Moving average of daily deaths may give a good indicator of the progress of the pandemic

12 08 2009

I’ve been thinking about how we map the progress of the pandemic here. The Department of Health and Ageing’s daily confirmed cases are good – but we all assume they understate the true number of Swine Flu cases out there by 1 to 2 orders of magnitude. Why? Because the overwhelming number of people who get Swine Flu are not being tested and so never show up on the figures.

So, the risk with these figures is that we may get a fall off in daily new cases and therefore think we have hit the peak of the pandemic, when in fact all that has happened is that testing protocols have been changed.

In other words, we need measures that are not so affected by human decisions. One of those measures, I think, is the number of people dying with Swine Flu each day in Australia. That’s a number that is not going to be affected by decisions on testing, I assume. (If you’re that sick, they’ll test you. The only issue here is that in a normal flu season they might not test you, and simply put down your cause of death as pneumonia, but in a pandemic, my guess is that they will test). So, if we see this measure peak and start to fall sustainably, we can be pretty sure we’re over the hump of the pandemic and coming down the other side.

So, I’ve graphed the seven day moving average of daily deaths of people with Swine Flu in Australia. Note, I’ve said deaths 7 day move avg deathsof people with Swine Flu, not because of Swine Flu. These people have died and they had Swine Flu at the time – but that doesn’t tell us that it was Swine Flu that was responsible.

Anyway – here’s the graph. And as you can see, at the moment, on average, 3.7 people are dying each day in Australia with Swine Flu. That’s the highest it’s been since the start of the pandemic here, and it is showing no sign that we have reached the peak yet. Remember that you can click on this graph to blow it up.

I’ve also updated my usual two graphs with the latest figures from the Department of Health and Ageing. Here they are (clicking on these graphs will blow them up):

7 day move avg australiaaust hospitalisations




2 responses

13 08 2009
Barnaby Dawson

You say that “If you’re that sick, they’ll test you.”

I don’t think thats true. There’s too many people dying all the time from causes that might be flu related for them to test even a small fraction of them.

In any case this assumes that those that die will have primarily viral pneumonias. Thats not the case for seasonal flu (in fact such deaths are rare). Unless the death is a viral pneumonia the virus has quite likely left the body by the time any test might be done so a negative test is not good evidence that someone did not die as a result of swine flu.

Some hard numbers. Scientists from Imperial (London) have estimated the CFR as 0.24%, Professor Ferguson as 0.4% other UK scientists as 0.5% and the UK government as 0.1-0.35%.

In the UK only 41 deaths have been detected as of today, whereas many more than 100,000 were believed to have been infected two weeks ago. At a 0.1% CFR rate (the lowest of the credible estimates I’ve seen) we would have expect about 100 deaths by now. So we are missing at least half of the deaths.

If the CFR were actually 0.5% and 200,000 people were infected then we would expect about 1,000 deaths by now and we’d be missing 96% of the deaths.

We just can’t assume that we’re detecting all the deaths when we have a track record of only detecting 5% of them.

In addition deaths aren’t such a great metric as deaths lag infections. So the peak of infections might occur a week or two before a peek in deaths (perhaps longer with modern intensive care). Your hospitalisation graph might be a bit better regarding lags and the statistical power is better too.

13 08 2009

> In other words, we need measures that are not so affected by human decisions.

Yes exactly. That’s why I too look at only two indicators: the number of fatal cases and the ILI-surveillance based charts (percentages of doctor visits with ILI symptoms)

However, when looking at the number of fatal cases as an indicator of pandemic progression, one must keep in mind a considerable time lag, due to:

a) time between onset of disease and death (median of 10 days, according to mexican and US studies)

b) time lag between death and the actual reporting (may be at least another week on average)

In the USA, the number of weekly reported new fatal cases (i.e. a similarily smoothed indicator than your 7-day moving average) is still on the rise (from 50 lats week to 70 deaths this week) despite the fact that the peak of the wave in terms of new infections has already occured 7 weeks (!) ago (according to the FluView ILI charts)

So this could be a very worrying sign that the CFR is increasing dramatically (decreasing number of infections + increasing number of fatalities *after* even ajusting for a very large delay of 7 weeks)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: