COVID-19: As numbers reduce – what now?

A little history first

I’ve been illustrating the impact of COVID-19 for about 4 months now – and the situation has changed quite a bit.

At the start of the outbreak I was measuring cases – and these were growing exponentially. Not long after, rapidly growing death numbers followed a similar pattern – it didn’t look good.

But things slowed – and numbers came down. We went from over 1000 cases per day on the island in April to under 10 at the end of June. But what’s next?

Well, no one knows for sure, but it seems like the virus will probably be a big concern for the remainder of 2020 at the very least. With testing at higher levels, and track & trace being used it looks like any outbreaks will be hopefully remain localised, and so monitoring this on a geographical level has never been more important.

How do we tell the story now?

I’m concentrating on new cases. It’s the earliest sign of a problem and with testing rates high, and numbers available per county/council it’s probably the best indicator we have.

I’ve created a view of what’s happening on the island that I’ve been updating this for a few weeks now:

Click to enlarge

I like the dramatic red. Attention grabbing and highlights the areas of concern straight away. But it doesn’t really give you much of a history – it only tells you when the last case was.

So here’s a heat map that gives a more comprehensive view of the last few weeks:

Click to enlarge

I’ve also created a similar chart – although visually quite different.

Click to enlarge

I really like the look of this – in this case ‘flat-lining’ is very much a good thing.

So now we’ve lots of detail – but admittedly, it’s a bit hard to look at this and be sure if things are getting better or worse. So, let’s measure how many counties are staying covid-free for at least a week. And while we’re at it, let’s measure 3+ days and 1+ days:

Click to enlarge

It’s clear from this that through June things were generally getting better – quite similar patterns North & South.

However, since the end of June things have not been so clear, and in many cases have been getting worse. The chart above allows us to see this at a glance, and I think it’s a really useful indicator because if there are any localised outbreaks within a county they’ll only get counted once.

But… that being said – I think it’s useful to keep an eye on what’s happening overall.

Click to enlarge

It’s clear from this that at an overall level things are no-longer improving. The charts at the bottom are green for the first couple of weeks, and then they go red.

The good news is that the number of cases are low, testing is high, and track & trace is in place – so even if we now have the same number of daily cases we were seeing in March – we’re in a much better position to handle it.

But, of course, in March we were going into lockdown – and now were coming out of it – so there’s a lot of unknowns over the coming weeks.


—————-   Notes  —————–


As with most datasets – there are problems. Here’s a few:

Denotifications / Adjustments – This data is published as cumulative numbers – we don’t get to see all of the dynamics happening under the bonnet. So, for example, a case could be reported in Sligo, but an old case could be denotified on the same day – and we therefore won’t know anything about it. It also means that you can end up with minus numbers, and it all get’s a bit messy. 

New methodologies – in NI there were 2 pillars – and when the 2nd pillar was introduced this was very hard to keep track of. To make things worse there was a deduping exercise carried out a week later. I managed to piece this together in the end – but it took 2 weeks before I could be sure the numbers were correct. In ROI something similar happened with the German testing numbers.

Data delays – Sometimes numbers are not published (e.g. weekends), and there’s a lag in getting the ROI numbers at county level. It means that you could have a really bad day, but the numbers at county level are quite low.

So, I’ve had to make a few compromises and assumptions with this data, but I think I’ve struck a fair compromise by trying to reflect things accurately, while also presenting this in a simple clean manner.