To HEART score, or not to HEART score – now that is a question. In the past five years it has become increasing common to hear Emergency Medicine providers and residents reference the HEART score when reviewing chest pain cases. This is especially true when they are justifying plans to discharge a patient.

At that stage of training, the HEART score may provide a helpful framework, but mid and late-career Emergency Physicians may not have heard it, utilize it, or find it as particularly helpful – so its time for a deep dive. 

 

The TL/DR;? Don’t tell me the HEART score, because I don’t find it helpful.

What is the HEART score?

The HEART score is an assessment tool for Emergency Department (ED) patients with chest pain.1

 

HEART

Image from Rebel EM

 

A score of 0-3 is generally considered low risk, 4-6 is considered moderate risk, and 7-10 is considered high risk. We will consider the derivation and evidence for this shortly.

 

The underlying context is that in the ED, we are worried about missing myocardial infarction (MI). An influential article by Pope2, published in the New England Journal in 2000, suggested the 2% of cases of acute myocardial ischemia are missed in the ED. This number has come under scrutiny3, but emergency physicians remain concerned about missed MI. It is likely that MI was more commonly missed prior to the advent of cardiac troponins in the 1990s, but things have changed.

 

A large study by Weinstock et al4 addressed this very question. This multicenter study reviewed a cohort of 11230 patients admitted to the hospital with chest pain who had negative troponins. Negative troponins were defined as an initial negative troponin, followed by a second negative troponin drawn 60 to 420 minutes from the initial test. A “clinically relevant adverse cardiac event” was defined as a life-threatening arrhythmia, an inpatient STEMI, cardiac or respiratory arrest, or death.

  • These events occurred in 0.18% of the cohort (20 patients), with a 95% upper confidence limit of 0.27%.
  • The event rate was even lower – 0.06% –  if, in addition to negative troponins, patients had normal vital signs and an interpretable, non- ischemic ECG.
  • They also found that 0.55% of the initial cohort developed positive troponin on the third or subsequent assay, but without these clinically relevant adverse events. It is notable that Weinstock and colleagues did not consider revascularization to be a “clinically relevant adverse event”. This is discussed below.

In the Weinstock study, the proportion of all chest pain patients who got admitted was close to 50%. The authors argue that the low event rate suggests that these admissions are not necessary. It is certainly unlikely that the rate of clinically relevant adverse events was any higher in the patients who were discharged.

Nonetheless, efforts continue to try to identify this tiny percentage of patients who have bad outcomes despite troponin-negative chest pain. Is this even possible? In my view, probably not. The reasoning comes from basic epidemiology.

 

Specificity – the Great Leveller

Let us do a thought experiment. Imagine that we had a diagnostic strategy (a test, an algorithm, a scoring system, whatever) to predict these bad outcomes in troponin negative chest pain. I will call it the Atlantis protocol. Let’s further assume that the Atlantis protocol has a sensitivity of 99% and a specificity of 95%. Sounds pretty good, right? OK, let’s apply the Atlantis protocol to 100,000 troponin negative chest pain patients, with an event rate of 0.18% as described in the Weinstock study. Here is what we find: 

HEART


What you can see is that for every true positive picked up by the Atlantis protocol, there are about 28 false positives (4991/178). This is because the outcome we are searching for has a very low prevalence. Even the high specificity of our imaginary Atlantis protocol cannot change this fundamental fact: when the pre-test probability of disease is low, testing will generate many false positives.

 

So, a priori, we can say that strategies like the HEART score are likely to be problematic, because clinically relevant adverse outcomes in troponin negative chest pain patients are very rare. The next evidence-based medicine principle to consider is the problem of composite outcomes.

 

Composite Outcomes

Many cardiology studies look at composite outcomes, and the existing HEART score studies are no different. Most looked for a primary outcome of Major Adverse Cardiac Events (MACE), defined as death, MI, percutaneous coronary intervention (PCI), coronary artery bypass graft (CABG), or angiography demonstrating procedurally correctable stenosis that was managed conservatively.5 Naturally, we need to do studies within the context of current practice. No ethics board would allow us to forgo interventions such as PCI and CABG in patients in whom they would typically be recommended, just so that we could wait to see what happened had we not intervened.

The problem is that some elements of the composite outcome are more important than others. Death matters, MI likely matters, PCI and CABG may matter in some patients.

A 2020 meta-analysis of the effect of PCI on death and MI concluded that PCI prevents death and MI in patients with unstable CAD, but has no effect on patients with stable CAD.6 This study suggested that there was benefit in non-STEMI ACS. But even this is controversial: a Cochrane systematic review7 of early-invasive (i.e. angiography) vs early conservative management in unstable angina and non-STEMI ACS showed that there was no difference in death, although there was a reduction in MI, angina, and rehospitalization at follow-up. Although the cardiology community seems to favour the early-invasive strategy8, it is reasonable to believe that, as with most therapies, the greatest benefit accrues to the highest risk patients—which are likely those who are troponin positive, especially in the era of high sensitivity troponin. 

So, if a patient is admitted with troponin negative chest pain, but we label it “unstable angina”, does that patient benefit from PCI? Or was the chest pain actually non-cardiac, but the patient just happened to have some coronary stenoses that were amenable to intervention?

Some components of the HEART score, such as cardiac risk factors, might prejudice admitting cardiologists to send a patient to the cath lab. Thus, such patients would ultimately be more likely to be classified has having a MACE, and the HEART score appears to have greater validity on the surface.

We will further consider how composite outcomes affected the HEART studies later in this article. First, the issue of cardiac risk factors needs further examination.

 

Cardiac Risk Factors

Medical students are often drilled to take an inventory of Framingham cardiac risk factors when they see a patient with chest pain. However, the predictive value of cardiac risk factors in assessing ED chest pain patients has long been questioned. A 1992 study9 looked at the association between acute ischemia and cardiac risk factors in ED chest pain patients. In women, there was no association. In men, diabetes and family history were associated with a higher risk of ischemia, but the relative risk increase was small (much lower than the risk increase associated with just having chest pain).

 

A landmark study of cardiac risk factors in ED chest pain patients was published in 2007 by Han and colleagues10. They retrospectively analyzed data on more than 10000 patients in the Internet Tracking Registry of Acute Coronary Syndromes. Results were stratified by age, which was itself a strong predictor of ACS.

  • In patients over the age of 65, going from zero to 4 cardiac risk factors had no association with the diagnosis of ACS.
  • In patients aged 40-65, having zero risk factors had a negative likelihood ratio of 0.53, and having 4 or more risk factors had a positive likelihood ratio of 2.13.
  • In the youngest age group, having zero risk factors had a negative LR of 0.17; having four risk factors had a positive LR of 7.39.

Likelihood ratios are tough to interpret. As a rule of thumb, negative LRs have to be around 0.1, and positive LRs have to be around 10, to be really helpful. If we imagine a 45 year old man having chest pain, with a pretest probability of ACS of 5%, how would risk factors change this? If he had zero risk factors, the Han study suggests his probability would drop to 3%. If he had 4 risk factors, his probability would rise to 10%. Would this change our clinical management?

The LRs are better for the under 40 age group, but that group starts with a much lower risk of ACS. So, the utility in that group is also questionable. 

A smaller study of ED chest pain patients by Body et al.11 found no correlation between risk factors and MI. Another study by Hess et al12 found that risk factors were only weakly predictive.

When I quote this evidence to my trainees, they have a hard time accepting it. They often respond, “Yeah, I understand that, but because the patient has risk factors I’m still more worried.” I respond that this is precisely the opposite of what they should be doing. They are being biased by the presence of risk factors, which are very poor predictors of ACS. They would be better off ignoring them.

Part of the problem is that we know these risk factors do cause coronary disease. So why do they not help in the ED? The short answer is, because the patient came to the ED with chest pain. Cardiac risk factors predict the development of CAD in the general population over time. BUT (and this is a big but), when a patient develops chest pain, and he or she chooses to come to the ED for evaluation, all bets are off. The patient is in a different population (namely the population of ED patients with chest pain) and the outcome is very different—not long- term CAD, but ACS on that particular visit. To know if risk factors are predictive of that outcome, we need to study that specific ED chest pain population, which is exactly what the above studies did.

Given that risk factors are lack predictive value, it is surprising to see them form such an important part of the HEART score. In fact, risk factors can give you as many points on the HEART score as troponins – which are actually very strong predictors of short-term cardiac events. Hmm. The problem likely lies in the historical basis for cardiac risk stratification was generated in an era without high sensitivity troponins. 

 

History of the Heart Score

The initial HEART score study, published in 2008, was a retrospective analysis of 122 ED patients with chest pain, performed at a community hospital in the Netherlands1. The investigators chose the components of the HEART score based on “clinical experience and current medical literature.” Ideally, clinical decision instruments are derived using a careful mathematical analysis of retrospective data, to identify the strongest predictors of the outcome.13 This does not seem to have been the case for the HEART score.

 

In this initial study:

  • 16/122 patients had MI
  • 2/122 died
  • 14/122 had PCI
  • 6/122 had CABG

Overall, 29/122 reached the composite outcome (some patients reached more than one endpoint, of course). The higher the HEART score, the greater chance of meeting the composite outcome. Interestingly, when they analysed the components of the score, only history, ECG, and troponin were significant predictors. Age and risk factors had nonsignificant trends toward association.

This was a very small study, but it is certainly hypothesis generating. There followed a slightly larger retrospective chart review published in 2010, performed at four hospitals in the Netherlands.14 This had 2161 patients, although only 910 presented with chest pain. In this cohort 158 had MACE: 

  • 92 MI
  • 82 PCI
  • 36 CABG
  • 13 death

Once again, of the five elements of the HEART score, only history, ECG, and troponin were independent predictors of MACE. The HEART score in aggregate was again predictive, with higher values increasing chance of MACE. Interestingly, the HEART score did not significantly predict MACE that occurred more than 6 weeks after presentation.

In 2013 came a multinational (again retrospective) study, but this time done using a database in the Asia-Pacific region15. They analysed 2906 patients, of whom 374 had MACE:

  • 19 STEMI
  • 353 non-STEMI
  • 19 emergency revascularization
  • 9 death

This study defined the endpoints slightly differently, and “elective” revascularization within 30 days was not considered part of the primary endpoint. Again, the HEART score worked, and the authors report that all elements of the score were statistically significantly associated with the outcome.

In 2013 we also had a prospective validation of the HEART score, done at 10 hospitals in the Netherlands.5 There were 2440 “unselected” chest pain patients in this study. Again, obvious STEMIs were diverted to a cath lab and not included. The primary outcome was our old composite friend, MACE, but combined AMI and death was a secondary outcome.

One interesting quirk was definition of positive troponin values. They state troponins were taken on patient arrival, and typically repeated after 6 hours. However, they only considered the initial troponin for the purposes of calculating the HEART score. This seems strange. Obviously, some AMI patients who present early will have a negative initial troponin. The effect of this methodology will be to have a subset of patients with lower HEART scores who are actually high risk. When these patients are aggregated with the rest of the cohort, it probably made the low-and-intermediate HEART categories seem higher risk than they really were. 

Higher HEART scores were associated with a higher risk of MACE, as one would predict. Furthermore, they state that all elements of the HEART score “differed significantly” between the MACE and non-MACE groups. When one looks at the raw numbers however, the distribution of scores for risk factors seem quite similar between the MACE and non-MACE groups. Also, they report that 53.6% of the MACE group had negative troponin, which may be influenced by their ignoring second and subsequent troponins. They report that the combination of ECG and troponin only had significantly worse performance than the whole HEART score, but again this needs to be taken with a grain of salt due to the definition of a positive troponin.

There were 407 patients with MACE, of whom 16 died and 155 had AMI. The major driver of MACE were the 251 PCI and 67 CABG, although there is of course overlap in these groups. Nonetheless, at most 42% of the MACE patients had MI or death.

A research letter published in 2018 by Sharp et al16 prospectively collected data at 15 hospitals in the Kaiser Permanente Southern California health system. There were 29196 ED encounters. Only 0.6% of patients had MI or death at 30 days. Higher HEART scores were associated with a higher risk of MACE. Likelihood ratios for low (0-3), moderate (4-6), and high risk scores (7-10) were 0.3, 1.6, and 4.9, respectively. Overall 6-week MACE rate was 1.5%, with 52% of these events being revascularization. Again, we have no data on which components of the score are most predictive.

There is also a small study by Leite et al17 looking at 233 chest pain patients in Portugal. The study was retrospective, as was grading of the history component of the score. There were only 22 ACS patients, of whom 10 had unstable angina. There is no analysis of which components of the HEART score were predictive, so this study adds very little.

The question we really need answered is, “How did the HEART score perform in troponin negative chest-pain patients?” By troponin negative, I mean serial troponin negative, like in the Weinstock study, but unfortunately it is impossible to answer this question from these studies.

 

 

Pragmatic Studies

Mahler et al18 published an RCT in 2015 evaluating the HEART pathway vs. usual care. The pathway is a variation on use of the HEART score. Essentially, if a patient has a HEART score of 0 to 3, and negative troponins, the patient goes home. If the HEART score is 4 or higher, or troponins are positive, the patient stays for cardiology or provocative testing, etc. This was a small study, with 141 patients in each arm, done at a single centre in North Carolina. Using the HEART pathway dropped the rate of advanced cardiac testing from 69% to 57%. Patients were discharged from the ED without advanced testing in 40% of cases compared to 18% with usual care. None of the patients discharged had MACE, in either group.

A 2017 paper by Poldervaart et al19 was done in 9 Dutch EDs. Every 6 weeks, one hospital was randomized to switch from usual care to using the HEART score. They found a slight difference (-1.3%) in the rate of MACE when switching to the HEART score. The rates of MI and death were similar, but the patients evaluated using the HEART score were less likely to be labelled as having unstable angina, more likely to have outpatient clinic visits, and less likely to have stress testing, angiography, or PCI.

A 2018 paper by Allen et al20 was a chart review of chest pain patients done over 30 months at an academic hospital in Florida. The authors compared the 15 months before implementation of a low-risk chest pain pathway, which was based around the HEART score, to the 15 months afterward. Their primary finding was that they increased the discharge rate of chest pain patients from 39.2% to 46.7%. There was no change in MACE. Interestingly, their “low risk” protocol involved a mandatory advanced diagnostic test (CT, stress test, nuclear study) for anyone with a HEART score of 4 or greater, who had not had one in the last year.

Clearly, their baseline rate of admission was very high, and the HEART score was only able to slightly increase their very low risk tolerance.

 

What does this all mean?

Taken together, these studies show the same thing. The HEART score predicts MACE (the composite outcome) but about half of that composite outcome is revascularization in patients who did not have MI. They went looking for CAD, they found it, they fixed it. Unfortunately, what the existing data do not tell us which components of the score predict death and MI, the outcomes that matter the most.

Actually, since positive troponin is part of the definition of MI, we can conclude that the rest of the HEART score tends to predict non-MI revascularization–which may not help patients.

In practice environments where there is low tolerance for discharging troponin-negative chest pain patients, use of the HEART score seems to give providers confidence that they can discharge the lowest risk patients. In those environments, that may be an improvement (like in the United States). In Canada, where I practice, most of the time we discharge these patients already.

We are left with a needle-in-a-haystack problem. There are a tiny percentage of chest patient patients who have negative serial troponins who may have bad outcomes. Finding those patients is a near statistical impossibility, because they are rare, and any future test or algorithm will likely generate many false positives.

The HEART score might, on one hand, make us feel better about sending people home who will be fine. On the other hand, widespread use of the score will tend to lead to over-investigation in patients who are older, and with cardiac risk factors. We will find some coronary artery disease, but it is questionable if this has meaningful patient centred outcomes. 

When I tell my residents not to use the HEART score, it is because troponins alone are already excellent predictors of the most important outcomes. Use of the score adds nothing beyond a careful history of the pain, patient age, ECG findings, and cardiac enzymes.

 

References

1 Six AJ, Backus BE, Kelder JC. Chest pain in the emergency room: value of the HEART score. Neth Heart J. 2008 Jun; 16(6): 191–196. doi: 10.1007/BF03086144
2 Pope JH et al., Missed diagnoses of acute cardiac ischemia in the emergency department. NEngl J Med 2000 Apr 20;342(16):1163-70. doi: 10.1056/NEJM200004203421603.

3 Newman D. Stress testing. SMART EM. 25 January,2012. http://www.smartem.org/podcasts/stress-testing-moment-clarity

4 Weinstock et al. Risk for Clinically Relevant Adverse Cardiac Events in Patients with Chest Pain at Hospital Admission. JAMA Intern Med 2015; 175(7):1207-1212. doi:10.1001/jamainternmed.2015.1674

5 Backus BE et al. A prospective validation of the HEART score for for chest pain patients at the emergency department. Int J Cardiol 2013; 168: 2153-2158. http://dx.doi.org/10.1016/j.ijcard.2013.01.255

6 Chacko L, et al. Effects of Percutaneous Coronary Intervention on Death and Myocardial Infarction Stratified by Stable and Unstable Coronary Artery Disease. Circ Cardiovasc Qual Outcomes 2020; 13: e006363. DOI: 10.1161/CIRCOUTCOMES.119.006363

7 Hoenig MR, Aroney CN, Scott IA. Early invasive versus conservative strategies for unstable angina and non-ST elevation myocardial infarction in the stent era. Cochrane Database of Systematic Reviews 2010, Issue 3. Art. No.: CD004815. DOI: 10.1002/14651858.CD004815.pub3. Accessed 22 March 2022.

8 Bavry AA. Non-ST-Segment Elevation Acute Coronary Syndromes: Lessons learned over the last decade. J Am Coll Cardiol 2017 Apr;69(15):1894-1896.

9 Jayes RL, et al. Do patients’ coronary risk factor reports predict acute cardiac ischemia in the emergency department? A multicenter study. J Clin Epidemiol 1992 Jun;45(6):621-6. doi: 10.1016/0895-4356(92)90134-9

10 Han JH et al. The role of cardiac risk factor burden in diagnosing acute coronary syndromes in the emergency department. Ann Emerg Med 2007 Feb;49(2):145-52, 152.e1.doi: 10.1016/j.annemergmed.2006.09.027.Epub 2006 Dec 4

11 Body R, McDowell G, Carley S, et al. Do risk factors for chronic coronary heart disease help diagnose acute myocardial infarction in the Emergency Department? Resuscitation 2008; 79(1):41–5.

12 Hess EP, Brison RJ, Perry JJ, et al. Development of a clinical prediction rule for 30-day cardiac events in emergency department patients with chest pain and possible acute coronary syndrome. Ann Emerg Med. 2012;59(2): 115-2

13 https://emottawablog.com/2015/07/epi-lessons-part-4-decision-rule-articles/
14 Backus BE, et al. Chest pain in the emergency room: a multicenter validation of the HEART score. Critical pathways in cardiology. 2010;9(3): 164-169.

15 Six, AJ et al. The HEART score for the Assessment of Patients with Chest Pain in the Emergency Department: A Multinational Validation Study. Critical Pathways in Cardiology 2013; 12(3): 121-126.

16 Sharp AL, et al. The HEART Score for Suspected Acute Coronary Syndrome in the US Emergency Departments. J Am coll Cardiol 2018 Oct 9; 72(15): 1875-1877. https://doi.org/10.1016/j.jacc.2018.07.059

17 Leite L, et al. Chest pain in the emergency department: risk stratification with Manchester triage system and HEART score. BMC Cardiovascular Disorders 2015;15:48. DOI 10.1186/s12872-015-0049-6