To STAT or not to STAT

A passage about limited resources and optimizing imaging from The Emergency Mind: Wiring Your Brain for Performance Under Pressure by Dan Dworkis MD PhD:

Within the broader context of your responsibility however, there frequently will be significant variability in the relative urgencies of individuals being imaged. Some patients—like a person seemingly experiencing an acute stroke—do need to be scanned immediately. Others—such as a patient with abdominal pain, stable vitals, and a reassuring physical exam—while no less “deserving” of those resources, would receive nearly equal benefit from being scanned now as in an hour from now. Optimizing care across the field in this context would involve prioritizing CT scans for those patients who would receive outsized benefits from immediate imaging, even if this makes some other patients wait longer.

Put a different way: If everything is stat, nothing is stat.

Stat abuse is one of those crimes especially tempting to inpatient teams in busy hospitals. It’s natural to want answers (and dispo) as soon as possible, and we assume that we will get them faster if we increase the priority of the exam.

All a clinician knows is that sometimes something ordered routine takes forever and that ordering stat should generally result in it being performed faster. They may not even care if the read is prioritized in all cases so long as the patient is freed from the waiting and future transport.

It’s also human nature for there to be a distribution with certain individuals ordering an outsize proportion of “stat” exams. The negatives of over-ordering or inappropriate priority are almost always placed on other staff. In a zero-sum game, selfish behavior may be an optimal choice for individual success even if it makes the system less efficient overall. Hospitals very rarely scold their staff for such abuses.

I don’t think most clinicians even have any idea where along the spectrum their behavior falls. Knowledge of outlier performance one hopes might curb excesses, and that data would certainly be helpful for individuals to know (presuming those individuals are capable of feeling shame and said shame functions as a deterrent). Such information would have to be long-term and stratified well to be meaningful (we should expect different levels of stat exams as a fraction of orders from different hospital units, for example). Otherwise, data are dismissible.

Ultimately, pleading and punishment are often ineffective and/or undesirable.

A more helpful approach would include data to guide decision-making on a case by case basis:

The EMR should show in real-time the expected wait for different study types based on the current queue and exam types pending, both inpatient and outpatient (i.e. how many unnecessary exams are obtained during an inpatient stay due to fears of long delays for outpatient follow-up?). Yes, a routine study may unexpectedly get bumped further down the line, but a smart system would incorporate predictions based on the current patient census, admission diagnoses, time of year, and whatever else some machine learning algorithm would include its impenetrable black box of Skynet code.

It would be extremely helpful for all parties to know if an MRI should be expected today or tomorrow, sometime this afternoon or more likely at 3 am.

And so, yes, of course, people are working on this in the machine learning world. But hurry up. I for one will continue to welcome our AI overlords and their promised efficiency gains, but I’m still waiting.

Leave a Comment.