That headline might seem inflammatory, but it is actually a conservative estimate based on the current numbers. Let's take a look.

The CDC is currently reporting 4714 cases of the swine flu in the US and 4 deaths, although a fifth death (the third in Texas) is in the news today. It isn't really enough cases to calculate an accurate fatality rate, but it's what we have so let's use it. All of those deaths involved pre-existing contributing factors, but a lot of regular flu deaths do as well. 5 deaths in 4714 cases gives us 1 case in 943 cases, which is not much different from the 1 in 1000 fatality rate of the typical seasonal flu.

The estimate being bandied about a lot if 36,000 deaths annually from the regular flu, so this should be the same, no? No. The problem is that even if this new flu is no more deadly than your typical flu, it spreads much more easily according to a study done by Imperial College looking at early data from the Mexico breakout. How fast a flu spreads depends largely on how many people a typical patient will infect while contagious. This number is called R₀. In a seasonal flu, this number is typically around 1.3. The Mexico data suggests an R₀ value in the 1.4 to 1.6 range for swine flu. For comparison, the R₀ values for the last three pandemics are estimated to be between 1.4 and 2.0, so if the Imperial numbers are right, we are in pandemic territory.

A WHO report puts it more simply:

H1N1 appears to be more contagious than seasonal influenza. The secondary attack rate of seasonal influenza ranges from 5% to 15%. Current estimates of the secondary attack rate of H1N1 range from 22% to 33%.

Professor Neil Ferguson was the lead investigator on the Imperial College report and he provided a useful summary of the report's findings in a BBC interview:

This virus really does have full pandemic potential. It is likely to spread around the world in the next six to nine months and when it does so it will affect about one-third of the world's population.

To put that into context, normal seasonal flu every year probably affects around 10% of the world's population every year, so we are heading for a flu season which is perhaps three times worse than usual – not allowing for whether this virus is more severe than normal seasonal flu viruses.

Three times worse. This means that if the typical flu does kill 36,000 Americans each year, 108,000 for this year would not be an unreasonable estimate. So far, so bad.

Unfortunately, you can't just multiply population by infection rate and fatality rate and get a good estimate. There is also the question of how well the health care community can respond. As WHO puts it:

Finally, the quality of health services influences the impact of any pandemic. The same virus that causes only mild symptoms in countries with strong health systems can be devastating in other countries where health systems are weak, supplies of medicines, including antibiotics, are limited or frequently interrupted, and hospitals are crowded, poorly equipped, and under-staffed.

Well, we're in the good old USA, right? We'll be fine!

Not so fast.

In the typical flu season, 5-20% of the population gets sick. Let's call it 10%, so that's about 30 million people. The CDC estimates that there are about 200,000 hospitalizations each year from flu-related problems. So, that means about 6.6 hospitalization per 1000 cases. And our health care system handles that just fine.

Now let's look at this flu so far. 4714 cases in the country and 173 hospitalizations. There's the problem. This suggests 36.7 hospitalizations per 1000 cases, about 6 times that of a normal flu. And let's not forget that we're looking at 3 times as many cases as well. Hospitals could be looking at 3.6 million admissions. Unfortunately, according the Wall Street Journal there are "fewer than one million hospital beds in the U.S."

Now, the swine flu hospitalizations won't be all at the same time, so we need to figure out what all this really means. A study in JAMA gives a median of 3-5 days per hospitalization for people aged 5-49, depending on complaint. Let's say it is a median of 4 days for all swine flu patients. 3.6 million admissions multiplied by 4 days means 14.4 million bed-nights taken by swine flu patients. If there are 1 million beds, there are 365 million bed-nights available, so the swine flu would only take about 4% of the capacity. That is assuming an even distribution, although it is more likely to be a fairly noisy normal distribution. A study by Health Affairs in 2001 found that the metropolitan area's hospitals were operating at 68% capacity, averaged over 365 days. Details of the day-to-day variation in this capacity usage were not given in the study, although it seems reasonable to assume some ebb and flow.

This may not be the right way to measure capacity, however. An article published by the American College of Emergency Physicians points out that while this measure (bed-nights over the course of a year) may say everything is fine, it might not be on any given day. In this study, which focused on new Jersey, they found that while the state's capacity as measured annually was fine, the hospitals were at 85% capacity on almost 58% of the days and over 90% on 25% of the days.

Throw a surge of flu cases at those hospitals on a bad day and there may well be a lot of people not getting the care they need. And that may well increase the number of fatalities, even if it doesn't make the virus itself any more deadly. I wish I could find some more numbers to get a better feel for this variable and its effect; thus far, this is all I have. That said, it is enough to support the idea that that the 108,000 number is likely to be an underestimation.

I may write more about this as more numbers become available, but without a major change in the data about transmissibility, fatality, or care capacity, I stand by the headline. Between May 2009 and May 2010, I expect 100,000 Americans to die from swine flu, and maybe 2 million worldwide. Maybe it will take longer. Maybe it will die out, although that seems unlikely at this point.

I hope I am wrong.