The AITRIAGE technology, which uses biological parameters to calculate patients’ risk score within minutes, could transform patient management in the emergency department.
by Dr Ong Eng Hock, Marcus
Chest pain is one of the leading causes of visits to the Emergency Department (ED). The patients at the ED could have varied levels of risk of complications in the acute phase of treatment, which is the period less than 72 hours from the occurrence of the chest pains. Physicians try their best to identify and predict high-risk patients for timely intervention of treatable conditions, and low-risk patients for management to reduce the necessity for excessive investigation and monitoring.
Challenges for Triaging Chest Pain
Physicians have a difficult task of identifying which chest pain patients to admit, which to continue monitoring, and which to discharge. Among the most popular tools to assess the risk of major adverse cardiac events (MACE) in chest pain patients, are the Thrombolysis in Myocardial Infarction (TIMI), Global Registry of Acute Coronary Events (GRACE) and history, ECG, age, risk factors, and initial troponin (HEART) scores.1,2 These clinical scores usually require laboratory results that take a few hours to be processed. Furthermore, patients with non-diagnostic initial electrocardiograms (ECG) are often put through a prolonged period of evaluation (eight hours or more), with serial ECGs and blood tests required. This may be challenge for the busy EDs, where manpower and resources are limited.
The Role of AITRIAGE
Working in the emergency department, Professor Marcus Ong understands the benefits of being able to predict patient outcomes quickly, with a high degree of accuracy for clinical triage. This is the process of determining the priority of patients’ treatments based on the severity of their condition. With a view towards transforming patient management, especially in the ED, the research and development of AITRIAGE technology was pursued. This was done through machine learning to optimise the performance of the predictive algorithms, which identify at-risk patients quickly by using real-time patient data. AITRIAGE was designed for the rapid and objective stratification of chest pain patients by assessing the risk of 30-day MACE. The technology uses patient records such as ECG, blood pressure and peripheral capillary oxygen saturation (SPO2), together with Heart Rate Variability (HRV) parameters to calculate a risk score within minutes. This allows clinicians to possibly identify chest pain patients who are at high risk of a MACE and require priority investigation and management. For low-risk patients, this means that unnecessary laboratory tests can be avoided, and a shorter diagnostic workup can be ordered instead. The accelerated triage workflow using AITRIAGE could potentially shorten the care duration for low-risk patients from eight-12 hours to two hours, reducing ED overcrowding and saving hospital resources.
Artificial Intelligence (AI) for Risk Prediction
At the core of the AITRIAGE application is a novel AI-based risk stratification algorithm that was developed at Singapore General Hospital (SGH). The algorithm leverages key inputs from the patient’s HRV, including time-domain, frequency domain and non-linear parameters to increase its predictive accuracy.
Analysing a patient’s HRV is a potentially useful approach that can be applied at the point of clinical assessment and is currently undergoing clinical trials at several hospitals in Singapore. Reduced HRV has been shown to be an independent predictor of cardiac death and mortality after myocardial infarction (MI)3 and has been shown to correlate with poor short-term outcomes for successfully resuscitated patients with out-of-hospital cardiac arrest.4 Altered HRV spectrum has also been found to be an indicator of severity in congestive heart failure,5 hypertension,6 coronary artery disease7 and MI.8
At present, the system has been clinically validated to outperform popular risk scoring tools such as TIMI, MEWS and GRACE scores, and is comparable to the HEART score even without the input of the patient’s troponin levels. Furthermore, new patient data can be processed by AITRIAGE’s “machine learning” algorithm to further enhance the predictive accuracy of the risk stratification model.
Clinical Impact of AITRIAGE
The AITRIAGE system is suitable for large-scale deployment across healthcare facilities with its small technological footprint. This technology has the potential to distinguish those who require urgent treatment of serious problems from those who do not have life-threatening conditions, in a quick and cost-effective manner. In the ED, where resources are scarce, patients who are triaged to have a low risk of progressing to a MACE can potentially be discharged faster with a higher certainty, freeing up valuable bed space and manpower. Meanwhile, patients at a high risk of a MACE can be identified earlier and intervention can be delivered sooner. AITRIAGE is also being developed for conditions such as sepsis, trauma and heart failure.
Beyond the ED, AITRIAGE could also be potentially utilised for ambulatory monitoring and risk assessment in pre-hospital care, Intensive Care Units, hospital wards, clinics and home-based monitoring. This could especially improve the efficiency of healthcare systems in larger, developing countries where specialist care is generally centred around the cities.
The AITRIAGE technology is currently undergoing multi-centre clinical trials for further validation and improvements. The team plans to make it commercially available by 2021, following regulatory approvals. [APBN]
- Backus BE, Six AJ, Kelder JH, Gibler WB, Moll FL, Doevendans PA. Risk scores for patients with chest pain: Evaluation in the emergency department. Curr Cardiol Rev 2011;7:2–8.
- Liu N, Ng JCJ, Ting CE, Sakamoto JT, Ho AFW, Koh ZX, et al. Clinical scores for risk stratification of chest pain patients in the emergency department: an updated systematic review. J Emerg Crit Care Med 2018;2:1–16.
- Bigger J, Fleiss J, Steinman R, Rolnitzky L, Kleiger R, Rottman J. Frequency domain measures of heart period variability and mortality after myocardial infarction. Circulation. 1992;85(1):164-71. PubMed PMID: WOS:A1992GY58200020.
- Chen W, Tsai T, Huang C, Chen J, Kuo C. Heart rate variability predicts short-term outcome for successfully resuscitated patients with out-of-hospital cardiac arrest. Resuscitation. 2009;80(10):1114-8. doi: 10.1016/j.resuscitation.2009.06.020. PubMed PMID: WOS:000271336400007.
- Guzzetti S, Mezzetti S, Magatelli R, Porta A, De Angelis G, Rovelli G, et al. Linear and non-linear 24 h heart rate variability in chronic heart failure. Auton Neurosci. 2000;86(1-2):114-9. PubMed PMID: 11269916.
- Guzzetti S, Piccaluga E, Casati R, Cerutti S, Lombardi F, Pagani M, et al. Sympathetic predominance in essential hypertension: a study employing spectral analysis of heart rate variability. Journal of Hypertension. 1988;6(9):711-7. PubMed PMID: 3183374.
- van Boven AJ, Jukema JW, Haaksma J, Zwinderman AH, Crijns HJ, Lie KI. Depressed heart rate variability is associated with events in patients with stable coronary artery disease and preserved left ventricular function. REGRESS Study Group. Am Heart J. 1998;135(4):571-6. PubMed PMID: 9539469.
- Poulsen SH, Jensen SE, Moller JE, Egstrup K. Prognostic value of left ventricular diastolic function and association with heart rate variability after a first acute myocardial infarction. Heart. 2001;86(4):376-80. PubMed PMID: 11559672.
About the Author
Ong Eng Hock Marcus, Ph.D.
Professor and Director Health Services and Systems Research Programme Duke-NUS Medical School
Senior Consultant, Director of Research and Clinician Scientist
Department of Emergency Medicine, Singapore
General Hospital Head, Data Analytics, Health Services Research Center, SingHealth Services Medical Director, Unit for Pre-hospital Emergency Care
Senior Consultant, Hospital Services Division, Ministry of Health
Chairman, Pan Asian Resuscitation Outcomes Study