The GRACE risk score accurately predicts the likelihood of in-hospital death and 6-month all-cause death and nonfatal myocardial infarction composite endpoint events in patients with ACS. The model has good predictive ability, with a C statistic of 0.85. It also performs well in two external validation sets. With the increase of risk score, the in-hospital mortality increased [11, 15]. This scoring system is derived from the largest ACS registration study in the world [16]. However, it contains many variables which may not be readily available at the time of admission [17]. In the GRACE risk score (2.0) proposed in 2014, the admission Killip classification and serum creatinine were modified to include the use of diuretics, making the initial data easier to obtain. The C index of death, which can predict both short and long-term mortality, exceeds 0.82 in the overall population at 1 and 3 years when using the French registry of Acute ST-elevation and non-ST-elevation Myocardial Infarction (FAST-MI) 2005 cohort [18]. Another study cohort also confirmed the good predictive ability of this scoring system [19]. Because the GRACE risk model was derived nearly 20 years ago, fewer Asians were initially included in the early database. However, in subsequent studies, the GRACE scores conducted in Asian populations also showed good predictive power, which further confirmed that despite the advancements in modern therapy and management, GRACE scores continue to accurately classify patients with ACS [20, 21, 22].

Although biomarkers have now been included in risk scores, several studies have not confirmed that high sensitive cardiac troponin (hs-cTn) and B-type natriuretic peptide (BNP) can improve the GRACE score [17]. However, a recent study showed that the area under the curve (AUC) of the GRACE risk estimate after growth differentiation factor 15 (GDF-15) adjustment increased from 0.79 to 0.85 (p 0.001) in the validation cohort using biomarkers. The GRACE score was also enhanced once the N-terminal fragment brain natriuretic peptides (NT-proBNP) were included [23]. Investigating the value of biomarkers in clinical risk scores needs to be further explored.

The TIMI risk score was first proposed in 2000. It performs well in terms of risk stratification for the prediction of 30-day mortality in ST segment elevation myocardial infarction (STEMI) patients after admission. In addition, the TIMI score in patients with unstable angina/non-ST segment elevation myocardial infarction (UA/NSTEMI) also performs well in patients seen in the emergency room with chest pain. The variables included in these two scoring systems are mainly derived from the data and electrocardiogram (ECG) that are easily available in the emergency room. These indicators are good at predicting short-term prognosis and long-term major adverse cardiovascular events (MACE) [24, 25, 26, 27]. It has been suggested to use a modified TIMI score (mTIMI, range 0–10) which gives some variables more weight. The goal was to determine whether improving the risk stratification of the TIMI risk score by giving more weight to ischemic ECG abnormalities and troponin elevations could help to more safely discharge patients following 12-hour troponin testing. While the mTIMI risk score outperformed the original score (AUC 0.87 versus 0.77, p 0.001), neither score by itself is sensitive enough at scores $>$0 to permit early and safe discharge without follow-up care. Therefore, in patients with normal ECG and negative troponins, the utility of TIMI and mTIMI scores for risk classification is limited [28, 29, 30].

A study proposed a prediction model for the prognosis evaluation of a new antiplatelet drug based on the TIMI risk score. Vorapaxar, a new antiplatelet drug, can be effectively used for secondary prevention of stable patients with identified atherothrombosis [31, 32, 33]. To evaluate the safety and effectiveness of vorapaxar in secondary prevention of patients with an acute myocardial infarction (AMI) using the results of the thrombin receptor antagonists in the secondary prevention of atherosclerotic thromboischemic events and myocardial infarction hemolysis (TIMI50) test, a TIMI risk score for secondary prevention (TRS2°P) score was developed to predict long-term recurrent cardiovascular (CV) events [34, 35]. This score is effective in determining the likelihood of repeat MACE in various risk categories, and shows a high predictive ability in external validation. After patients are stratified according to the TRS2°P, high risk groups such as elderly patients with myocardial infarction and patients with complications, can also benefit from the treatments recommended by guidelines such as antiplatelet medication [36, 37]. Patients with a high risk of recurrent CV events often have complications following anti-platelet medication and invasive therapy. Careful follow-up is required in these patients to minimize future MACE and will be the subject of future studies in these high-risk ACS patients.

Although the TIMI risk score is a powerful tool which can be used for risk stratification of ACS patients in the emergency room, it should not be used as the only means to determine the disposition of patients. Studies have found that NT-proBNP outperforms the TIMI score in predicting death following an AMI [9, 38]. In addition, the combination of baseline NT-proBNP, C-reactive protein, creatinine level, and inflammatory markers in the TIMI risk score provides more data regarding risk stratification and prognosis following an ACS [39, 40, 41].

Recently, a biomarker-based risk model for MACE within one year after admission for Chinese patients with ACS was proposed, which highlights the significant role of NT-proBNP in predicting MACE in ACS patients. The final model combines NT-proBNP with six other clinical variables, and showed better discrimination in the validation cohort (C statistic 0.79, 95% CI 0.73–0.85), and was superior to the GRACE and TIMI risk scores [42]. The European Society of Cardiology (ESC) now recommends that NT-proBNP be used as a prognostic factor for patients with coronary artery disease (Class IIa, level B) [43]. NT-proBNP is a neurohormone peptide secreted by ventricular cells and is closely related to ventricular dysfunction. Future studies will continue to define the role of NT-proBNP in risk stratification of ACS patients in clinical practice.

The ACEF score is a straightforward and practical risk prediction technique since it only includes three independent criteria. It was initially developed for patients undergoing elective heart surgery to assess perioperative mortality. In order to compute the ACEF score, the following formula was used: age (years) / ejection fraction (%) + 1 (in case serum creatinine values were $>$2 mg/dL). Studies have confirmed its accuracy compared to more complicated risk scores [44]. Subsequent studies showed that the ACEF score also had important prognostic value for patients with ACS. A cohort study validated the ACEF score, and demonstrated that these three factors could independently predict the outcome of patients with ACS following coronary revascularization and produce a predictive value comparable to the GRACE score [10]. The ACEF score has an independent predictive effect on 1-year mortality with a strong AUC of 0.79, according to data from the Korean Acute Myocardial Infarction Registry, which included data from an ACS cohort undergoing percutaneous coronary intervention (PCI) [45]. When applied to patients with a non-ST-elevation ACS (NSTE-ACS), it demonstrated superior discrimination compared to other complex risk stratification models [46].

In addition, other risk factors have been added to the ACEF score and it has been combined with other scores to further enhance its performance. The carotid plaque score (cPS) was assessed with data from carotid ultrasonography. By combining cPS with the modified ACEF score, the degree of freedom of MACE was 71% and 31% (p 0.001) for lower and higher scores at 5 years. When combined with ordinary ACEF scores in ACS, cPS enhances predictive values [47]. When diabetes, a common risk factor for patients with coronary heart disease, is included in the new ACEF diabetes comprehensive score, better accuracy and calibration factors are achieved [48].

Many complications other than cardiovascular adverse events can also be predicted by the ACEF risk score. In a study, the ACEF score accurately predicted additional clinical outcomes, such as bleeding, in addition to in-hospital mortality [49]. In patients with myocardial infarction who have ST segment elevation after a coronary intervention, a high ACEF score predicted the occurrence of contrast-induced acute kidney damage (CI-AKI) [50]. It was also reported that the ACEF score achieves good performance in identifying the adverse prognosis of high-risk patients with complex coronary lesions after PCI, including bifurcation lesions and chronic total occlusions [51, 52].

One of the variables included in the ACEF score, the ejection fraction, will change with the degree of myocardial ischemia, so that the timing of evaluating the ACEF score is particularly important. These studies have demonstrated that the ACEF score can provide a new and simple tool for daily clinical practice to stratify the risk of patients with ACS. Despite the fact that the ACEF score is simple to use and has performed on par with more complex models, long-term validation studies in various populations, hospitals, and nations are still required to assess its role in ACS patients.

Although many studies have proposed a variety of prognostic risk scores for ACS, an appropriate score for patients admitted for the first time with ACS still needs to be developed. Based on the data from patients with AMI from ACS-1 registries in Quebec and Canada, a C-ACS risk score has been developed, and verified in patients with ACS in four large data sets. The C-ACS score, which varied from 0 to 4, was generated using logistic regression modeling. One point was given for each of the following variables: age 75 years, Killip $>$1, systolic blood pressure 100 mmHg, and heart rate $>$100 beats per minute. This score has a C statistical value of 0.79. Notably, when the C-ACS score is 0, there is a potential to accurately identify 97% of short-term survivors, and to evaluate the possibility of in-hospital death in ACS patients [53]. In one study, the C-ACS score outperformed age in predicting in-hospital mortality among patients with AMI [54]. Some studies have shown that not only does the C-ACS risk score perform well in predicting infections that may occur following PCI in patients with AMI, the data suggests that these patients are more prone to develop contrast-induced nephropathy after PCI when the C-ACS risk score increases [55, 56]. This score can be obtained by calculating a number of variables that are not based on blood tests and ECG interpretations. It performs well for predicting both hospital and long-term death, and can be easily calculated, making it better suitable for diagnosing and treating ACS in the emergency department as well as early risk stratification following admission.

Over 45,000 patients from the Portuguese ACS Registry were included in the study of the ProACS risk score [57]. In all the research cohorts as well as independent external validation cohorts, the score has satisfactory discrimination ability [58, 59]. The ProACS risk score does not perform well in predicting long-term prognosis compared to the GRACE risk score [11]. Although in this study, the ProACS had a high recognition rate and a similar C-statistic compared with the most effective risk stratification score, it was under calibrated in the NSTE-ACS cohort. The ProACS risk score was derived to facilitate the use of instant information to help determine early risk stratification after admission, to assist in making timely and effective decisions. Therefore, it mainly focuses on short-term results, while the results of long-term follow-up are insufficient, and the included variables are limited.

According to the baseline data of 62,546 unselected patients with ACS from multiple hospitals in China, the CCC-ACS score was recently developed to predict in-hospital mortality in these ACS patients. This score differs from the China acute myocardial infarction (CAMI) risk score proposed in 2018, which has the same forecasting ability as the GRACE score, but it contains up to 16 variables, and only the computational complexity reduces its applicability [60]. The CCC-ACS score includes seven variables, which are different from other risk scores. The variables included in the CCC-ACS score take into account the evaluation of patients before blood testing. Except for ST segment changes on the ECG, the other variables mainly focus on the patient’s vital signs and medical history. The AUC of this new risk score in the training dataset was 0.84 (Hosmer-Lemeshow goodness-of-fit test p = 0.1), and it also performed well in the validation dataset [61]. These two scoring models have the drawback that the research population is close to 100% non-white and exclusively Asian, which prevents them from being really generalized until they are further validated in diverse populations and nations.

CMR is a valuable tool for determining the risk of heart failure. It characterizes myocardium by using a variety of different imaging parameters, which has been widely accepted as a reference standard for quantifying chamber size and ejection fraction [62]. The predictive ability of CMR in myocardial infarction (MI) is constantly being explored. At present, some studies have confirmed its role in AMI and other special types of myocardial infarction, such as unrecognized myocardial infarction (UMI) and myocardial infarction with nonobstructed coronaries (MINOCA) [63, 64, 65, 66]. Recently, the CMR risk score was proposed for risk categorization in STEMI patients. It includes left ventricular ejection fraction (LVEF), microvascular occlusion (MVO) and myocardial infarction (MI) size. In the derivation cohort, the score performed well in predicting the 1-year composite endpoint, and even exceeded the GRACE score [67]. Compared to GRACE and transthoracic echocardiography-LVEF, the CMR score offers additional prognostic classification and may have an impact on how patients with STEMI are managed [68]. For patients with an STEMI with an LVEF 50% by echocardiography, selective use of CMR can significantly improve the prediction of MACE [66]. In addition, a study has shown that in all time periods, CMR had a similar predictive value for the main endpoint [69]. The extrapolation of this risk scoring model still needs to be verified in larger multicenter research cohorts.

Compared to young patients, elderly patients with AMI have more comorbidities, and an increased risk of death following an AMI may result from poorer physiological reserves and more dysfunction (including physical ability and cognition). Functional decline has been linked to limitation in mobility as a potential mechanism [70, 71]. In a study of 3006 patients $\ge$75 years old who survived an AMI after discharge, Dodson and colleagues proposed a model to predict 6-month mortality risk in this population, the SILVER-AMI mortality risk model, which included around 9.5% of the patients who were non-white. In addition to the more common clinical features, the final risk model also includes four features specifically designed for the elderly: hearing loss, poor mobility, weight loss and poor health as reported by patients. The model has a good capacity for discrimination (AUC of the validation queue = 0.84) and is properly calibrated (Hosmer-Lemeshow p $>$ 0.05) [72, 73, 74]. Using this scoring model, a 180-day readmission risk model for elderly patients with AMI was established. The functional mobility of patients is also emphasized in the variables included in this risk model. Similar information is obtained in the verification queue, where the model’s differentiation ability is 0.68. In addition, over 40% of participants were hospitalized after 180 days following an AMI [75]. This work was the first to provide a mortality risk model for senior patients following discharge from the hospital with an AMI. However, the study still has some limitations. First, although the study was fully validated internally, the performance of this prediction model was not centrally evaluated in the external database. In addition, the access to information about related dysfunction needed to be evaluated by this score, is relatively limited. It is worth noting that the elderly often have multi organ and multi system chronic diseases, which may lead to inaccurate reporting of the cause of death. A summary of main risk models for determining the prognosis of ACS are provided in Table 1 (Ref. [11, 15, 18, 24, 42, 53, 58, 60, 61]).