In this real-world population-based study, we showed that the modified SMIR score performed similarly to the GRACE 2.0 score in a multi-ethnic Asian population to predict all-cause mortality at 1 year after STEMI and NSTEMI.
Inter-ethnic differences in STEMI patient outcomes have been published previously. Previous studies carried out both locally15,16,17 and abroad18 suggested inter-ethnic differences in outcomes such as mortality. Although there are established coronary risk factors, such as smoking, hypertension, hyperlipidemia, and diabetes mellitus, these risk factors cannot fully explain the observed interethnic variations in outcomes.19. Ethnic differences also existed in possible pathophysiological factors such as economy, lifestyle, anthropometry, and patient susceptibility to cerebrovascular disease.16.18. It should be noted that these factors are not included in contemporary risk scores such as the TIMI and GRACE 2.0 scores.3.4, and are also difficult to determine in the treble setting. As such, there is a need to assess the relevance of contemporary risk scores for predicting outcomes within a multi-ethnic or ethnic-specific population.
The GRACE registry initially included 123 hospitals from 14 countries in Europe, North and South America, Australia and New Zealand20. This registry initially had no participation from Asian countries and therefore the original GRACE score derived was not obtained from Asian patient data2. The subsequently updated GRACE 2 registry expanded the recruitment to involve 154 hospitals, this time including hospitals in Asia (including China)20. Nevertheless, the updated GRACE 2.0 score was only derived from the old registry and was validated in a French cohort3. In the Asian context, studies of the GRACE 2.0 score have been carried out in ethnically homogeneous populations such as in JapansevenVietnamese8 and Chinese21 populations. The Japanese study was a single center validation study of 412 STEMI patients who underwent PPCI. This study showed a good AUC of 0.92 to predict 360-day mortalityseven. The Vietnamese study was carried out on 217 patients from a single center diagnosed with unstable angina, NSTEMI and STEMI. The authors used the score to stratify their patients, but did not specifically study the predictive performance of the GRACE 2.0 score.8. Fu et al. in China developed the CAMI-NSTEMI score based on 5775 patients from the Chinese Acute Myocardial Infarction (CAMI) Registry. They showed that the CAMI-NSTEMI score was superior to that of the GRACE score (AUC 0.81 vs 0.72, p 21. We found that the performance of the GRACE 2.0 and modified SMIR scores were similar, whether in all AMI patients or in ethnically specific patients.
In the modified SMIR score, we found that higher LVEF was associated with reduced all-cause mortality at 1 year. LVEF is currently not a component of the TIMI and GRACE 2.0 scores. LVEF has previously been shown to be associated with increased mortality in post-MI patients22. Therefore, it was worth considering the use of LVEF as a risk predictor variable for AMI patients. Previously, it was difficult to perform a dedicated transthoracic echocardiography study in the acute setting due to time constraints. However, with the advent of point-of-care echocardiography with portable devices, the patient’s LVEF can be quickly obtained at the bedside.23. Future risk scores may consider the use of variables that were not readily available before.
Additionally, there are notably emerging risk stratification tools for AMI patients beyond published risk scores. Emerging approaches, such as metabolomics-based risk stratification, may play a role in future risk stratification beyond currently clinically available variables.24.25. Identified soluble biomarkers, such as those for myocardial fibrosis, may play a role in determining the severity of acute myocardial infarction26. The authors also reported machine learning-based methods for risk stratification of AMI patients using big data approaches, with results that appear to outperform traditional risk models.27.28. It is not unlikely that in the future, risk prediction will incorporate a combination of clinical, haematological, biochemical, echocardiographic and electronic health record-based information, appropriate to the local context, to provide risk stratification. personalized risks for each AMI patient. Nevertheless, until this technology becomes mature and widely available, and also in practice areas where resources are limited29traditional risk scores will remain relevant.
Strengths and limitations
This study used a large national AMI patient database based on mandatory reporting to ensure near-complete case coverage. It also minimized the selection bias. Data linkage with the National Death Registry ensured accurate and objective determination of outcomes. Another strength of this study is that this scoring system is based on the population in contemporary treatment, both in secondary prevention and in revascularization.
Nevertheless, we recognize several limitations to this study. While the ethnically diverse population of Singapore is ideal for this study, no superiority in using the modified SMIR score over the popular and validated GRACE tool has not been demonstrated. Thus, the scientific and clinical contributions of our findings do not seem high. Nevertheless, this study fills the gap in the literature by investigating the GRACE 2.0 score in a multiethnic Asian population that is currently lacking and demonstrating that GRACE 2.0 is likely to be applicable to other Asian populations that are predominantly of Chinese descent. Malay or Indian. As our study focused exclusively on PCI patients alone, our results cannot be extrapolated to patients without PCI such as the thrombolysed population. However, thrombolysis as a reperfusion strategy is rarely used, at least in Singapore. Although we found that the modified GRACE 2.0 and SMIR scores were able to correctly classify patients into low categories ( 60%) risk, we were unable to compare observed mortality for finer subgroups at different levels of predicted risk due to small sample sizes. Clinicians should apply their own clinical judgment if they require more granular risk stratification. Further studies are needed to optimize the performance of the indicated scores in predicting all-cause mortality at 1 year. Also, the points corresponding to the categories of some prognostic components, such as age at onset of acute myocardial infarction and Killip’s class, are nonlinear, but these components were used for the regression models. Therefore, the clinical interpretability of these components should be made with caution.