On the theme of knowing when and when not to follow the diktats of emergency medicine, one of the greatest challenges for a practicing ER doc is chest pain. Missed MI is still the biggest driver of malpractice costs, and last I hear, ER docs still send home something like 2% of patients who are having MI or unstable angina. Not good. So over the last decade we’ve gotten all these chest pain observation units and rapid rule-out protocols and early stress tests and all sorts of protocol-y goodness to fulfill every ER doctor’s goal of never sending home an MI.

And it’s good, and works. At least, for most cases. Consider if you will:

Mr Smith is 58 years old. He smokes, and was diagnosed with hypertension and high cholesterol several years ago. He is treated with medicines for these, but is not particularly compliant about taking them. He has a strong family history of accelerated cardiovascular disease, with a father who died of an MI in his 40s and a younger brother who has had a CABG. He presents with 24 hours of stuttering chest pain. It is episodic, lasting 2-10 minutes, dull, midsternal, without radiation or associated symptoms. It occurs sporadically both at rest and with exercise. On arrival, his ECG and troponin are normal, and he rates his pain as 5/10.

So this is a pretty straightforward case, isn’t it? Slam dunk, admit to Card Tele, rule out & stress test. See? Protocol-driven medicine is fun and easy.

Oh, I forgot to mention something:

Mr Smith has previously had two MIs, has five stents in place, and says the pain he is having today is exactly the same as the last time he had an MI.

That gets your attention, doesn’t it? I just ramped up my level of concern quite a bit. In this case, I am probably calling a cardiologist to see the patient in the ER and starting him on heparin and a nitro drip.

But I also forgot to mention a couple of other details:

Mr Smith had his last cardiac cath eight months ago, showing patent stents. His stents are three years old. He had a negative nuclear stress test three months ago. He also has a crippling anxiety disorder and has visited the ER for chest pain twelve times over the past year. He has been admitted seven times, ruling out each time.

Oh. Well, that does change things, doesn’t it?

This is where protocol-driven medicine breaks down. Chest pain observation units are great for undifferentiated chest pain. but for someone with well-known, recently studied disease, they are less useful. Mr Smith is a real patient — I changed nothing from the patient I saw yesterday. And I see a Mr Smith every single day I work.

The academic emergency physician will say, rightly, that I should treat the third Mr Smith exactly the same as the second one, because you cannot know when his noncardiac chest pain is noncardiac and when it is cardiac. A risk-averse doc will assert that he just admits any patient like this, because he does not want to run the risk of ever ever getting sued. But that is not practical or sustainable in the real world. I only have so many beds in the obs unit! There are only so many times you can admit someone for observation without objective evidence of active disease before you have to admit it’s pointless. No matter where you personally set that threshold, there will be a patient who will visit you in the ER more than that.

I recall in residency a guy with known CAD who visited the ER for chest pain 550 times in a three-year span. We kept his ECG on the wall for easy comparison. After a while we stopped treating him with nitro and just gave him orange juice, which fixed his chest pain. But I digress.

If you work in an ER, someday you are going to send home a patient who presented with chest pain with a history of CAD. If you don’t, then you are a crummy doctor with no clinical judgement. It’s bad medicine and a poor stewardship of resources to admit every patient with chest pain. The difference between a good ER doc and a bad one, between an experienced physician and a robot, is acquiring the judgement to know where to draw the line, and how to do so safely.

I sent Mr Smith home, after talking to his cardiologist, observing him for six hours with serial ECGs and troponins, and arranging next day follow-up in the cardiology clinic. In this case, for this person, that seemed reasonable. For other patients, some of them do get admitted, depending on a million sometimes subjective variables — how many ER visits, when they were last studied, how old the stents are, how the patients look, how bad their disease has been, how long the pain has been going on, etc etc etc. There’s no good protocol for that.

Someday I am going to be wrong. In fact, I have been wrong, though with care there have been no bad outcomes. I can live with that — you have to be able to live with that if you are going to survive long working in the ER.

This is the art of medicine. This ability to recognize patterns, to integrate a lof of variables and clinical data points and come out with an accurate, back-of-the-envelope estimate of risk, that is the hallmark of a true physician. It somes with time. We all start off as algorithm-driven neophytes and some never seem to progress beyond that point. But for the Mr Smith I see every day, who doesn’t want to be admitted to the hospital again (he never does), but he also doesn’t want to die, he really values having a “good doctor.”

“Shadowfax” is an emergency physician who blogs at Movin’ Meat.

Submit a guest post and be heard on social media’s leading physician voice.