What's the benefit of diabetes screening?

Asymptomatic guy, obese, no high blood pressure. Do you screen him for diabetes? The USPSTF says the evidence is Incomplete. "Would a hemoglobin A1C [diabetes test] change your management?" I ask. Always my first question - I'm a skeptic to a fault. "Sure," comes back the answer. "If it was 8, you'd start metformin, right?"

Well, maybe. But that's the problem of the screen. If their number is 8, we put them in the Diabetes box. Then we "know" that we need to get their A1C at 7 . . .

But why do we know that? The evidence isn't so great that 8, say, is all that much worse than 7 with regard to clinical outcomes in an asymptomatic patient without evidence of micro- or macrovascular disease. Yes, if the number were 9, 10, 11, 12, then the answer becomes more and more definite, but you're going to start seeing symptoms somewhere in that range anyway.

[links to come, I hope]

A case of ... what?

Recently I was staying with relatives, which gave me the chance to read the New York Times in print. It felt old-timey. I chanced upon an article in Lisa Sanders's Cases series, whose tropes can be summarized as follows:

1. Woman faints.

2. The doctors can't figure out what's wrong with her.

3. Bad Doctor says it's all in her head:

A neurologist in New York carefully examined her and her now thick chart and pronounced definitively that there was nothing wrong with her and that she should try to relax and maybe take up yoga.

4. Good Doctor notices a few key features and makes the diagnosis:

Ledereich watched as the patient calmly sat up. “I know what you’ve got!” he told her excitedly. Her sudden collapse looked as if a switch had been thrown and all her muscles just turned off. Ledereich realized that although it looked like syncope, it wasn’t; she hadn’t actually lost consciousness. What she probably had, Ledereich told her, was something called cataplexy, and that meant that she also had narcolepsy.

So far so good. But the treatment didn't cure the attacks:

But for reasons that neither the patient nor her doctors understand, after about six weeks, [the fainting spells] returned. At first, just occasionally. Then almost daily.

Thereafter she is left to do (more or less) what the Bad Doctor suggested: integrate her new diagnosis into her life.

The patient has learned to cope with her unusual condition; she no longer drives. And when she feels the warning signs, she tries to alert those around her to tell them not to worry. She’s part of a small community, andby now, most know her well enough not to call 911.

There are implications left unexplored here. First: that diagnoses can be partially but not entirely therapeutic. As Up To Date says about cataplexy, "these symptoms are often improved by medications." Often, but not always.

Second, that so much hinges on how the diagnosis is conveyed. Bad Doctor indicated that the woman affected with cataplexy "should just relax" - an abrupt and unhelpful direction, but not, for all that, unfounded. There is a connection between anxiety and cataplexy (and other sleep disorders) remarked upon in the literature.

Finally, a question is left unanswered (and unasked) at the end of the piece. What does the patient know that she has? Does she identify with her diagnosis of cataplexy in a way in which she wouldn't identify with a diagnosis of anxiety or other psychiatric disorder? Does the partial failure of GBH to treat her cataplexy at all detract from her trust/confidence in the diagnosis? In short, what does the patient think of all this?

I don't know why I'm in the hospital!

Continuing to tweak my thoughts about doctor-patient diagnostic discordance (i.e. doctor says: she's in the hospital for X; patient says: my doctor told me I'm in the hospital for Y), this time for the International Conference on Communication in Healthcare to be held Oct 4-7 in Miami Beach. I will be heading to the sukkah, not to the surf (I'm arriving 10/5 if anyone wants to look me up), but I am looking forward to it nonetheless.

I'm also working on an application for an AHRQ grant on the same topic so I can reproduce my Bellevue pilot study at the Johns Hopkins Hospital.

What are we to learn at the bedside? A re-examination of Verghese's essay "Culture Shock"

Abraham Verghese's essay "Culture Shock" (pdf) made a lasting impression when I came upon it last night during a quiet period at work. He writes beautifully about the real patient, with all his spots and signs - as opposed to what Verghese calls the "iPatient," the simulacrum found inside the electronic medical record but nowhere else.

I do realize that we residents, no matter where we train, hone our skills on the iPatient's indices ("The iPatient's blood counts and emanations are tracked and trended like a Dow Jones index, and pop-up flags remind caregivers to feed or bleed") while getting ever farther away from the bedside physical exam done on the real patient. This article makes as powerful a case as any I've read for the re-centering and re-honing of my skills, and it comes at just the right time, when I have the chance to make a transition to be the kind of doctor I want to be. 

But Verghese is confused in his defense of the physical exam - he doesn't know what rationale he wants to focus on, or how he feels about physical diagnosis as justified (or questioned) by evidence-based medicine. Here he is in one place:

If one eschews the skilled and repeated examination of the real patient, then simpl diagnoses and new developments are overlooked, while tests, consultations, and procedures that might not be needed are ordered.

This is the argument from efficiency, or maybe from diagnostic rigor - exceeded somewhat by Verghese's clear affection for the physical exam as a pedagogic and maybe, even, an esthetic, cultural, and moral tool. But it's not clear whether Verghese believes that the physical exam does improve diagnosis or efficiency. Just paragraphs later, Verghese takes another turn:

Younger physicians often argue that physical signs lack an "evidence base." Clearly some signs are helpful, some are not, and we need continued study in this area. But recognizing erythema nodosum or decreased breath sounds and dullness over a larg pleural effusion is worthwhile in and of itself.

The physical exam's actual use in diagnosis is again feinted toward, but without making a real case one way or the other ("we need continued study in this area," the academic physician's classic copout that I know I use at the end of every article I write).

Near the end of the essay, and most confusingly, Verghese takes yet another tack when describing with understandable pride the teaching of the bedside physical exam he coordinates with his chief residents:

We teach that physical findings should be considered biomarkers, phenotypic markers. ... An enlarged spleen, Roth's spots, a Virchow's node, and jugular venous distention are all biomarkers that should be factored in with the high calcium level, the abnormal MRI, and other data to arrive at a true picture of the patient. Failure to recognize these biomarkers is an oversight akin to not seeing a key laboratory value in the chart.

But this comparison is double-edged. If a finding on the physical exam is like a biomarker, then it is like any other diagnostic test, which can be ignored, re-interpreted, or even not tested at all based on the prior probabilities the physician approaches the patient with. Perhaps - given the patient - I might prefer the information given by X-ray to my own physical exam. Or maybe, given the vagaries of varying echocardiography reads, I might privilege my own cardiac exam.

This is why I find Verghese's essay, though moving and personally challenging to my own too-ingrained love of EHRs, to be ultimately unsatisfying. If medicine is a culture, it changes. If the physical exam is to be a practical part of the diagnostic art, and not a relic, it too must change. Let's find out which parts of the diagnostic exam work, and why. We know that no physician does the "head-to-toe exam" for more than a fraction of his or her patients, so which parts should be done when? When is it useful to look for Roth's spots as a diagnostic adjunct rather than as a fascinating bedside pedagogical tool of limited clinical import? (Probably rarely.) Given our limited time with patients, should we not build rapport and understanding by asking more detailed histories at the bedside, rather than indulging in percussion of parts which have no diagnosis to yield up? 

I take Verghese's wisdom and his eloquence but I look for rigor elsewhere, trying to spend my time with the patient in ways that build our therapeutic relationship and find a true diagnosis efficiently.

She doesn't know why she's in the hospital?

I'm still thinking about an all-too-common hospital situation: doctor and patient don't agree on the reason why the doctors put the patient in the hospital. I'm giving a revised talk about it on Wednesday to my primary care colleagues. Comments welcome!

The only skill that matters in treating a patient?

The ubiquity of UpToDate is not without its troubling features (although, to be fair, most UpToDate articles include more references to evidence-based medicine than old-school textbooks - or old-school colleagues - ever entertain). So Darshak Sanghavi's article in Slate serves a useful purpose.

But he really didn't mean to write this, did he?

However, the sheer abundance [of knowledge taught in medical school] crowds out an important—in fact, the only—skill that matters in treating a patient: how to critically appraise published clinical trials.

The only skill that matters? How about:

1. talking to the patient (not trivial!)
2. eliciting the patient's wishes and preferences
3. diagnosis (including the use of diagnostic tests)
4. elucidating the treatment options and formulating a question
5. judging what the best evidence is for the particular clinical question
6. applying the best evidence to the clinical question
7. discussing treatment options with the patient
8. ensuring patient compliance/adherence/agreement/cooperation/investment

Pediatrics (the author's specialty) can't be that different!

Why diagnose?

Just Google.