I’ve been spending the week at the Infectious Disease Modeling Summer School here at UW. It’s very interesting, and good for me to learn more about how people in my new field think (especially people in my new field, outside of my little institute…)
I’ve discovered a pet peeve during this week of presentations, though. I’ve seen a lot of numerical examples where the numbers work out perfectly… a little too perfectly. If you split 1000 people into an experimental and control group by choosing a random subset of 500, fine. But if you look within that group to see how many have a trait that occurs independently with probability 0.2, you do not often find exactly 100 in group A and 100 in group B. I think a little more complexity in the numbers makes the example easier to understand.
I’m sure that you, my loyal reader, can generate random numbers from a multitude of distributions, if you wanted to spend the time. But if you’re busy, busy, busy, then you can have wolfram alpha do all the work. It actually comes through for that one: “sample Binomial(500, .2)“.
I wrote two months ago about the mysterious differences in health care costs that I found so intriguing in a talk by Jonathan Skinner. (That was two months ago? Really?) Since then, the surgeon/author Atul Gawande has brought the mystery to the national stage. In a long story for the New Yorker, he gave the non-technical version of Skinner’s talk, and today he addressed some of the feedback that this article has received over the last month.
His short answer to the mystery is this:
Analysis of Medicare data by the Dartmouth Atlas project shows the difference is due to marked differences in the amount of care ordered for patients—patients in McAllen receive vastly more diagnostic tests, hospital admissions, operations, specialist visits, and home nursing care than in El Paso.
But that is not the end of the story. It only takes a sentence to explain the “proximal” cause of these cost differences, but it takes the whole article for Gawande to do justice to his theory on the underlying cause, and his is certainly not the only theory.
Since his theory of the root cause of this inequality is centered on physicians putting profit over patients, it has made some doctors uneasy. Greg Roth, a physician that I work with hadn’t had time to read the article when we last chatted, but he did attend Skinner’s talk with me two months ago. Greg told me about a detail that has emerged as doctors put Gawande’s article under their microscopes: we might be making a mountain out of molehill-sized mystery.
Investigative reporter have to get the story, and raking the muck way out in the tail of this distribution turned out to be a good bet this time. But McAllen is 6 standard deviations above the mean (not to imply that this distribution is normal… should it be?) How much impact would it have, for the whole population, if the outliers were greatly improved?
If through anti-fraud policing, better culture, and general hard work, the top 10% of hospitals reduced their cost per patient to the national average, that would reduce the average cost by 3.6%. Outliers show what is possible, but making a big change involves more than outliers.
The political situation in Iran has been in the news and on the nets a lot this week. I hope that the friends and families of all my Iranian colleagues are safe. I’m thinking of you.
I haven’t had time to write anything this week because I am up to my neck in this Seven-Samurai-style software engineering project. You know, where a bunch of untrained villagers (that’s me) need to defend themselves against marauding bandits (that’s the Global Burden of Disease 2005 Study), so they have to learn everything about being a samurai (that’s writing an actual application that people other than this one villager can use) as quickly as possible.
I guess this analogy is stretching so thin that you could chop it with Toshirō Mifune’s wooden sword. But, if anyone knows how a mild-mannered theoretical computer scientist can get a web-app built in two weeks, holler. If you prefer to explain in terms of wild-west gunslingers, that is fine.
Here’s my game plan so far: I’m going to make the lightest of light-weight Python/Django apps to hold all the Global Disease Data, and then try to get my epidemologist doctors to interact with it on the command-line via an interactive python session.
The rest of this post is basically a repeat of the Django tutorial, but specialized for building a data server for global population data. As far as interesting theoretical math stuff, hidden somewhere towards the end, I’ll do some interpolation with PyMC’s Gaussian Processes using the exotic (to me) Matérn covariance function. Keep reading →
I was reluctant to enter this media frenzy about H1N1 flu (or whatever we end up calling it…), but only 8% of telephone respondents are “not concerned at all” about these events, so I thought I’d say something more than nothing.
Information technology’s main contribution so far has been the rapid spread of misinformation: for example, eating pork is no less safe than usual, despite rumors to the contrary twittering around the globe.
But there is an opportunity for IT to shine a little bit, too. I’m optimistic about Ushihidi’s web2.0 approach to “crowdsourcing crisis information”. Definitely something I can spend too much time looking at.
Here’s a half-baked post that I started months ago. I decided to rush it to press for Earth Day, which is today.
The first U.S. auction for carbon emission pollution rights occurred in December of 2008. It raised over $38.5B, which will go to six states in New England. From ScienceNOW Daily News:
The auction’s premise is that putting a price tag on pollution–so-called carbon trading–will eventually reduce emissions industrywide. Companies must pay for the right to emit greenhouse gas emissions and are penalized for excess pollution.
RGGI states, picture
The ten states shown in dark green are participating in RGGI. Observers are represented in lime green.
How did the auction work? online, reserve price, open to investors and environmental groups, required for power companies in RGGI states. Not required for manufacturing or transportation. Any earth-day-interested readers out there to fill in these details? Or, to do a little follow up research about how things have gone? (I wrote this last December.)
Finally, here is a humorous critique of carbon trading, based on the observation that carbon credits are a scarce resource. This is highlighted by a paired example from cheatneutral. I find it compelling.
Do you remember last summer’s health scare around the housing market collapse? There was a theory that all the swimming pools in all the foreclosed houses in California would become major mosquito breeding grounds, leading to major crops of mosquitoes, leading to West Nile virus or maybe even the reintroduction of malaria in the US.
There have been some fun ideas for tackling this potential problem, like filling the foreclosed pools with exotic fish. But I woke up today to learn about my new all-time favorite approach: let skateboarder to drain the pools and skate in them. (thx @omarkhalifa)
Bonus points opportunity for my influential readers: WSJ reports that local disease control agencies are doing aerial surveillance for abandoned pools. Can you convince them to release their aerial photos of abandoned pool locations to the local skaters?
Health Economist Jonathan Skinner gave a talk at IHME about a week and a half ago. He told us about his work on the Dartmouth Atlas of Healthcare, and showed us some of the numbers he’s crunched on the variation of Medicare costs by region. He has found this mysterious, 2.5x variation between the cost of care between expensive regions (like Miami) and inexpensive regions (like Seattle). It seems like a great mystery, and I’ve been puzzling over it for a week now. Any theories? I’m partial to network effects.
I guess I’m one to follow the latest fads. I have a blog, right?
I held off even considering “Twitter” for a long time, however. Who cares what I’m doing, right now, in 140 characters?
But that’s not actually what twitter is about (at least its not all that twitter is about). It’s more like having an IRC chat room, but in a public park. But the fauna is synthetic.
Anyway, I’m giving it a try. You can see how it’s going for me here.
My ego does depend a little bit on how many “followers” I have, but I’ve got practice dealing with this. When I was a college radio deejay, I usually had no idea if anyone was listening out there in radio land, so I’d put on my most depressive college radio voice, get on the mic and ask for callers. Then while I waited to see if anyone would call in, I’d dedicate this song to myself:
I’ve been spending the week at the Infectious Disease Modeling Summer School here at UW. It’s very interesting, and good for me to learn more about how people in my new field think (especially people in my new field, outside of my little institute…)
