June 3, 2016

Diagnosing those predictable images.

It didn’t take long in my residency training program for newbies to learn that there were certain imaging study/clinical history combos that yielded very predictable results. To the point that a cynical mind (insert self-referential comment here) might soon wonder whether actually looking at the images had any more of an impact upon the interpretation than, say, the statistical margin of error.

For instance, a CXR with clinical history of “chest pain” almost invariably turned out to be normal. Aside, perhaps, from findings that neurotic sticklers such as myself insist on putting in reports (slight scoliosis, age-appropriate degeneration of shoulders), which other, surely saner rads can omit without subsequently lying awake at night suffering crises of conscience.

The phenomenon could, no doubt, serve as a springboard for all sorts of philosophical contemplation or discourse. What might it mean, that such diagnostic outliers exist? Are too many chest X-rays being ordered for pain that should be clinically manageable without usage of ionizing radiation? Are the vast majority of chest pain etiologies not realistically detected by CXR? Is there a disproportionate amount of the chest pain population that isn’t “real” (malingering, psychosomatic, chronic, etc.)?

The upshot, in any event, was that the daily workload contained an onslaught of studies with a very low pretest probability of significant abnormality. Asymptomatic positive PPDs were another common example.

“It’s gonna be normal” study/history combos are not the only example of this phenomenon. There are also the less tidy instances where the history and imaging type lets you know that you will be reporting abnormalities…that will absolutely never be pathology the referring clinician expected or cares about. Unlike a quick-n-easy negative chest X-ray, however, they’ll eat up plenty of your time and trouble.[[{"type":"media","view_mode":"media_crop","fid":"49126","attributes":{"alt":"Predicting imaging","class":"media-image media-image-right","id":"media_crop_9849690659422","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"5911","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 166px; width: 170px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"©Danny Smythe/","typeof":"foaf:Image"}}]]

The prime example is the “fever of unknown origin” or “leukocytosis” whole-body scan that didn’t bother utilizing contrast of any type. It is a virtual certainty that these will be chock full of stuff you have to rewindow, view in multiple planes, measure, compare against priors, etc. There will likely as not be horrible patient positioning and motion to make it even more of a chore. And the payoff is that there will either be nothing suggestive of an infectious nidus…or there will be about two dozen possibilities you have to report (possible pneumonia, pyelo, cholecystitis, colitis, cystitis, osteo…you get the idea).

Happily, many of the “it’s gonna be abnormal, but not in any useful way” combos are less chorelike. Right upper quadrant sonos for transaminitis, for instance, never have a smoking gun…but maybe the liver is kinda-sorta big, or a little echogenic. Or at least the sonographer said so (see my recent column on the matter).

I could go on…”altered mental status” head CTs, “dizzy” carotid Dopplers in 20-year-olds, or “R/O appy” sonos in corpulent adults, but for every one I mention you’ll probably already have a couple examples from your own experience in mind.

Do any rads take this ball and run with it, a la John Edward, and fake psychically churn out reports on these cases practically without looking at them because they’re 99% certain what the results will be? Gosh, I hope not.

Do any do the opposite, recognizing that they have developed a bias about these cases, and buckle down to scrutinize the images extra-attentively in compensation? It would be nice to think so, but such saintly individuals would, at some point, realize the paradox of thus giving other cases less than 110%, and their heads might explode while trying to figure out how to reconcile these things.

What do I do? Clearly, I take note of these trends for a decade or so of my practice, and eventually write a blog about them. Maybe I’ll write a “part 2” in 2026.