Application of HEART, GRACE 2.0 and TIMI scores for risk stratification of chest pain: a single-centre external pilot study in a Sri Lankan population Authors: C. Welhenge, K. M. Fernando, J. N. Fernando, P. C. Deshapriya, M. L. D. Lekamge, A. Kasturiratne, A. P. De Silva Published: 2026 ABSTRACT BackgroundHEART score is widely used for risk stratification in patients with chest pain. TIMI and GRACE 2.0 scores are recommended for prognostication in patients with acute coronary syndrome (ACS). This study aimed to explore feasibility of conducting a larger study to evaluate the effectiveness of HEART score in a Sri Lankan population for risk stratification in patients presenting with chest pain, and to compare its diagnostic accuracy with GRACE 2.0 and TIMI scores.MethodsData were collected from 74 patients presenting with chest pain. HEART, GRACE 2.0 and TIMI scores were calculated for each patient. The predictive accuracy of the HEART score for diagnosing ACS and the predicting the occurrence of major adverse cardiac events (MACE) within 6 weeks was assessed and compared with TIMI and GRACE 2.0. scores.ResultsThe area under receiver operating characteristic curve (AUC-ROC) for the HEART, GRACE 2.0 and TIMI scores, for the diagnosis of ACS, were 0.889 (95% CI: 0.8171–0.9609), 0.805 (95% CI: 0.6758–0.9349), 0.812 (95% CI: 0.6961–0.9278) respectively, without any statistically significant pairwise difference between the scores. For prediction of MACE, the AUC-ROC values were 0.905 (95% CI: 0.8437–0.9669), 0.721 (95% CI: 0.5934–0.8493), and 0.767 (95% CI: 0.6467–0.888) for the HEART, GRACE 2.0 and TIMI scores respectively, with statistically significant differences observed between the scores.ConclusionsExploratory evidence from this pilot study suggests that the HEART score demonstrates good diagnostic accuracy for ACS and short-term risk of MACE in this cohort. Its diagnostic accuracy for ACS appears comparable to GRACE 2.0 and TIMI scores while demonstrating superior predictive accuracy for MACE. However, given the small sample size and high event rate, these findings should be interpreted with caution and require validation in larger studies. Introduction Chest pain accounts for majority of emergency medical admissions globally, including in Sri Lanka [1, 2]. However, only about 20% of these patients are ultimately diagnosed with acute coronary syndrome (ACS), which requires immediate treatment and inpatient care. Delays in diagnosing ACS, referred to as “rule out delays” contribute to overcrowding at the emergency department (EDs) [3]. Risk stratification of these patients is required for early identification of individuals who require urgent and more aggressive treatment [4]. Many patients presenting with chest pain are low risk and can be safely discharged with outpatient follow up [5]. Identification of patients at high risk helps physicians reduce morbidity and mortality while enabling better utilization of the limited healthcare resources, particularly in resource limited settings [4]. International guidelines recommend the use of risk stratification tools in patients with chest pain due to their superiority over clinical assessment used alone and their benefits in patient outcomes [6]. Expert cardiology panels recommend the utilization of dedicated risk scores for discriminating high risk patients for diagnostic and therapeutic decision making [7–9]. The HEART score was developed in 2008 for risk stratification in patients with chest pain. It categorizes patients into low, intermediate or high-risk groups according to their short-term risk of major adverse cardiac events (MACE). These events include acute myocardial infarction, the need for percutaneous coronary interventions (PCI) or coronary artery bypass grafting (CABG) and death within six weeks [5]. The use of this score has been recommended as a quick, easy to use, reliable predictor of ACS at the ED as a triage tool [4]. Many studies throughout the years have validated the HEART score for various populations including those in Netherland, United Kingdom, United States and South Korea [10–13]. A study conducted in India has validated the HEART score for Indian population, suggesting relevance for the broader South Asian region [13]. Global Registry of Acute Coronary Events (GRACE) score was published in 2003, to assess the in-hospital mortality risk of patients with ACS [13]. GRACE 2.0, a more accurate version of the score was subsequently developed, for improved risk stratification and prediction of the risk of death in the acute setting and in the longer term (six months post discharge) [14]. Both scores were developed and externally validated extensively in multiple regions throughout the world, including in South Asia [15–17]. A GRACE score cut off, of 140 has been recommended to identify the high risk NSTE-ACS patients who require urgent coronary interventions in the European society of Cardiology (ECS), American heart association (AHA) and American college of Cardiology (ACC) guidelines. Thrombolysis in Myocardial Infarction (TIMI) is another score, which was published in 2000. It was originally developed in patients with NSTEMI (Non-ST elevated myocardial infarction)/UA (Unstable angina) for prognostication and therapeutic decision making [18]. It has also been validated in countries throughout the world, including South Asia, but less extensively when compared with the GRACE score [19–22]. Although developed for use in patients with NSTEMI/UA, the score has demonstrated consistent performance in patients with potential ACS in the emergency setting [23]. Nonetheless, all these scores have yet to undergo validation for the Sri Lankan population. The purpose of this study was to assess feasibility of a larger study, on the effectiveness of the HEART score, at predicting the probability of ACS and early risk stratification, in patients presenting with chest pain in a Sri Lankan healthcare setting. Additionally, it sought to evaluate the diagnostic and predictive accuracy of the HEART score in comparison with the extensively validated TIMI and GRACE scores. Methods Participants This external validation pilot study was conducted at a single tertiary care centre in Sri Lanka from July to December 2024. Data were collected from 74 adult patients presenting to the ED with chest pain. All consecutive patients above 18 years presenting with chest pain, who consented to participate were included, regardless of the eventual diagnosis to reflect real-world emergency department data. Exclusion criteria included: pregnant and post-partum patients (up to 3 months), patients with chest pain following trauma and those with known familial cardiac diseases such as inherited cardiomyopathies, channelopathies or familial hypercholesterolemia. At the time of the study, PCI facilities were not available for emergency admissions at the relevant hospital. Hence, patients diagnosed with ST-elevation myocardial infarction (STEMI) who were eligible for PCI were transferred to a specialized PCI centre and were not available for follow up. However, patients who received thrombolytic therapy were included, to represent the full spectrum of patients who present with chest pain to the emergency department. Data Data collection was performed using an interviewer-administered electronic questionnaire. Additional required data was retrieved from the bed head tickets (BHTs) and the hospital’s digital laboratory information system. Serial ECGs were obtained in the ED and subsequently in the ward as clinically indicated. ECG interpretations made by the attending medical officers were documented and independently verified by the research team. hs-cTn I level, and other biochemical parameters were requested from the biochemistry laboratory as a routine practice from the ED or the ward immediately upon admission. The biochemical investigations were traced from the laboratory, once available and the relevant section of the questionnaire filled. The data collection process occurred independently of patient management, and the calculated scores were not disclosed to or used by the treating clinical team during decision-making. The patients were reviewed prior to discharge to assess them for the final diagnosis, which was made by the treating clinical team based on the history, examination, serial ECGs, troponin values and other relevant investigations. The required information was obtained from the BHTs and confirmed with the treating clinical team. Six-week follow-up data on the occurrence of MACE were collected from the patients and personal medical records. Data preparation Data was collected to a spreadsheet and recoded for statistical analysis. The HEART score was calculated from the collected independent scores using R Studio software. GRACE 2.0 and TIMI scores were calculated using validated web based freely available software. Outcome The diagnosis of ACS was made by the treating clinical team in accordance with the 2023 ESC guidelines for the management of acute coronary syndromes [8] and was cross checked by the research team. Based on the clinical history, ECG findings and the hs-cTn I level, ACS was further classified as unstable angina (UA), Non-ST elevated myocardial infarction (NSTEMI) and ST elevated myocardial infarction (STEMI). Occurrence of acute myocardial infarction (AMI), the need for PCI or CABG and death within six weeks of the onset of chest pain were considered MACE. Only NSTEMI and STEMI were considered as AMI. Statistical analysis Categorical data was presented using frequencies and percentages. The effectiveness of the HEART, GRACE 2.0 and TIMI scores, were assessed with, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUC-ROC), as measures of predictive accuracy. Delong’s test for two correlated ROC curves was used to compare the AUC-ROC of the different scores. All statistical analyses were performed using R Studio (Version 2025.05.0 + 496). Statistical significance was defined as p < 0.05 two-sided. Results Participants Data were collected from 74 patients presenting with chest pain to the ED and followed up for 6 weeks (Figs. 1 and 2). Baseline characteristics of the patients are illustrated in Table 1. The hospital at which the study was carried out receives patients from the immediate vicinity of the hospital as well as from other primary and secondary care hospitals. Therefore, the selected cohort of patients included both low risk and high-risk patients, which minimized the selection bias. However, the set of patients coming with chest pain, which was obviously non-cardiac in nature, could not be included as those patients did not warrant cardiac Troponin testing and were discharged from the hospital within few hours without extensive testing. Fig. 1Flow of the patients through the study Flow of the patients through the study Fig. 2Distribution of the patients through the study based on the diagnosis and occurrence of MACE Distribution of the patients through the study based on the diagnosis and occurrence of MACE Table 1Patient characteristics of the study groupPatient CharacteristicsNumberPercentage (%)Study group74100Male4257Female3243 Co-morbidities ACS (55) Non-ACS (19) Number Percentage (%) Number Percentage (%) Patients with anyco-morbidity4581.81263.1Diabetes2545.5421.1Smoking1832.7421.1Hypertension3360421.1Hypercholesterolemia1629.1421.1History of IHD2545.5631.6History of heart failure23.600Family history of CAD11.800History of stroke11.815.3CKD59.100Bronchial asthma11.8210.5COPD11.800Hypothyroidism00210.5Atrial fibrillation11.800GORD11.800Valvular heart disease0015.3Pulmonary hypertension0015.3Parkinson disease0015.3Thalassemia trait11.800History of tuberculosis11.800Dilated cardiomyopathy23.600None1018.1736.8 Table 1Patient characteristics of the study groupPatient CharacteristicsNumberPercentage (%)Study group74100Male4257Female3243 Co-morbidities ACS (55) Non-ACS (19) Number Percentage (%) Number Percentage (%) Patients with anyco-morbidity4581.81263.1Diabetes2545.5421.1Smoking1832.7421.1Hypertension3360421.1Hypercholesterolemia1629.1421.1History of IHD2545.5631.6History of heart failure23.600Family history of CAD11.800History of stroke11.815.3CKD59.100Bronchial asthma11.8210.5COPD11.800Hypothyroidism00210.5Atrial fibrillation11.800GORD11.800Valvular heart disease0015.3Pulmonary hypertension0015.3Parkinson disease0015.3Thalassemia trait11.800History of tuberculosis11.800Dilated cardiomyopathy23.600None1018.1736.8 Patient characteristics of the study group Percentage (%) Percentage (%) Patients with any co-morbidity Out of the 74 patients, 42 (57%) were male and 32 (43%) were female. A total of 60 patients (81.1%) were found to have cardiac-type chest pain, while 14 patients (18.9%) were classified as having non-cardiac chest pain. Patients diagnosed with stable angina, unstable angina, NSTEMI and STEMI were considered to have cardiac type chest pain. The characteristics of the chest pain among the study participants are presented in Table 2. Table 2Chest pain characteristics of the study groupACS (55)Non-ACS (19) Chest pain localization N Percentage (%) Chest pain localization N Percentage (%) Central3461.8Central1578.9Left side1527.3Left side421.1Right side59.1None11.8 Character of Chest pain N Percentage (%) Character of Chest pain N Percentage (%) Aching type35.5Aching type15.3Pleuritic type23.6Pleuritic type210.5Tightening type4378.2Tightening type1473.7Burning type610.9Burning type15.3Non-specific11.8Non-specific15.3 Onset of Chest pain N Percentage (%) Onset of Chest pain N Percentage (%) Sudden2341.8Sudden947.4Insidious3156.4Insidious1052.6Non-specific11.8 Radiation N Percentage (%) Radiation N Percentage (%) To neck/jaw814.5To neck/jaw526.3To left upper limb1221.8To left upper limb736.8To right upper limb11.8To right upper limb15.3To bilateral upper limbs23.6To bilateral upper limbs15.3To back23.6To back210.5None3767.3None736.8 Associations N Percentage (%) Associations N Percentage (%) Autonomic symptom3054.5Autonomic symptom736.8Exertional dyspnoea2138.1Exertional dyspnoea631.6Orthopnoea and PND35.5Orthopnoea and PND0Palpitations0Palpitations15.3Cough712.7Cough526.3Wheezing47.3Wheezing15.3Dyspeptic symptoms23.6Dyspeptic symptoms210.5Abbreviations: ACS; acute coronary syndrome, PND; paroxysmal nocturnal dyspnoea Table 2Chest pain characteristics of the study groupACS (55)Non-ACS (19) Chest pain localization N Percentage (%) Chest pain localization N Percentage (%) Central3461.8Central1578.9Left side1527.3Left side421.1Right side59.1None11.8 Character of Chest pain N Percentage (%) Character of Chest pain N Percentage (%) Aching type35.5Aching type15.3Pleuritic type23.6Pleuritic type210.5Tightening type4378.2Tightening type1473.7Burning type610.9Burning type15.3Non-specific11.8Non-specific15.3 Onset of Chest pain N Percentage (%) Onset of Chest pain N Percentage (%) Sudden2341.8Sudden947.4Insidious3156.4Insidious1052.6Non-specific11.8 Radiation N Percentage (%) Radiation N Percentage (%) To neck/jaw814.5To neck/jaw526.3To left upper limb1221.8To left upper limb736.8To right upper limb11.8To right upper limb15.3To bilateral upper limbs23.6To bilateral upper limbs15.3To back23.6To back210.5None3767.3None736.8 Associations N Percentage (%) Associations N Percentage (%) Autonomic symptom3054.5Autonomic symptom736.8Exertional dyspnoea2138.1Exertional dyspnoea631.6Orthopnoea and PND35.5Orthopnoea and PND0Palpitations0Palpitations15.3Cough712.7Cough526.3Wheezing47.3Wheezing15.3Dyspeptic symptoms23.6Dyspeptic symptoms210.5Abbreviations: ACS; acute coronary syndrome, PND; paroxysmal nocturnal dyspnoea Chest pain characteristics of the study group Abbreviations: ACS; acute coronary syndrome, PND; paroxysmal nocturnal dyspnoea Patients diagnosed with unstable angina (n = 9, 12.2%), NSTEMI (n = 35, 47.3%) and STEMI (n = 11, 14.9%) were considered to have ACS. Based on history, ECG findings and hs-cTn I value, ACS was excluded as the cause for chest pain in the rest (Table 3). Table 3Diagnosis of patients with chest pain at discharge from the wardDiagnosisNumberPercentage (%)Stable angina56.8Unstable angina912.2NSTEMI3547.3STEMI1114.9ACS5574.3Non-ACS1925.7Abbreviations – NSTEMI – non-ST segment elevated myocardial infarction, STEMI – ST segment elevated myocardial infarction, ACS – Acute coronary syndrome Table 3Diagnosis of patients with chest pain at discharge from the wardDiagnosisNumberPercentage (%)Stable angina56.8Unstable angina912.2NSTEMI3547.3STEMI1114.9ACS5574.3Non-ACS1925.7Abbreviations – NSTEMI – non-ST segment elevated myocardial infarction, STEMI – ST segment elevated myocardial infarction, ACS – Acute coronary syndrome Diagnosis of patients with chest pain at discharge from the ward Abbreviations – NSTEMI – non-ST segment elevated myocardial infarction, STEMI – ST segment elevated myocardial infarction, ACS – Acute coronary syndrome Among patients diagnosed with ACS at the time of discharge, MACE was observed in 48 patients (87.3%) at six weeks. In contrast, only 3 patients (15.8%) in the non-ACS group experienced MACE. The types of MACE observed in each group are presented in Table 4. Table 4Occurrence of MACE at 6 weeksACS (55)Non-ACS (19)None712.7%1684.2%Myocardial infarction4785.5%315.8%PCI47.3%15.3%CABG00Death35.5%0Total MACE4887.3%315.8%Abbreviations – PCI – Percutaneous coronary interventions, CABG – Coronary artery bypass grafting, MACE – Major adverse cardiac events Table 4Occurrence of MACE at 6 weeksACS (55)Non-ACS (19)None712.7%1684.2%Myocardial infarction4785.5%315.8%PCI47.3%15.3%CABG00Death35.5%0Total MACE4887.3%315.8%Abbreviations – PCI – Percutaneous coronary interventions, CABG – Coronary artery bypass grafting, MACE – Major adverse cardiac events Occurrence of MACE at 6 weeks Abbreviations – PCI – Percutaneous coronary interventions, CABG – Coronary artery bypass grafting, MACE – Major adverse cardiac events The total HEART score, GRACE 2.0 and TIMI scores were calculated retrospectively for each patient and categorized into low-risk, intermediate-risk, and high-risk groups for risk stratification (Table 5). The risk categories were compared against the diagnosis at discharge. GRACE score was not available in two patients due to missing data, and these patients were removed from analysis involving the GRACE score. The distribution of patients among risk categories between the ACS and non-ACS group is illustrated in Table 5. Table 5Distribution of the HEART, GRACE and TIMI score risk categories based on the diagnosis of ACS and occurrence of MACEHEART score risk categoryACS (n = 55)Non-ACS (n = 19)Low (0–3)00%842.1%Intermediate (4–6)2647.3%1157.9%High (7–10)2952.7%00% GRACE score risk category ACS (n = 53)Non-ACS (n = 19)Low (≤ 108)1630.2%1473.7%Intermediate (109–140)2750.9%526.3%High (≥ 140)1018.9%00% TIMI score risk category ACS (n = 55)Non-ACS (n = 19)Low (0–2)1629.1%1473.7%Intermediate (3–4)2341.8%421.1%High (5–7)1629.1%15.3% HEART score risk category With MACE (n = 49)Without MACE (n = 25)Low (0–3)0832%Intermediate (4–6)2040.8%1768%High (7–10)2959.2%00% GRACE 2.0 score risk category With MACE (n = 48)Without MACE (n = 24)Low (≤ 108)1633.3%1458.3%Intermediate (109–140)2347.9%937.5%High (≥ 140)918.8%14.2% TIMI score risk category With MACE (n = 49)Without MACE (n = 25)Low (0–2)1326.5%1768%Intermediate (3–4)2142.9%624%High (5–7)1530.6%28%Abbreviations – ACS – Acute coronary syndrome, MACE - Major adverse cardiac events Table 5Distribution of the HEART, GRACE and TIMI score risk categories based on the diagnosis of ACS and occurrence of MACEHEART score risk categoryACS (n = 55)Non-ACS (n = 19)Low (0–3)00%842.1%Intermediate (4–6)2647.3%1157.9%High (7–10)2952.7%00% GRACE score risk category ACS (n = 53)Non-ACS (n = 19)Low (≤ 108)1630.2%1473.7%Intermediate (109–140)2750.9%526.3%High (≥ 140)1018.9%00% TIMI score risk category ACS (n = 55)Non-ACS (n = 19)Low (0–2)1629.1%1473.7%Intermediate (3–4)2341.8%421.1%High (5–7)1629.1%15.3% HEART score risk category With MACE (n = 49)Without MACE (n = 25)Low (0–3)0832%Intermediate (4–6)2040.8%1768%High (7–10)2959.2%00% GRACE 2.0 score risk category With MACE (n = 48)Without MACE (n = 24)Low (≤ 108)1633.3%1458.3%Intermediate (109–140)2347.9%937.5%High (≥ 140)918.8%14.2% TIMI score risk category With MACE (n = 49)Without MACE (n = 25)Low (0–2)1326.5%1768%Intermediate (3–4)2142.9%624%High (5–7)1530.6%28%Abbreviations – ACS – Acute coronary syndrome, MACE - Major adverse cardiac events Distribution of the HEART, GRACE and TIMI score risk categories based on the diagnosis of ACS and occurrence of MACE Abbreviations – ACS – Acute coronary syndrome, MACE - Major adverse cardiac events An exploratory analysis was performed using all three scores to assess their ability to diagnose ACS and predict MACE within six weeks. In intermediate and high-risk categories of all the scores, the score was classified as positive, while in the low-risk category it was classified as negative. Cut off values (clinical and optimal) and the sensitivity, specificity, PPV and NPV of each score in predicting a diagnosis of ACS and the risk of occurrence of MACE is illustrated in Table 6. Table 6Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of HEART, GRACE 2.0 and TIMI scores in predicting ACS and risk of MACE at the clinical cut off and the optimal thresholdPrediction of ACSHEART (N = 74)GRACE 2.0 (N = 72)TIMI (N = 74) At a threshold of 4 At a Threshold of 109 At a threshold of 3 Sensitivity % (95% CI) 100 (94–100) Sensitivity % (95% CI) 70 (57–82) Sensitivity % (95% CI) 71 (59–83) Specificity % (95% CI) 42 (20–64) Specificity % (95% CI) 74 (54–94) Specificity % (95% CI) 74 (54–94) PPV % (95% CI) 83 (74–92) PPV % (95% CI) 88 (78–98) PPV % (95% CI) 89 (78–98) NPV % (95% CI) 100% (63–100) NPV % (95% CI) 47 (29–65) NPV % (95% CI) 47 (29–65) Youden’s optimal threshold 5.5 (6) Youden’s optimal threshold 74 Youden’s optimal threshold 1.5 (2) Sensitivity % (95% CI) 73 (61–85) Sensitivity % (95% CI) 96 (91–100) Sensitivity % (95% CI) 87 (78–96) Specificity % (95% CI) 90 (77–100) Specificity % (95% CI) 58 (36–80) Specificity % (95% CI) 58 (36–80) PPV % (95% CI) 95 (88–100) PPV % (95% CI) 86 (77–95) PPV % (95% CI) 86 (77–95) NPV % (95% CI) 53 (36–70) NPV % (95% CI) 85 (66–100) NPV % (95% CI) 61 (41–81) Prediction of MACE HEART ( N = 74) GRACE 2.0 ( N = 72) TIMI ( N = 74) At a threshold of 4 At a Threshold of 109 At a threshold of 3 Sensitivity % (95% CI) 100 (93–100) Sensitivity % (95% CI) 67 (53–80) Sensitivity % (95% CI) 74 (61–86) Specificity % (95% CI) 32 (14–50) Specificity % (95% CI) 58 (39–78) Specificity % (95% CI) 68 (50–86) PPV % (95% CI) 74 (64–85) PPV % (95% CI) 76 (63–89) PPV % (95% CI) 82 (70–93) NPV % (95% CI) 100 (63–100) NPV % (95% CI) 47 (29–65) NPV % (95% CI) 57 (39–74) Youden’s optimal threshold 5.5 (6) Youden’s optimal threshold 112.5 (113) Youden’s optimal threshold 1.5 (2) Sensitivity % (95% CI) 80 (69–91) Sensitivity % (95% CI) 65 (52–78) Sensitivity % (95% CI) 90 (82–98) Specificity % (95% CI) 88 (75–100) Specificity % (95% CI) 71 (53–89) Specificity % (95% CI) 52 (32–72) PPV % (95% CI) 93 (85–100) PPV % (95% CI) 82 (71–93) PPV % (95% CI) 79 (67–91) NPV % (95% CI) 69 (50–88) NPV % (95% CI) 50 (32–68) NPV % (95% CI) 72 (52–92)Abbreviations – ACS – Acute coronary syndrome, PPV – Positive predictive value, NPV – Negative predictive value Table 6Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of HEART, GRACE 2.0 and TIMI scores in predicting ACS and risk of MACE at the clinical cut off and the optimal thresholdPrediction of ACSHEART (N = 74)GRACE 2.0 (N = 72)TIMI (N = 74) At a threshold of 4 At a Threshold of 109 At a threshold of 3 Sensitivity % (95% CI) 100 (94–100) Sensitivity % (95% CI) 70 (57–82) Sensitivity % (95% CI) 71 (59–83) Specificity % (95% CI) 42 (20–64) Specificity % (95% CI) 74 (54–94) Specificity % (95% CI) 74 (54–94) PPV % (95% CI) 83 (74–92) PPV % (95% CI) 88 (78–98) PPV % (95% CI) 89 (78–98) NPV % (95% CI) 100% (63–100) NPV % (95% CI) 47 (29–65) NPV % (95% CI) 47 (29–65) Youden’s optimal threshold 5.5 (6) Youden’s optimal threshold 74 Youden’s optimal threshold 1.5 (2) Sensitivity % (95% CI) 73 (61–85) Sensitivity % (95% CI) 96 (91–100) Sensitivity % (95% CI) 87 (78–96) Specificity % (95% CI) 90 (77–100) Specificity % (95% CI) 58 (36–80) Specificity % (95% CI) 58 (36–80) PPV % (95% CI) 95 (88–100) PPV % (95% CI) 86 (77–95) PPV % (95% CI) 86 (77–95) NPV % (95% CI) 53 (36–70) NPV % (95% CI) 85 (66–100) NPV % (95% CI) 61 (41–81) Prediction of MACE HEART ( N = 74) GRACE 2.0 ( N = 72) TIMI ( N = 74) At a threshold of 4 At a Threshold of 109 At a threshold of 3 Sensitivity % (95% CI) 100 (93–100) Sensitivity % (95% CI) 67 (53–80) Sensitivity % (95% CI) 74 (61–86) Specificity % (95% CI) 32 (14–50) Specificity % (95% CI) 58 (39–78) Specificity % (95% CI) 68 (50–86) PPV % (95% CI) 74 (64–85) PPV % (95% CI) 76 (63–89) PPV % (95% CI) 82 (70–93) NPV % (95% CI) 100 (63–100) NPV % (95% CI) 47 (29–65) NPV % (95% CI) 57 (39–74) Youden’s optimal threshold 5.5 (6) Youden’s optimal threshold 112.5 (113) Youden’s optimal threshold 1.5 (2) Sensitivity % (95% CI) 80 (69–91) Sensitivity % (95% CI) 65 (52–78) Sensitivity % (95% CI) 90 (82–98) Specificity % (95% CI) 88 (75–100) Specificity % (95% CI) 71 (53–89) Specificity % (95% CI) 52 (32–72) PPV % (95% CI) 93 (85–100) PPV % (95% CI) 82 (71–93) PPV % (95% CI) 79 (67–91) NPV % (95% CI) 69 (50–88) NPV % (95% CI) 50 (32–68) NPV % (95% CI) 72 (52–92)Abbreviations – ACS – Acute coronary syndrome, PPV – Positive predictive value, NPV – Negative predictive value Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of HEART, GRACE 2.0 and TIMI scores in predicting ACS and risk of MACE at the clinical cut off and the optimal threshold Abbreviations – ACS – Acute coronary syndrome, PPV – Positive predictive value, NPV – Negative predictive value In this exploratory analysis, ROC curves were developed to compare the three scores for a diagnosis of ACS and predicting risk of MACE separately (Fig. 3). The AUC-ROC of the three scores are illustrated in Table 7. The AUC-ROC of the scores were compared against each other with the Delong’s test. Compared to GRACE 2.0 score and TIMI score, HEART score had a higher AUC for the diagnosis of ACS as well as in predicting risk of MACE. The variations in the AUC-ROCs were not statistically significant, apart from the differences between the HEART score and GRACE 2.0, and the TIMI score in predicting MACE. The Youden’s index was also calculated for each score from the ROC curves to demonstrate the optimal threshold for each score and the sensitivity, specificity, PPV and NPV of each score was calculated at this optimal threshold (Table 6). Fig. 3ROC curves (N = 72) of HEART, GRACE 2.0 and TIMI scores (a) for a diagnosis of ACS (b) for the occurrence of MACE at 6 weeks ROC curves (N = 72) of HEART, GRACE 2.0 and TIMI scores (a) for a diagnosis of ACS (b) for the occurrence of MACE at 6 weeks Table 7AUC-ROC of HEART score, GRACE 2.0 score and TIMI score for a diagnosis of ACS and predicting risk of MACE and statistical significance of difference between the ROC curves (DeLong’s method)ACSMACEAUC-ROC95% CIAUC- ROC95% CIHEART (N = 74)0.8890.8171–0.96090.90530.8437–0.9669GRACE (N = 72)0.8050.6758–0.93490.7210.5934–0.8493TIMI (N = 74)0.8120.6961–0.92780.76730.6467–0.888ACSMACEDelong’s testp value95% CIDelong’s testp value95% CI HEART vs. GRACE 0.1713-0.0391–0.21980.01260.0415–0.3248 HEART vs. TIMI 0.08029-0.0093–0.16340.00620.0392–0.2368 GRACE vs. TIMI 0.8009-0.1394–0.10760.6091-0.1666–0.0645Abbreviations – ACS - Acute coronary syndrome, MACE – Major advance cardiac events, CI – Confidence interval Table 7AUC-ROC of HEART score, GRACE 2.0 score and TIMI score for a diagnosis of ACS and predicting risk of MACE and statistical significance of difference between the ROC curves (DeLong’s method)ACSMACEAUC-ROC95% CIAUC- ROC95% CIHEART (N = 74)0.8890.8171–0.96090.90530.8437–0.9669GRACE (N = 72)0.8050.6758–0.93490.7210.5934–0.8493TIMI (N = 74)0.8120.6961–0.92780.76730.6467–0.888ACSMACEDelong’s testp value95% CIDelong’s testp value95% CI HEART vs. GRACE 0.1713-0.0391–0.21980.01260.0415–0.3248 HEART vs. TIMI 0.08029-0.0093–0.16340.00620.0392–0.2368 GRACE vs. TIMI 0.8009-0.1394–0.10760.6091-0.1666–0.0645Abbreviations – ACS - Acute coronary syndrome, MACE – Major advance cardiac events, CI – Confidence interval AUC-ROC of HEART score, GRACE 2.0 score and TIMI score for a diagnosis of ACS and predicting risk of MACE and statistical significance of difference between the ROC curves (DeLong’s method) Abbreviations – ACS - Acute coronary syndrome, MACE – Major advance cardiac events, CI – Confidence interval Discussion Interpretation This exploratory analysis, of the distribution of risk categories, demonstrated that HEART score outperformed other scores in identifying low and high-risk patients in diagnosing ACS and predicting MACE. The HEART score demonstrated strong diagnostic accuracy, with an AUC of 0.889 for the diagnosis of ACS. The optimal threshold of the score was found to be 6. At this threshold, a sensitivity of 72.7% with a PPV of 83.3%, shows that the HEART score can be used to complement a diagnosis of ACS. A NPV of 100% with a specificity of 89.5% at the optimal threshold indicates that the score has good discriminatory ability in ruling out ACS. Further, the HEART score demonstrated excellent discriminatory power in risk prediction for six-week MACE, with an AUC of 0.9053. At the optimal threshold of 6, the score accurately identified patients at high risk of MACE (PPV = 93%) while maintaining its power in safely identifying patients at low risk of MACE (Sensitivity = 80%, Specificity = 88%). Few studies over the years have attempted to evaluate the diagnostic and risk stratification accuracy of GRACE score in patients with ACS [24, 25], although it was originally intended for mortality prediction [13]. Our exploratory results indicated a strong diagnostic accuracy of the GRACE 2.0 score for a diagnosis of ACS (AUC-ROC = 0.805). At the optimal threshold of 74, the score was best for ruling in ACS (Sensitivity = 96%, PPV = 86%). But it can be used to safely rule out ACS as well (NPV = 85%). TIMI score was similarly developed for the purpose of predicting mortality and the risk of ischemic events in patients diagnosed with ACS [18]. Few studies have evaluated the applicability of the score in patients with undifferentiated chest pain [23, 26]. With an AUC-ROC of 0.812 for a diagnosis of ACS, our results demonstrated that TIMI score, has a good diagnostic accuracy in ACS, at an optimal threshold of 2 (Sensitivity = 87%, PPV = 86%). Given the lack of statistical significance in the difference of AUC-ROCs of the three scores, their diagnostic accuracy is comparable. AUC-ROC of 0.721 and 0.7673 respectively for the GRACE 2.0 and TIMI scores, demonstrates that these scores have good discriminatory power, in risk prediction for occurrence of MACE at six weeks. At an optimal threshold of 113, which is different to the one used for predicting ACS, GRACE 2.0 showed moderate predictive accuracy (Sensitivity = 65%, PPV = 82%) of MACE at six weeks. TIMI score was better at identifying patients at high risk of MACE (Sensitivity = 90%, PPV = 79%) at an optimal threshold of 2. However, the HEART score outperformed GRACE and TIMI scores when AUCs of the three scores were compared, and the difference was statistically significant. The primary aim of this pilot study was to explore the feasibility of a larger study to evaluate the diagnostic accuracy of the HEART, GRACE 2.0 and TIMI scores in patients with chest pain. Therefore, the statistically significant results obtained from this exploratory analysis confirms the feasibility of a larger study. Limitations Our exploratory results demonstrated a MACE rate of 66% in this cohort of patients, in contrast to the usual 10–20% MACE rate in previous similar validation studies. This cannot be explained solely with the fact that the study took place at a tertiary care hospital, because the hospital received primary and secondary care admissions as well as tertiary care admissions. However, the set of patients with clinically obvious non cardiac chest pain did not undergo cardiac Troponin testing, and hence were not included in the study, and this might have partially contributed to this high MACE rate which may have artificially elevated the observed event rate. The tertiary care centre where this study was conducted was not a designated primary PCI centre at the time of data collection. Hence the patients diagnosed with STEMI who required urgent PCI were transferred to external facilities and were not included in the study group. However, in the Sri Lankan context, only a limited number of patients are accepted for primary PCI. Hence the exclusion of these patients is unlikely to have had a significant impact on the overall results of the study. Although current guidelines recommend a pharmaco-invasive strategy for patients diagnosed with STEMI, in the Sri Lankan healthcare setup, where resources are limited, not all patients who receive thrombolysis proceed to coronary angiogram routinely. In addition, CABG facilities are limited, often leading to long wait times beyond the study’s follow-up period. Therefore, the actual incidence of MACE might be slightly higher than reported in this study. However, since MACE included myocardial infarction as well as PCI and CABG, the patients who warranted these modalities of treatment were classified as positive for MACE. Hence the effect on the overall results was considered to be minimum. The study included all consecutive admissions with chest pain with a complete data set, irrespective of the final diagnosis, to reflect the full spectrum of real-world emergency department presentations in routine clinical practice, including those subsequently diagnosed with STEMI. This allowed us to evaluate whether these scores correctly classified high-risk patients into the high-risk category, although this approach may introduce a degree of spectrum effect. Although the sample size was adequate for exploratory analysis, it remains insufficient for definitive validation, and findings should be interpreted cautiously. Sample size calculations based on Hanley & McNeil and Obuchowski’s methods confirmed that ≥ 50 participants would be adequate to detect differences in AUC ≥ 0.1 with 80% power at α = 0.05. Nonetheless, a larger, multicentre study would enhance statistical power and improve the generalizability of the findings across the broader Sri Lankan population. Usability of the model in the context of current care Although the HEART score is already used in emergency care settings in Sri Lanka, it has not been formally validated for the local population. This study provides exploratory evidence supporting its validity among Sri Lankans. By utilizing HEART score at the EDs, ACS can be ruled out rapidly in low-risk patients presenting with chest pain and hence provide the rationale for early discharge of the low-risk patients from the ED, helping to reduce overcrowding at the ED and inpatient wards. Use of the GRACE 2.0 score in the ED may not be practical given the complex nature of the score as well as the requirement for serum creatinine. However, TIMI score may be used in the ED, considering its less cumbersome nature. In resource limited settings like in Sri Lanka, the use of risk stratification tools helps in cost effective utilization of the healthcare facilities. Given the excellent discriminative power of the HEART score to predict MACE, the HEART score can also serve as an effective communication tool between emergency physicians and cardiologists when determining urgency for referral and intervention. Refined risk stratification of patients presenting with chest pain extends beyond the emergency department. Although GRACE 2.0 score is cumbersome to be used at an emergency setting, it can be used easily in a ward setup. Identifying high risk patients assists clinical teams in therapeutic decision making and avoiding over treatment in low-risk patients. Our exploratory results demonstrate the feasibility of the utilization of all three scores in the Sri Lankan population, for this purpose. These scores assist clinical teams in, treatment escalation decisions, establishing monitoring frequencies, identifying patients with residual or persistent risk despite the standard treatment who might benefit with novel treatment options etc. Therefore, the utilization of these scores needs to be encouraged in developing countries like in Sri Lanka, especially in public healthcare facilities for the optimal utilization of the limited resources.