Wednesday, August 06, 2008

Separating clinicians from automatons: the long tail

DB has written a series of posts on this concept (here, here, here, here, here and here). There’s little I can add but I wanted to get them into my own links, and going through the exercise below has helped my understanding of the concept.

The long tail refers to the horizontal tail of a power law graph. This is a polynomial function which has garnered great interest because of its applicability to a large number of phenomena in commerce and nature. When applied to differential diagnosis the horizontal axis can be visualized as an array of diseases of increasing rarity the farther away from the origin. The vertical axis represents the probability of the given disease explaining the patient’s presentation. A small number of common diseases are clustered toward the origin (to the left). Toward the right (in the long tail) are uncommon diseases, becoming increasingly rare as the graph approaches the horizontal axis as an asymptote.

Individual diseases in the long tail are uncommon. But, because the tail is long (there are many rare diseases), in the aggregate a significant number of patients is represented. That principle, a challenge for the clinician, is explained here.

DB uses the sore throat as an illustration. (If you think sore throats are boring read his posts and listen to his recent Grand Rounds presentation at UAB---scroll to July 30). Pneumonia is another example. Pneumococcal pneumonia would be near the origin. Blastomycosis, ANCA associated pulmonary capillaritis and bronchoalveolar carcinoma, diseases which can present as “pneumonia”, are in the long tail. For recurrent abdominal pain irritable bowel syndrome would belong on the left, with celiac disease and acute intermittent porphyria occupying positions progressively to the right. And so on.

The challenge of the long tail is knowing when to enter it and, once you do, to generate a wide enough differential diagnosis to encompass the disease the patient has, and finally to select appropriate tests to pinpoint the diagnosis. That’s why the long tail separates clinicians from automatons. Algorithms and guidelines won’t help. Up to Date may not even help! What’s needed is judgment along with a vast fund of knowledge about diseases. Key to knowing when and when not to enter the long tail, as DB explains, is knowledge of natural history. How long, for example, should it take your patient with pneumonia to get better? At what point, as a corollary, should you start searching for another diagnosis?

The description of this cognitive process and the contrast between clinician and automaton makes a compelling case for the revival of the original concept of the internist.

Image by Hay Kranen. Source: Wikipedia.


Michel Accad, M.D. said...

On reading this post I immediately thought of Herb Fred, the author of the Texas Heart Journal editorial about which you and others have previously commented in March 2006. In the early 1990's I had the distinct privilege of working as a resident and chief medical resident under Dr. Fred, a true giant of medicine and an unsurpassed medical educator. One of his best crafted teaching tool was a method he devised to ensure that the "long tail" is captured each and every time. He published on this method in 2003 (The Morning Report Card. Res Staf Phys, 49:22-26 (September) 2003; unfortunately not available online).

After obtaining the history, performing the physical, and getting the routine labs, the method calls for a formulation of the "clinical problem" in 5 words of less. Those 5 words have to convey what is most specifically known about the case but absolutely avoid the speculative. The diagnostic work that follows entails listing the DDx beginning with the most likely disease, and then ordering the more sophisticated tests as necessary to nail the diagnosis, if possible. Short of a positive gold standard of pathology, the diagnosis should always remain provisional or "probable."

For example, the clinical problem for a patient presenting with rapid onset of fever, green sputum, and a lower lobe infiltrate on a CXR should never be "pneumonia" but instead "lobar infiltrate and fever" since there are probably 14 non-infectious possibilities than can present in this fashion. Of course, most of the time a "probable" diagnosis is sufficient to initiate treatment, but at least the formulation of the clinical problem avoids tunnel vision, especially if the patient fails to respond appropriately or follows an unexpected course.

Needless to say, in a DRG-driven world of "efficiency," there is little concern for diagnostic accuracy or circumspection. Nevertheless, the lesson for me has been invaluable, even though I frequently fall in the trap of "premature closure."

Zo said...

I expect this deeply interesting post speaks to the lack of diagnosis for myself and those in my extended family with active Hereditary Coproporphyria (or some mutation thereof.)

I am a member of an online porphyria group, and as a group, "porphs" are met with rejection or downrigh hostility, more often than not, told, for example, "You can't have that!"

As I say, interesting.