Interactive data: The Dartmouth Atlas of Health Care points the way

Today’s New York Times has a story on regional variation in the availability and cost of health care. The story is accompanied by a “multimedia interactive graphic” — that is, a Flash visualization that charts the following variables on a U.S. map:

  • Reimbursements
    • Total
    • Acute care
    • Outpatient
    • Surgical
  • Surgery Rates
    • All
    • Heart bypass
    • Knee replacement
    • Mastectomy
  • Enrollees

For each mapped variable, mousing over the displayed hospital referral regions yields the local, state, and national values for that variable.

It’s nicely done. There’s no question that, as of mid-2007, this is cutting-edge data interactivity for the mainstream. But times are changing fast. The Times sourced this data from the Dartmouth Atlas of Health Care. It took me five minutes to download the surgical data, upload it to Dabble DB, and publish a similar map along with a complete tabular dump.

Of course I cheated by aggregating only to the state level, which Dabble DB can do easily, rather than to the level of hospital referral regions. And I left out the national averages. But still, it’s striking to see what can be accomplished in a few minutes with no programming.

It’s even more striking to see what you can do directly on the Dartmouth site. Suppose you were thinking about having a knee replacement done in Keene, NH, and you wondered how many of these procedures are done at the Keene hospital, at other hospitals around the state, and in Boston and New York. Here’s the answer:

Knee Replacement per 1,000 Medicare Enrollees
HSA Level Rates (2003)
Area Population Rates Ratio to
*Keene , NH 8,047 5.07
Laconia , NH 8,116 8.60 1.70 29
Concord , NH 13,428 8.49 1.68 46
National Average 28,767,985 6.88 1.36 52,049
Lebanon , NH 9,483 6.73 1.33 16
State: New Hampshire 148,431 6.54 1.29 218
Manchester , NH 20,748 6.32 1.25 26
Rochester , NH 5,806 6.25 1.23 7
Nashua , NH 16,706 6.16 1.22 18
Dover , NH 8,462 5.16 1.02 1
Exeter , NH 10,624 4.96 0.98 -1
Boston , MA 67,651 4.90 0.97 -11
Manhattan , NY 166,112 3.32 0.66 -290

Wow! Hats off to the team at the Dartmouth Atlas of Health Care. This is the sort of thing that will change expectations about what interactive data ought to mean.

Posted in .

7 thoughts on “Interactive data: The Dartmouth Atlas of Health Care points the way

  1. Great stuff, Jon. The emergence of these sorts of tools for adhoc and on-the-fly data analysis is one of the unrealized consequences of “Web 2.0”. Too many people see the specific applications, and don’t realize that the process can be generalized. Admittedly, it’s still far easier for those of us who can hack on code, but Dabble, Pipes, Google Spreadsheets, etc., etc. are starting to change the nature of analysis.

  2. Great post, showing how “good enough” can get you online quickly.

    Here’s another quick example – more like two hours instead of five minutes, but that included registering, and tracking down the data.

    Right now it’s primitive: meaningless totals, no control over the legend, colors, no ability to handle anything other than numeric data …. but it’s good enough for a first effort.

  3. Thanks so much! The first thing I did was search out information on the three people credited for the NYTimes infographic; Erin Aigner, Matthew Bloch, and Vu Nguyen. I learned that:

    – ArcInfo is *the* system for creating data maps (there is also an ArcViewer)

    – Ms. Aigner is very accomplished in the field of GIS and has a great interview over at

    – The School of Journalism and Mass Communication at the University of North Carolina at Chapel Hill is turning out some award winning students

    – The journalism of data has some surprising challenges, such as the fact that after you finish a data display, you are often obligated to destroy/discard the data! That means no libraries for future mash-ups or reference. Many other are revealed in Ms. Aigner’s interview.

    – Programmer/Cartographer Matthew Bloch has a great development site!

    Thanks for that key to this treasure trove!

  4. I’m one of the people (along with Vu Nguyen and Erin Aigner) who worked on the NYTimes interactive graphic that John mentions in his post. Generalizability is also important to us, as there often simply isn’t time to put together an interactive graphic from scratch. For this piece, we used a new web mapping framework that we’re developing in-house.

    Re: the previous comment, ArcInfo is an essential GIS tool, but of limited use for creating interactive web maps. The link you posted was to a free online cartographic tool called MapShaper that I created last year when I was a grad student at the University of Wisconsin, Madison. Feel free to try it out :)

  5. “For this piece, we used a new web mapping framework that we’re developing in-house.”

    You’re working for the NYTimes, then? Excellent!

    Do you think we’ll see the definition of interactivity expand to include the kinds of social processes being enabled by Swivel, Dabble DB, Many Eyes, etc?

Leave a Reply