaks will only show once they are well advanced (or in decline).**** Also, the data only includes lab-confirmed cases and is therefore heavily dependent on the testing strategy and on whether people are aware of being ill. Finally, not all positive test results can be attributed to an area, and these are omitted as well.
Still, I think visualising the available data in a map gives some indication of the regional activity of the epidemic. In addition to the incidence rate by lower tier local authority (England)/NHS Health Board (Scotland)*****, the visualisation shows the development of new cases over time in those areas which are currently most affected. This provides some information on whether activity is still rising in these areas or not.
While the data for England (and likely also for Scotland) only gives an incomplete picture of new infections in the last 5 days, it still contains some interesting information on new local outbreaks. If the incidence in the last week recorded in the data (which is incomplete) is higher than the one in the previous week (with mostly complete data), this is a sign of activity picking up. This is analysed in the following graph.
It goes without saying that an economist can only say very little if any on epidemiology. My colour and thresholds choices should therefore not be interpreted as led by epidemiological science, as a value judgement (“high” or “too high” incidence) or as entailing any advice on when action needs to be taken. They rather show “buckets” with comparable incidence. Each new threshold is roughly double the size as the next.
* David McCandless entertaining TED talk contains some very good visualisations: https://www.ted.com/talks/david_mccandless_the_beauty_of_data_visualization?referrer=playlist-art_from_data&language=en#t-519423 as does the classic, by the international health expert Hans Rosling: https://www.ted.com/talks/hans_rosling_the_best_stats_you_ve_ever_seen.
** Note that PHE has been mapping weekly incidence rates for English upper tier local authorities since end of May in their weekly surveillance report (https://www.gov.uk/government/news/weekly-covid-19-surveillance-report-published). Since the beginning of June, this map also includes pillar 2 test results. However, the data is only reported weekly and does not include Scotland, Wales or Northern Ireland. It is also not at the lowest area level available (lower tier local authority for England) but at one level up.
*** See https://coronavirus.data.gov.uk/about.
**** Due to the reporting delays, not too much should be read into whether an area is in one colour category on the map or the next one up or down, in particular in areas with low incidence or in areas with low population.
***** So far, I could not find equivalent datasets for Northern Ireland and Wales. However, both Northern Ireland (https://app.powerbi.com/view?r=eyJrIjoiZGYxNjYzNmUtOTlmZS00ODAxLWE1YTEtMjA0NjZhMzlmN2JmIiwidCI6IjljOWEzMGRlLWQ4ZDctNGFhNC05NjAwLTRiZTc2MjVmZjZjNSIsImMiOjh9 ) and Wales (https://public.tableau.com/profile/public.health.wales.health.protection#!/vizhome/RapidCOVID-19virology-Public/Headlinesummary) have very useful dashboards which allow looking at various cuts of their data.