Cardiac Risk Scores and Missed Myocardial Infarction
Missed myocardial infarctions in ED patients prospectively categorized as low risk by established risk scores.
Singer et al. (2017) American Journal of Emergency Medicine.
Risk scoring for patients with chest pain and concern for ACS has been used to reduce unnecessary admissions and minimize the risk of adverse outcomes for discharged patients. These authors sought to compare a number of these scores with clinical gestalt as techniques to detect myocardial infarction, in conjunction with conventional two-set troponin testing. Most of the existing literature on these scores was performed using high-sensitivity troponin testing not yet available in the United States.
This was a multi-center secondary analysis of data obtained from a prospective trial of a high-sensitivity troponin assay using each institution’s own existing troponin assay. It compared HEART (score 0-2), TIMI (score 0), GRACE (score 0-50), EDACS (score 0-15), with “low-risk” clinician gestalt. All patients were admitted and charts were reviewed up by a panel of cardiologists for presence of MI during admission or subsequent provocative testing. For each scoring method, the authors computed and compared C-statistics for each score using the chart review as a gold standard. They also computed the percentage of patients judged as low-risk by each scoring method and the “miss rate” of false negatives. Finally, they computed score cut-offs required to maintain a 0% miss rate and re-counted the percentage of patients judged as low-risk
18% of 434 patients admitted had myocardial infarctions. EDACS had the highest percentage of patients judged as low-risk (52%), followed by clinical impression (26%), HEART score (32%), TIMI (7%), and GRACE (4%), which had the lowest percentage of low-risk patients. GRACE also had the highest miss rate (6.3%), followed by HEART (3.6%), EDACS (0.9%), and TIMI and clinical impression, which both had a 0% miss rate. These were not statistically significantly different from each other, but EDACS did have a slightly lower upper limit of normal for miss rate (3.5%) than clinical impression (3.9%) and TIMI (14.1%), while HEART (8.7%) and GRACE (32.3%) also had higher upper limit of normal for miss rate. C-statistics were higher for EDACS and HEART than for TIMI and gestalt, which were all higher than GRACE. Adjusting score cut-off’s to generate a 0% miss rate reduced the low-risk percentage for each score, resulting in EDACS having the highest percentage (33.6%), followed by clinical impression (still 26%), HEART (19.5%), TIMI (still 7%), and GRACE (3.3%), which had the lowest percentage.
EDACS performed best in almost all categories, except miss rate, and even there its upper limit of normal was still the lowest even though its point estimate was not. Clinical impression tied for best performance in miss rate with TIMI, being the only scores with a 0% miss rate, but performed better or equivalently in all other categories. GRACE appeared to have the least useful test characteristics. The study was notably limited by its relatively small size and the diverse assays used for troponin testing, with the authors noting that many of the missed MI cases involved point-of-care troponin tests that are known to be suboptimal. The absence of data on medium-term mortality or major adverse cardiac events also limits the applicability of these findings, as its unclear what effect admission vs. discharge would have on any “low-risk” patients. Overall, this study supports the use of clinical impression and complements much of the existing literature on use of these scores in the setting of high-sensitivity troponin testing.