The Global Registry of Acute Coronary Events (GRACE) risk score is less accurate at predicting in-hospital mortality for ethnic minorities with non-ST-segment elevation myocardial infarction (NSTEMI), a new British study suggests.
As reported in the European Heart Journal, the GRACE 2.0 score provided good discrimination for in-hospital mortality in White and ethnic minority patients (area under the curve for both, 0.87).
Results showed, however, that the score was well calibrated in White patients, with an estimated to observed (E:O) mortality rate ratio of 0.99, but overestimated risk in Black, Asian, and other ethnic minority patients, with an E:O ratio of 1.29.
The overestimation of risk was present in all three ethnic groups, and greater for ethnic minority women than their male peers (E:O ratio, 1.31 vs 1.27).
“Before we got the results, I thought this is going to show we’re grossly underestimating the risks because these patients have a high prevalence of diabetes and have previous vascularization, and when our analysis suggested we’re overestimating the risk, that was an eye opener,” said senior author Mamas A. Mamas, MD, Keele University, Stoke-on-Trent, United Kingdom.
“But to my mind, it’s not whether you over or underestimate the risk, it’s that it’s not calibrated well in a particular group of patients and so you could be potentially disadvantaging these patients, either giving them a treatment that may not be appropriate for them or withholding treatment that is,” he told theheart.org | Medscape Cardiology.
Mamas said the growing racial and ethnic disparity in outcomes for patients admitted during the pandemic with acute coronary syndrome sounded the alarm about structural racism and bias, but also called into question the tools used in deciding treatment. For example, a 2020 report showed that pulse oximetry failed to detect hypoxemia approximately three times as often in Black people as in White people.
European Society of Cardiology guidelines for the management of NSTEMI give the GRACE score a class 1 recommendation to guide the timing of an invasive strategy and a class 2a recommendation to guide prognosis, but little is known about its performance in specific ethnic groups. The score takes no account of ethnicity or comorbidities such as diabetes, hypertension, or previous coronary revascularization, and, Mamas said, was derived from and validated in predominantly White populations.
“Most risk stratification tools, when they’re being assessed, will often look at that performance in men versus women, but I think we should go further than that and should also mandate their performance in different ethnic groups,” Mamas said.
Lower Mortality, Better Care?
For the study, the researchers calculated the GRACE score — characterized as low, medium, or high risk — for 326,160 NSTEMI admissions in England, Wales, and Northern Ireland in the prospective Myocardial Infarction National Audit Project registry between January 2010 and March 2017.
Of these admissions, 26,976 were in ethnic minorities, including Black patients (Caribbean, African, Black British, any other Black background), Asian patients (Indian, Pakistani, Bangladeshi, Asian British, any other Asian background), and patients of other non-White ethnicity (mixed group, including Chinese).
Across all three risk groups, ethnic minority patients were about 5 years younger and presented with significantly more cardiometabolic risk factors, including diabetes, hypertension, and hypercholesterolemia.
Despite this, in-hospital mortality was 2.9% in ethnic minorities and 5.2% in White patients (P < .001).
Age is one of the strongest determinants of outcomes in patients with MI, and it might be that the relationship between age and outcomes is different in ethnic minorities, suggested Mamas. Prehospital social, environmental, and behavioral differences could also explain the mortality difference.
Notably, the use of invasive coronary angiography and percutaneous coronary intervention was similar in the two cohorts at low and intermediate risk, with higher rates of both procedures in ethnic minorities at high risk. Minorities at all risk levels were more likely to undergo coronary bypass surgery and to receive dual antiplatelet therapy and statins at discharge.
“That’s a good thing, I’d say, but the bigger thing for me is it could quite as easily have gone the other way because we didn’t know what the performance was,” Mamas said.
Race as Proxy
The GRACE score can still be used in practice, he said, but the results support testing its performance in other regions with different ethnic groups and recalibrating it as needed.
Mamas noted that the QRISK score used in the United Kingdom to predict cardiac events in primary prevention has race as one of its factors. A 2021 study also identified distinct race-specific contributors to heart failure using race-specific machine-learning risk models that were better able to predict 10-year heart failure risk than traditional non-race-specific models.
“I think really the AHA, ESC, and so forth should come out with some sort of best-practice statement to say when you’re developing these scores you need to validate them in different age groups, different genders, different ethnic minorities, races,” Mamas said. “Because until then, we’re going to have this sort of situation where we’re using scores that may disadvantage one group of patients.”
The present study is one of the largest applications of the GRACE score in diverse populations and provides intriguing findings, said Vasan Ramachandran, MD, DM, a professor of medicine and epidemiology at Boston University Schools of Medicine & Public Health, who was not involved in the study.
“We all know that race is a social construct and when you find this kind of overestimation, and they nicely refer in this paper to how the ethnic minority patients presented at a younger age and at a higher burden of risk factors, race in these people may be a proxy for a number of other determinants like healthcare access, education,” he told theheart.org | Medscape Cardiology.
There is strong pushback against using race in risk prediction models, Ramachandran said, noting that his recent in silico analysis reported significant differences in 10-year estimates of cardiovascular disease in Black and White patients using sex-specific and race-specific pooled cohort equations.
“If you really want to correct and understand risk and the biology of risk, you need to be focusing on the root causes that underlie the racial disparities,” he said.
An editorial will be published to accompany this study but was not available at press time.
Mamas and Ramachandran reported having no relevant financial relationships. Coauthor disclosures are listed in the paper.
Eur Heart J. Published online February 24, 2022. Full text
Follow Patrice Wendling on Twitter: @pwendl For more from theheart.org | Medscape Cardiology, follow us on Twitter and Facebook.