Proving Anything With Numbers

 

 

# 3307

 

 

Just about everyone, it seems, wants to take the numbers we get on confirmed H1N1 cases, hospitalizations, and fatalities in order to somehow quantify the severity of this novel influenza virus.

 

Depending on what numbers you decide to use, you can generate a wide spectrum of results.

 

Yesterday, in Premature Extrapolations, I highlighted one newspaper columnists opinion that seasonal flu kills 250 times more people than swine flu based on 2,000 NYC estimated deaths each year and only 8 swine flu deaths to date.    

 

Today we’ll attempt to play with several  larger sets of numbers, and demonstrate why these calculations aren’t terribly reliable.

 

In their 45th update on the H1N1 global situation, the World Health Organization today announced that there have been 25,288 confirmed cases of the virus globally and 139 deaths.

 

One of the things we want to figure out is the virulence of this virus. 

 

Just how deadly is it?  

 

Seasonal flu is thought to kill about 1 person out of every 1,000 who contract it (.1%).   So we naturally want to see how this new virus compares.

 

Using the `official numbers’ from the WHO,  we can divide the total number of cases by deaths (25,288/139) and get 1 death out of every 181 cases.

Or about 5 times higher than seasonal flu.

 

But . . . is this estimate reasonable?

 

It certainly sounds reasonable.  We are, after all, using the official WHO numbers. 

 

But I can hear many of you saying that this incorporates the early high fatality numbers of Mexico, and that those numbers skew the results for the rest of the world.

 

In Mexico, the `official numbers’ are 5717 cases and 106 deaths (5717/106=53).   This suggests a CFR (Case fatality ratio) of nearly 2%. 

 

If you remove Mexico from the mix, you have 19571 cases and only 33 deaths.  Or about 1 death in every 593 cases (not quite double the CFR of seasonal flu).

 

So . . .is that our answer?

Well . . . maybe.   But the United States shows 13217 confirmed cases and 27 deaths.  And that gives one death out of every 489 cases. 

 

It becomes more complicated when you add in the deaths reported, but not yet added to the US tally. 

 

That number (last I checked) was 37 deaths.

 


So . . .  13217/37= 357.    

 

Or about 3.5 times deadlier than seasonal flu.

 

But of course, this doesn’t take into account the untested and unconfirmed cases, now does it?

 

Here in the United States the CDC has acknowledged that the actual number of cases is probably 20 times higher than the official count.   Most states are only testing those hospitalized or otherwise seriously ill with influenza-like illnesses.

 

So . . . can we safely divide the mortality rate of .35% by 20? 

 

If so, then the true CFR of H1N1 would be about 1/6th that of seasonal flu. 

 

It would therefore be a gift from god if H1N1 supplanted the other seasonal viruses, because it would cut our yearly death toll from 36,000 a year to only about 6,000 a year!

 

This doesn’t seem terribly likely.  So we must be missing something.

 

The problem here is, every day in the United States more than 6,000 people die.  That’s about 180,000 deaths a month.  Most are never tested for the H1N1 virus (most, as in 99%). 

 

It is not only possible, but it is likely, that we are missing some H1N1 related deaths here.   How many?  

 

We don’t know. 

 

Just like we don’t know how many people actually die as a result of seasonal influenza each year. 

 

Doctors routinely sign death certificates for their patients as long as they had some underlying condition (or were of an age) where death wasn’t totally unexpected. 

 

Quite often the cause of death put on the death certificate is something generic, based on the patient’s history of chronic illness, not on a post-mortem exam.

 

Unless someone dies in the hospital while being treated for influenza and/or pneumonia, there is a good chance that influenza will not be directly blamed for their death. 

 

And their death won’t end up in the official statistics.

 

That is why, instead of counting the number of death certificates listing influenza/pneumonia as the cause of death each year, the CDC relies on mathematical models and estimates.

 

It is probably fair to say, that even with these gaps in our surveillance, that the United States does a better job of monitoring the causes of deaths than does 90% of the rest of the world. 

 

So if we are missing cases, other nations are missing cases too.

 

 

So what is the answer?   How virulent is this novel H1N1 virus?

Based on the few examples above, we can show this virus as being anywhere from 1/6th the virulence of seasonal flu to 5 times deadlier.

 

And if we wanted to skew the numbers even more, we could take the Australian experience, where they’ve had more than 1,200 cases and NO Fatalities!

 

Based on that sizeable dataset, you could argue that H1N1 flu isn’t deadly at all.   

 

But we know that isn’t true.

 

There’s a reason why the CDC, the WHO, and virologists around the world have been slow to put a number on the virulence of this virus.   

 

Given the vagaries of global testing, surveillance, and reporting, it simply isn’t easy to calculate.

 

Professor Neil Ferguson, who uses considerably more sophisticated methods than I’ve demonstrated today, published Pandemic Potential of a Strain of Influenza A (H1N1) : Early Findings  last month giving a preliminary CFR of about .4%  or about 4 times greater than seasonal flu.

 

The CDC continues to refer to this virus as being on a par with seasonal flu, or perhaps a bit more virulent.  They admit that small, but important increments in virulence, would be very difficult to detect.  

 

Obviously, I don’t know what the CFR of this virus is.  Nobody does, and it is likely to vary considerably from one population to the next.

 

My sense of it is, based on what we’ve seen thus far, mirrors that of the CDC. 

 

That it is currently comparable – or perhaps a little greater than – seasonal flu.  That could change, of course.

 

That isn’t to say that this virus isn’t a serious threat.

 

As a novel influenza virus, it is expected that it could infect double or even triple the number of people over the next year than we would expect to be stricken by seasonal flu.  

 

So even at the virulence level of seasonal flu, it could have double or triple the impact.  Any increase in virulence would compound the effect.

 

With a health care delivery system plagued by a serious lack of surge capacity, that could prove a real problem this fall.

 

As time goes by we’ll probably get a better handle on all of this, although I expect the number to remain elusive, and vary considerably over time.  

 

While these back-of-the-envelope calculations are interesting,  I don’t recommend putting much faith in them. 

 

 

                *             *             *             *              *

 

Note: Due to a family medical emergency (not me, this time), I will be away from my desk a good deal for the next couple of days.

Related Post:

Widget by [ Iptek-4u ]