Monday, April 27, 2009

Want To Learn ACH? Watch More House! (

I really like the television show House. This surprises me as I am not a big fan of medical mysteries generally and do not find the House character particularly likable. It occurred to me recently, however, that I might like the show more for its method than its characters or story.

The main method used in each of the shows -- differential diagnosis -- bears a remarkable resemblance to one of my favorite intelligence analysis techniques, Analysis of Competing Hypotheses (ACH).

OK, OK, I know this sounds weird, but bear with me for a moment. In the first place, making the connection between medicine and intel is old hat. My colleague Steve Marrin does it all the time.

In the second place, these two methods even sound similar. Differential diagnosis "involves first making a list of possible diagnoses, then attempting to remove diagnoses from the list until at most one diagnosis remains." ACH "requires an analyst to explicitly identify all the reasonable alternatives and have them compete against each other for the analyst's favor, rather than evaluating their plausibility one at a time." Sort of sounds the same...

The true test, I thought, would be to use an ACH matrix to map a differential diagnosis. The medical lingo is beyond me, though, so I figured I would search out someone who had summarized a particularly good House episode and use their abstract.

Enter Scott.

Scott (FNU) is a former Air Force doctor now in general practice in Illinois who summarizes and critiques every episode of House over at his blog, Polite Dissent. Very cool, slightly obsessive and enormously useful for this exercise.

I chose the "Let Them Eat Cake" episode from Season 5 as Scott said the medicine was pretty good (he gave it an "A").

Scott starts his summary like this: "Emmy, a thirty year-old fitness instructor, is filming an infomercial when she experiences sudden difficulty breathing and collapses, breaking her ankle in the fall. She is admitted to House’s service for evaluation and all the initial tests were normal. Taub suspects her of steroid use, Kutner mentions environmental allergies, and Cuddy suspects exercise induced asthma." (Note: Taub and Kutner are both members of Dr. House's team. Cuddy is house's boss)

Mapping this to an ACH Matrix (and I am using the PARC ACH 2.0.3 software here), looks something like this:

I am making some assumptions here, of course. The team considers the asthma hypothesis to be the best but Scott is quiet as to why. I am just guessing that the fact that the initial tests were all normal was inconsistent with the other two hypotheses.

Anyway, Scott continues: "The last (i.e. exercise induced asthma) seems the most likely, so the team sets about to recreate Emmy’s episode, the best they can with her broken ankle. Sure enough, while in the middle of exercising, she once again collapses and is found to be pulseless." Scott doesn't say this explicitly but I have to assume that this extreme reaction is not what they expected and is inconsistent with any of the current hypotheses.

This means it is back to the drawing board for House and the team. Interestingly, an intel analyst in a similar position would do exactly the same thing. Scott continues his narration, "Kutner suggests she may have Carcinoid syndrome." None of the existing evidence rules this out so, Scott continues, "A CT is obtained which shows no carcinoid tumor..." Intel analysts in a similar position would also task additional collection in an attempt to disprove the unclear hypothesis.

The IMINT -- sorry, I mean CT Scan -- reveals something new, though. Emmy has had gastric bypass surgery! This has, in Scott's words, "them rethinking their differential diagnosis: now diabetic neuropathy (nerve damage caused by diabetes) and sleep apnea are added."

Reviewing the bidding after the most recent round shows a matrix that looks something like this:

As the show progresses both the diabetic neuropathy and the sleep apnea diagnoses/hypotheses are eliminated and several more (well, LOTS more) hypotheses are added to the mix. If you are interested in how all this plays itself out, you can read the rest on Polite Dissent along with Scott's deconstruction and evaluation of the episode.

As I thought about other episodes of House, I realized that many of the common problems that ACH deals with repeatedly come up. For example, it is not unusual for the doctors in House to suspect deception (e.g. patients that lie) when it comes to some of the information they collect. Likewise, the story often turns on an assumption concerning a patient that turns out to be false. Deliberately deceptive information is daily fare for the intel analyst and we came up with Structured ACH as a way to get at underlying assumptions.

Once you start to think about this idea, it really starts to grow on you. Next time you watch House, think about ACH. If you come up with anything neat, post it in the comments!


Christophe Deschamps said...

Hi Kristan,

I just found this articles and I told myself that in fact it may be dr House that copied Heuer :-)

Kristan J. Wheaton said...


Many thanks! Very interesting article.