I’ve spent nine years working in an emergency department, which means I’ve also spent nine years performing electrocardiograms at triage. With a couple thousand waiting-room ECGs under my belt, I’ve noticed a few things:
- A lot of ECGs are ordered at triage.
- Many of them are performed on low-risk patients.
- Very few of those ECGs lead to a change in initial management.
Oh, and one more thing:
- Most physicians hate signing triage ECGs.
I can’t say I blame them. Triage ECGs interrupt their workflow and, worse than just slowing them down, that distraction can lead to errors. As an additional frustration, these patients often end up being seen by a different doctor in the department, and no one likes making diagnostic or treatment decisions that another provider will have to deal with. Plus, if the signing physician does happen to find something wrong, there’s always a nagging concern that the patient will end up being added to their already full group and need to be seen immediately—further interrupting flow.
Triage ECGs bring work and distraction.
So it’s understandable why many emergency physicians are rejoicing at the publication of a new study by Hughes, Lewis, Katz, and Jones: “Safety of Computer Interpretation of Normal Triage Electrocardiograms” .
Below are some great opinions on the article from the target demographic: emergency physicians.
- Can We Trust Our Computer ECG Overlords? by EM Literature of Note
- Triage ECGs: Reducing Interruptions in a Busy ED by R.E.B.E.L. EM.
- How unreliable are computer algorithms in the diagnosis of STEMI? by Dr. Smith’s ECG Blog
They’re all well thought-out posts with perfectly reasonable conclusions. Dr. Salim Rezaie and Dr. Anand Swaminathan from R.E.B.E.L. EM also make it clear that they wouldn’t want to eliminate physician over-reads of “normal” triage ECGs, just delay them to reduce interruption.
My only issue with that approach is that the emergency physicians I’ve work with are always busy. Whether they read a triage ECG now or in fifteen minutes, they’re going to be doing something when I try to slide that paper in front of their screen.
As the only tech at triage, I’ve also got somewhere between one and thirty patients out in the waiting room, so if I don’t get that ECG signed right after it’s performed, I’ll get distracted myself and occasionally forget. It’s a lose-lose situation.
Maybe it’s just the practice at the hospital I worked at, but I think we’d make more progress at reducing interruptions by investigating a stricter triage protocol for ECGs. Though the rate of “normal” ECGs in this study was 26%, in my experience it’s somewhat uncommon to see a tracing with the words “Normal ECG” printed at the top in the ED (I’m just making this up, but I’d guess something < 10% at triage).
It’s not the “normal” ECGs that are weighing on my physicians.
I believe a bigger burden comes from ECGs that are read as “abnormal” in some way by the computer but were not indicated in the first place and do not change the patient’s management in the waiting room. If we could cut back on the number of triage ECGs that are performed on low-risk patients with non-cardiac complaints, I think that would make a bigger dent in reducing interruptions.
But I digress…
While seeing the computer spit out “Normal ECG” certainly stratifies a patient as lower risk, it’s not the same as “no risk.” That’s a problem, because triage ECGs are (or should be) performed in a moderate-to-high risk population with complaints suggestive of an acute cardiac process. In that cohort, the negative LR suggested by this study won’t “rule-out” concerns as well as it would in a general ED population getting ECGs as part of their routine workup.
As Dr. Steve Smith summarizes in his post on the topic, “Computer algorithms that make the diagnosis of ‘normal’ are usually correct, but is usually good enough?”
We don’t have an answer, and I don’t expect to make any headway on that issue with this blog post, but I thought it would be interesting to share a few subtle STEMIs from my collection that were read as normal by the computer.
It’s not data, and certainly not scientific, but it’s interesting (to the folks who like ECGs, at least).
- Hughes KE, Lewis SM, Katz L, Jones J. Safety of Computer Interpretation of Normal Triage Electrocardiograms. Acad Emerg Med. 2017;24(1):120-124. doi: 10.1111/acem.13067