There is a physician at a Midwestern institution who is known to come on service and order an echo bubble study on several (all?) patients the first day he is on service. He has doubtless impressed a generation of medical students when the occasional PFO (patent foramen ovale) is found and he gets to wax prolific about shunts and other mechanisms of hypoxemia for a few minutes as they listen with fawning and rapt attention. What they do not know is that he finds PFOs about 10% of the time he orders his echos, and this is just above the rate of PFOs in the general population. He's basically finding PFOs at random.
There is a physician at a prominent east coast institution who thinks everybody has a pulmonary embolism (PE). When she is on service, everybody gets rushed to the CAT scanner (the proverbial "donut of truth") or the nuclear medicine suite for a scan for PE. They also get lots of ultrasounds of the extremities and other ancillary tests for PE. Surely some "unexpected" PEs have been found, as have some pseudoembolisms - small filling defects on the CAT scan that are not emboli or which are emboli that could have been ignored without consequence. And surely, countless scans were done that yielded nothing, but we don't register those. We register only the brilliant find, the guy who had PE and PJP pnuemonia, the attending surprising and impressing everyone.
Open the NEJM and turn to the Case Records of the Massachusetts General Hospital and tally how many labs, serologies, X-rays, scans, biopsies, etc. are done in these challenging cases. In the Case Records, the diagnosis is ultimately made, even though sometimes it's a diagnosis that nothing can be done about and the patient dies soon afterward. The problem is that these zebra hunting safaris are not limited to the cases where a rare and interesting diagnosis lurks in waiting for the sly and perspicacious physician who can pat himself on the back in the Case Records - these safaris go on all the time, and the misses are not recorded.
This situation arose because there is no penalty for all this testing, save for the occasional diagnostic misadventure where somebody is actually harmed by the pursuit of a diagnosis with a vanishingly small probability. Somebody else pays for the excess testing - either Medicare or a third-party payer. (The patient pays too, in terms of Death by 1000 Needlesticks, but we ignore this.) If there is no cost to the expedition, but even a minuscule chance of finding a treasure, physicians are going to launch the expedition.
So the next time you're impressed that your attending found some PFO or some PE that nobody else expected, ask yourself: Is this discovery attributable to his or her keenly calibrated sense of pre-test probability, or is it simply a function of excessive testing?
If CMS were to start reporting individual physician testing volume and unique patient encounters, we could have an actual crude estimate of this phenomenon.