ED Chest Pain Rules: Follow Your HEART ?

The HEART Score is yet another clinical prediction tool to identify emergency department (ED) chest pain patients who are at low risk for a short-term major cardiac event (MACE).1 The acronym is clever—HEART: history (H), electrocardiogram (ECG, E), age (A), risk factors (R), and troponin (T). Each of the five variables is assigned 2 points for convincing evidence, 1 point for partial evidence, and 0 points for no convincing evidence. Another older clinical prediction tool, the TIMI (Thrombolysis in Myocardial Infarction) risk score,2 also includes history, ECG, age, risk factors, and cardiac biomarker findings, but the scoring system is less finely divided; each variable has two values, not three. For example, HEART assigns 2 points for troponin more than three times normal limit, 1 point for troponin one to three times the normal limit, and 0 points for troponin of not more than the normal limit. In contrast, TIMI assigns 1 point for “positive cardiac biomarker” and 0 points otherwise. In this issue, Fernando et al.3 present a systematic review of test accuracy studies that evaluated the HEART score as a predictor of MACE. They identified 29 studies (28 distinct cohorts) assessing the sensitivity and specificity of a HEART Score > 3 for predicting MACE in ED patients who present with chest pain. In seven of the cohorts, the authors were able to compare the performance of HEART > 3 to TIMI > 1. Overall, the studies comprised more than 44,000 ED visits. In the studies evaluating both the HEART and the TIMI scores, the HEART score performed better. The authors conclude that “the HEART score has excellent performance for prediction of MACE in ED chest pain patients and should be the primary clinical decision instrument used for the risk stratification of this patient population.” This conclusion may be premature. The HEART Score does appear to discriminate better than the TIMI Score, but the authors present no evidence that it discriminates better than other scores or an experienced clinician. In the 1980s, we saw ED chest pain decision rules from Tierney et al.,4 Goldman et al.,5 and Pozen et al.6 The Tierney rule did not even perform better than clinicians in its own derivation set. Sensitivity and specificity were 81 and 86% for the rule and 87 and 78% for clinicians. A comparison of the Goldman and Pozen rules to the admitting decisions of family medicine and primary care internal medicine residents found that neither rule “could reduce unnecessary admissions without seriously increasing the rate of inappropriate discharges.” In 1998, the Pozen rule (also called ACI-TIPI—acute cardiac ischemia time-insensitive predictive instrument) was evaluated in a clinical trial. During the intervention periods, the rule-produced probability of ACI was printed across the top of ECGs for ED patients with chest pain. This did not affect sensitivity for ACI, nor did it increase the specificity of attending physicians or supervised residents, only of unsupervised residents. Now, in addition to HEART, we have the Emergency Department Assessment of Chest Pain Score (EDACS) in original and simplified versions,7, 8 the Vancouver Chest Pain Rule,9 and others. Which rule is best, and is it better than the judgment of an experienced clinician? In 1954, Meehl, famously showed that simple prediction rules tend to perform better than trained professionals making subjective judgements, and we are aware of the report by Backus et al.11 that reweighting the elements of the HEART score resulted in only moderate improvement. However, we find it difficult to believe that a score which gives the same weight to age ≥ 65 years as it does to troponin of more than three times normal discriminates better than a clinician who weights the abnormal troponin more highly. All other factors being the same, a 69-year-old woman with a normal troponin has the same HEART score as that of a 43-year-old man with a troponin of more than three times the normal limit. According to a strict interpretation of the HEART score, if they both have a family history of coronary artery disease and an ECG showing nonspecific repolarization abnormality, they both have a HEART score of 4, which is “positive” using the cutoff of 3. If the HEART score is the “primary clinical decision instrument used for risk stratification,” then both patients should be treated the same way, whether that means sending them both to urgent cardiac catheterization or admitting them both for monitoring and serial troponins. Despite Meehl and his successors, is it possible that the experienced clinician may be able to preserve the HEART Score's high sensitivity for MACE while admitting and catheterizing fewer patients? One possible reason is that HEART itself requires clinician judgment. The patient history is rated as “slightly suspicious” (0 points), “moderately suspicious” (1 point), or “highly suspicious” (2 points), where “suspicious elements include: retrosternal pain, pressure, radiation to jaw/left shoulder/arms, duration of 5–15 minutes, exertional, diaphoresis, nausea/vomiting, relief of symptoms with sublingual nitrates.” One can see how two different interviewers might disagree on how many points to assign. They might also disagree on how many points to assign for risk factors, as no explicit definition of each is provided. Before we accept HEART as the winner of the chest pain decision rule derby, we would like to see it compared to other rules and most especially to the judgment of experienced clinicians who can weight variables differently and consider additional elements in a more nuanced way.