More Flu Data

Statistics as reported in most newspapers never quite satisfy me. It's not that they are lying, but there is never quite enough there. The flu is a great example. I was just reading this great article in Slate about how they came up with the 33,000 deaths per year from the flu http://www.slate.com/id/2218367. Turns out they did regression analysis and looked at the correlation between the severity of the flu season and the number deaths on a week by week basis. From this they estimate that 34,000 people died that would not have died otherwise. This is all well and good, but it doesn't answer the real question. How should the government react? What will the economic impact be? What's the difference between pandemic flu and seasonal flu?

Because, let's face it. The flu doesn't kill anybody. After all, we are all going do die. All the flu does is speed death up for some folks. Some articles make some cursory acknowledgment of this fact by noting whether the flu is killing young people or not. But this there is never a good measure associated with it. Reminds me of a case I was reading at about the time of the peanut contamination stories. A man was talking about how his mother had fought through cancer and was still weak and on a respirator but seemed to be making a comeback when she got sick and died from contaminated peanuts. Sure this is a tragic story and there is a chance that his mother would have had many more years if it hadn't been for the peanuts. But really, cancer killed her. Let's take a look at making these numbers meaningful.

For example, you could do the same regression analysis you did above but then continue the analysis moving forward to see if there is a subsequent dip death rate in the coming months. If the flu was killing people who were already pretty close to death, you might expect to see a decrease in death rate over the next year as those deaths that were 'paid forward' aren't recorded. After some work, rather than recording a 'mortality' rate you could state a rate for 'having more than 6 months taken off your life ' or 'number of years taken from population.'

Another death related statistic I've never liked is life expectancy. I've go "The Economist Pocket World in Figures 2008 edition" on my desk (of course.) I open it up and see that the life expectancy for Japan is 82.5 years. For the United States it's 78.2, just below Cuba and Denmark. But then on another page I can see that the median age in Japan 42.9. The United States doesn't even show up on the list because it's too low. Well, that puts the life expectancy in perspective. It's a lot easier to have a longer life expectancy if you don't have young people around to die and mess up your stats. It also shows how life expectancy in isolation doesn't help much for policy because you don't know who is dying.

For example, I am personally more interested in the life expectancy for a 40 year old male than I am for the entire population. A better chart would be percent chance of dying in a given year as a function of age. There are lots of cool things we could learn about and apply to policy. I would expect a spike in the first few years because of birth defects and childhood illnesses. After that you are almost certainly going to get a spike around 16 when US kids get driver's licenses and some relatively high rates for the next few years due to suicide and teenagers being crazy. Perhaps there would be another spike at 21 for hitting the legal US drinking age. It goes down until thirty and stays low through the peak parenting and earning years. Your just too busy to die. Things pick up in your fifties as bad living habits start to take their toll. I would guess you would see another spike around retirement but I suspect things level off for quite a while until you get truly old. These are all guesses and it would take a lot more data to come up with this sort of chart, but it sure seems more useful and more actionable.

And don't even get me started on dramatic illustrations of the amount of money in the economy especially since the banking crisis. The lesson is that real numbers take work to figure out. Even stats presented in good faith don't tell the real story, so you can be sure that numbers presented to further an agenda they are useless.