sinaiem.org

So, lots of big things were discussed today, but I’m going to focus on Dr. Judd Hollander’s talk, as it was crammed with insight on a very common problem – achieving disposition on the 8 million patients we seen annually with chest pain (this is national, not just Sinai). Of these 8 million, 3 million are sent home and so we admit 60-65% of chest pain, of which only 15% have real disease… Cardiologists hate us for this, but is there an alternative? What’s the evidence behind what we do?

In 1994, when first guidelines for UA came out, there were 130 randomized controlled trials in cardiac emergency management. Now there’s almost 1000 trials referenced in the 2007 guidelines.

What does this literature say about who’s safe to send home? Well, the more patients are sent home, the more misses occur – where miss is defined as MI / death / urgent revascularization. What miss rate can we live with? When we go best on impressions, we miss 3-6% — which is really unacceptable – even 2% missed MI makes you a liability for your department.

In looking for a safe group to discharge, Dr. Hollander started with the classic Lee and Goldman 1985 paper (Arch IM 1985;145:65 PMID: 3970650) multicenter trial. They looked at independent variables, and hey, even age alone doesn’t get you under 1% in the 20-29 year old cohort. Discharging those with “sharp or stabbing” chest pain still misses 5% of MI. Reproducible, positional, pleuritic pain is all around 5-10% of missed MIs.

Part of the I*trACS study (Han et al. Ann Emerg Med 2007;49:145 PMID: 17145112) of 17,000 patients shows the 5 classic Framingham risk factors (Diabetes, smoking, HTN, high cholesterol, family history) are almost completely useless in the ED. Even when rheumatologists were brought in to declare a patient has costochondritis, 6% of those patients had MI. Alternative diagnoses also had a 4% rate of ‘bad stuff’ at 30 days (this is from another I*trACS paper from last year –  Acad Emerg Med., 2007:14:215) What about discharging patients with a normal EKG? 3-5% missed MI. But normal EKG will at least predict for your patient a very low risk of death, and very low risk of VFib.

Going through Goldman’s risk management scheme [link to EMCrit], some truths become apparent:

• Don’t talk to a chest pain patient until you’ve seen their EKG – the EKG is far more predictive of disease than any element in history or physical.
• Even when STEMI seems likely, 20% do not actually have STEMI.
• If the EKG is normal and the patient’s pain has lasted continuously for 48 hours, and markers negative, there’s less than 1% chance of missing MI.

There’s more to the algorithm of course, but it’s all really a way of trying to get to a less than 2% or less than 1% miss rate. You’ll note there are frequent references to EKG and markers and elements of the story, but no reference to diabetes or smoking or classic risk factors – and yet this was in the mid-80’s! Way ahead of its time, in a way.

One segment of the population where risk factors actually do have some value is in the under 40 year old group– especially when these young adults have normal EKGs, they can be safely sent home. Add on one normal set of markers, and your miss rate goes down to 1.4% (Marsan et al. AEM 2005;128:26-32, PMID: 15635134).
What about more complex scoring schemes? The TIMI risk score consists of 7 items we fill out at Sinai all the time:

• Age > 65
• 3 or more classic risk factors
• Known CAD > 50%
• ST segment changes on ECG
• 2 or more anginal events in past 24 hours
• ASA use within 7 days prior
• Elevated cardiac markers

These are all kind of logical items, based on co-morbidities, symptoms and CAD history — and in (Chase, et al. Ann Emerg Med. 2006:48:252, PMID: 16934646) they showed that risk rises steadily with more points. A low score is not low enough to discharge, however, but a high score can drive placement.

But TIMI makes use of markers, and markers are really what’s driving a lot of our disposition. And even though we’re always taught about CK-MB rising first and troponin staying elevated longer, the truth is sensitivity of troponin and MB for detecting MI at the time of presentation is about the same. Specificity of MB is actually better than troponin I or T – 97% vs. 93%, with an odds ratio of MI much higher for MB than for troponin (25 vs. 9-11, respectively). Troponin does have better prognostic value in angina patients (Storrow, I*trACS, Ann Emerg Med 2006 Dec;48(6):660-5, PMID: 17112930).

But look at this – even a negative troponin and a Goldman score less than 4% still had a 4.9% risk of bad outcome in 30 days (Limkakeng et al. AEM 2001; 8:696, PMID: 11435183). Hollander used this as a segueway into brief observation periods — how patients and administrators prefer them, and about 15% of OBS patients end up with signs of ischemia and admissions.

My favorite part of his talk came here, when Dr. Hollander digressed into the uselessness of telemetry beds. Tele got its start in the 1970’s when true STEMI patients developed VFib but were revived with defibrillation. The idea was that defibrillation could help even low-risk chest pain patients, should they develop VFib. But it would only work if they were monitored and identified before the next morning’s rounds. There was never evidence to support VFib was likely in low-risk patients or that monitoring improved outcomes. After years of wastefulness, Estrada, in two retrospective studies (Am J Cardiol 1994;74:357 and Am J Cardiol 1995;76:960) did not identify VT/VF in 338 patients and then in a prospective study of thousands had a 0.8% rate of upgrades as a result of arrhythmia and a 0.3% rate of transfer to CCU as a result of telemetry findings. Schull’s retrospective study (AcadEM 2000;7:647 PMID: 10905643) of over 8000 patients showed 20 cardiac arrests on the tele floor, only 9 were detected and acted upon by monitoring. Only 3 survived to discharge – one was saved by the monitor, one wasn’t clear how the arrest was noted, and one arrest was picked up by the patient’s roommate. So roommates might be as valuable as tele.

Dr. Hollander has studied this himself (Hollander et al. Annals EM 2004;43:71, PMID: 14707944 and Journal of Emergency Medicine , Volume 33 , Issue 1 , Pages 53 - 60 PMID: 17630076). Now at Penn (HUP) they track their low risk (TIMI<2 or Goldman < or = 7) patients that were marker negative – over 1000 patients – and none of them had sustained VT/VF, bradydysrhythmias, or preventable CV events. Notably, in this group, 13% of the patients had ACS and real disease in need of workup (which sounds about right, given the first stats presented in this post) — but none of these patients needed telemetry monitoring.

The talk then moved into imaging. First, echocardiography: A study from Peels (Am J Cardiol 1990;65:687 PMID: 2316447) showed that, in 43 patients with acute chest pain, echo missed 7% of AMI and had many false positives — so detecting wall motion abnormalities is not as good as clinical criteria alone. Another study the following year (Sabin, Circ 1991;84:S85 PMID: 1884510) suggested a 1% miss rate on a small group of patients discharged with normal EKG and echo — not bad, but in the intervening years echo has not been compared to biomarkers to assess its comparative value.

How about sestamibi imaging to help aid disposition? It’s actually great, very sensitive and specific for MI… if you can inject while the patient’s in pain or shortly after. But it’s very cumbersome to set up and tough to administer (Tatum, Annals EM 1997;29:116 PMID: 8998090).

As for prior stress, well, it doesn’t seem to influence decision-making, and based on our experiences with MI shortly after stress, it shouldn’t. Looking for signs of ischemia in a stressed heart is not the same as predicting who’ll have plaque rupture and MI. Dr. Hollander also pointed to data suggesting that, despite their ok prognostic value, patients with normal stresses don’t even stay away from the ER for very long (he also notes that when a patient has a negative stress, he or she is told “go home but be careful, we can’t say you’re disease free” but a negative cath is a lot more reassuring).  So stress testing doesn’t affect ED use or ED decisions on admissions and cath (Shaver and Hollander, Academic EM 2004 Dec;11(12):1272-7 PMID: 15576516). A clean cath (not just 30% occlusion but truly clear artiers) seems to give you at least 2-5 years of very low risk of MI.

Finally, coronary CT angiography — now with 64 slice / dual source (essentially like 128 slice) scanners approaching 96-99% sensitivity, the coronary CTA is like a noninvasive catheterization.  At Penn they started getting coronary CTA on any patient with TIMI 0-2 and normal creatinine, before even markers were resulted — if the CTA was negative, they would be sent home. Their data was great, and echoed in the new findings from Gallagher (Annals EM 2007;49:125 PMID: 16978738) — 525 of 568 were scanned immediately, 81% were sent home with negative scans (<50% stenosis) and 19% admitted. Nobody with a negative CTA needed revascularization or had MI at 30 days. One was hit by a car. Those with admissions who got cath, the CTA matched up very well with their initial scans. CTA also led to cheaper and faster visits, and fewer revisits. He was very enthusiastic and higher quality machines should reduce radiation dosages, too.

Dr. Hollander then summarized his approach as follows: EKG first, clinical presentation a little, history of CAD, meds they’re on.  Then you stratify into low, intermediate and high risk categories, using TIMI or Goldman or the neural networks they sometimes use at Penn. Then adjuncts:  markers and imaging. After that, it’s time for disposition. High risk patients (EKG changes, CHF, arrhythmia, unstable vitals, drips, positive markers or positive sestamibi scans in the ED, or positive CCTA with good stories and markers) should go to the CCU. The low risk patients (with low-risk stories and normal EKG and normal sestamibi or CTA, if done) can go to the Obs unit, nonmonitored beds, or home if they’re at the lowest possible risk. The intermediate, tele-bed bound patients have equivocal stories, abnormal but not diagnostic EKGs, markers normal or slightly elevated, or scans showing old abnormalities.

It’s important to remember that with with much evidence out there, it’s possible to over-rely on some studies and discount others. So for other excellent evidence-based approaches to this problem, please read Dr. Corey Slovis’ approach and the thorough examination of chest pain at EMcrit.org.

Posted on Thursday, July 24th, 2008 at 5:45 am by Nick. Filed under Risk Stratification, Post-Conference Letter, Arrhythmias, Radiology, ACS, Blog.
You may post a comment.