Triage Model Research Articles

Automatic Bug Triaging via Deep Reinforcement Learning

Software maintenance and evolution account for approximately 90% of the software development process (e.g., implementation, testing, and maintenance). Bug triaging refers to an activity where developers diagnose, fix, test, and document bug reports during software development and maintenance to improve the speed of bug repair and project progress. However, the large number of bug reports submitted daily increases the triaging workload, and open-source software has a long maintenance cycle. Meanwhile, the developer activity is not stable and changes significantly during software development. Hence, we propose a novel bug triaging model known as auto bug triaging via deep reinforcement learning (BT-RL), which comprises two models: a deep multi-semantic feature (DMSF) fusion model and an online dynamic matching (ODM) model. In the DMSF model, we extract relevant information from bug reports to obtain high-quality feature representation. In the ODM model, through bug report analysis and developer activities, we use a strategy based on the reinforcement learning framework, through which we perform training while learning and recommend developers for bug reports. Extensive experiments on open-source datasets show that the BT-RL method outperforms state-of-the-art methods in bug triaging.

Leveraging Large-Scale Electronic Health Records and Interpretable Machine Learning for Clinical Decision Making at the Emergency Department: Protocol for System Development and Validation.

BackgroundThere is a growing demand globally for emergency department (ED) services. An increase in ED visits has resulted in overcrowding and longer waiting times. The triage process plays a crucial role in assessing and stratifying patients’ risks and ensuring that the critically ill promptly receive appropriate priority and emergency treatment. A substantial amount of research has been conducted on the use of machine learning tools to construct triage and risk prediction models; however, the black box nature of these models has limited their clinical application and interpretation.ObjectiveIn this study, we plan to develop an innovative, dynamic, and interpretable System for Emergency Risk Triage (SERT) for risk stratification in the ED by leveraging large-scale electronic health records (EHRs) and machine learning.MethodsTo achieve this objective, we will conduct a retrospective, single-center study based on a large, longitudinal data set obtained from the EHRs of the largest tertiary hospital in Singapore. Study outcomes include adverse events experienced by patients, such as the need for an intensive care unit and inpatient death. With preidentified candidate variables drawn from expert opinions and relevant literature, we will apply an interpretable machine learning–based AutoScore to develop 3 SERT scores. These 3 scores can be used at different times in the ED, that is, on arrival, during ED stay, and at admission. Furthermore, we will compare our novel SERT scores with established clinical scores and previously described black box machine learning models as baselines. Receiver operating characteristic analysis will be conducted on the testing cohorts for performance evaluation.ResultsThe study is currently being conducted. The extracted data indicate approximately 1.8 million ED visits by over 810,000 unique patients. Modelling results are expected to be published in 2022.ConclusionsThe SERT scoring system proposed in this study will be unique and innovative because of its dynamic nature and modelling transparency. If successfully validated, our proposed solution will establish a standard for data processing and modelling by taking advantage of large-scale EHRs and interpretable machine learning tools.International Registered Report Identifier (IRRID)DERR1-10.2196/34201

Predictive values of AI-based triage model in suboptimal CT pulmonary angiography

PurposeWe evaluated and compared performance of an acute pulmonary embolism (PE) triaging artificial intelligence (PE-AI) model in suboptimal and optimal CT pulmonary angiography (CTPA). MethodsIn an IRB approved, retrospective study we identified 104 consecutive, suboptimal CTPA which were deemed as suboptimal for PE evaluation in radiology reports due to motion, artifacts and/or inadequate contrast enhancement. We enriched this dataset, with additional 226 optimal CTPA (over same timeframe as suboptimal CTPA) with and without PE. Two thoracic radiologists (ground truth) independently reviewed all 330 CTPA for adequacy (to assess PE down to distal segmental level), reason for suboptimal CTPA (artifacts or poor contrast enhancement), as well as for presence and location of PE. CT values (HU) were measured in the main pulmonary artery. Same attributes were assessed in 80 patients who had repeat or follow-up CTPA following suboptimal CTPA. All CTPA were processed with the PE-AI (Aidoc). ResultsAmong 104 suboptimal CTPA (mean age ± standard deviation 56 ± 15 years), 18/104 (17%) were misclassified as suboptimal for PE evaluation in their radiology reports but relabeled as optimal on ground truth evaluation. Of 226 optimal CTPA, 47 (21%) were reclassified as suboptimal CTPA. PEs were present in 97/330 CTPA. PE-AI had similar performance on suboptimal CTPA (sensitivity 100%; specificity 89%; AUC 0.89, 95% CI 0.80–0.98) and optimal CTPA (sensitivity 96%; specificity 92%; AUC 0.87, 95% CI 0.81–0.93). ConclusionSuboptimal CTPA examinations do not impair the performance of PE-AI triage model; AI retains clinically meaningful sensitivity and high specificity regardless of diagnostic quality.

Triage in the time of COVID-19

Triage in the time of COVID-19

Times to endovascular treatment following two triage models

Times to endovascular treatment following two triage models

Service evaluation of the South Wales police control room mental health triage model: outcomes achieved, lessons learned and next steps

PurposeSouth Wales Police Mental Health (MH) Triage service was initiated to meet the Welsh Government MH priority of early intervention to prevent MH crisis. Community Psychiatric Nurses, based in the control-room, provide advice to police and control room staff on the management of MH-related incidents. The purpose of this paper is to evaluate the first 12 months of operation (January-December 2019).Design/methodology/approachService evaluation of the first 12 months of operation (January–December 2019). Data were analysed in relation to: MH incidents; repeat callers; Section (S)136 use/assessment outcomes. Police, health staff and triage service users were interviewed and surveyed to capture their opinions of the service.FindingsPolicing areas with high engagement in triage saw reductions in S136 use and estimated opportunity costs saving. Triage was considered a valuable service that promoted cross agency collaborations. De-escalation in cases of mental distress was considered a strength. Access to follow-on services was identified as a challenge.Practical implicationsTriage enables a multi-agency response in the management of MH-related incidents. Improving trust between services, with skilled health professionals supporting police decision-making in real time.Originality/valueThere is a gap in the research on the impact of police-related MH triage models beyond the use of S136. This project evaluated the quality of the service, its design and the relationship between health, police and partner agencies during the triage process. Multi-agency assessment of follow-up is needed to measure the long-term impact on services and users.

Abstract NS5: A Telestroke Nurse/Neuroradiologist Model For Extended Window Code Stroke Triage

Background and Purpose: Two distinguishing features of our stroke network include the routine involvement of a telestroke nurse (TSRN) for code stroke activations at non-thrombectomy centers, and immediate availability of neuroradiologist for imaging interpretation. With revascularization treatment decision making for patients presenting beyond 4.5 hours from last known well (LKW) being driven by imaging, we hypothesized that reliance on those resources could substantially decrease the routine involvement of a teleneurologist. Methods: On May 1, 2021, we implemented a workflow for code stroke activations beyond 4.5 hours from LKW that relied on a TSRN supported by a neuroradiologist. Criteria for a code stroke activation included any cortical sign on exam or suspicion for a basilar artery occlusion, as determined by the treating Emergency Medicine (EM) physician. All qualifying patients underwent a CTA with or without CTP imaging. Patients with a target occlusion on CTA confirmed by the neuroradiologist, the neurointerventionalist and neurologist at the thrombectomy center are immediately contacted to join the triaging team. Patients without a target occlusion on CTA are managed without the involvement of a teleneurologist, unless requested by the treating EM physician. Results: In our first 3 months utilizing this workflow, there were 509 extended window code stroke activations. In 64% (n=326) of these cases, a teleneurologist was not initially involved, representing a cumulative 58% absolute decrease in teleneurology consultations when compared to the prior three months. For 95% (n=311) of those cases, a target LVO was not identified. For the 15 LVO cases, the median CTA acquisition to LVO confirmation time was 14 minutes (IQR 11-25.5). Ninety-five percent (n=311) of non-LVO cases were subsequently managed by the EM physician. Of those cases, 90% (n=292) were admitted to the hospital. For 47% (n=138) of admitted cases, the patient had a non-stroke diagnosis. Thirty-nine percent (n=114) of all admitted cases were managed by a hospitalist physician without a neurology consult. Conclusions: A telestroke nurse/neuroradiologist model for extended window code stroke triage can substantially reduce the routine involvement of a teleneurologist.

Abstract TMP39: Impact Of Best Modeled Transport On Hospital Patient Volumes, Quantified Outcomes And Healthcare Cost Savings

Introduction: A previously published conditional probability field-based triage model predicts best transport options for ischemic stroke patients to access Endovascular Therapy. This model predicts whether Drip and Ship or Mothership transport results in a higher probability of excellent outcomes. To inform health system policy, it is critical to understand the change in annual hospital volume, the annual increase in the number of patients expected to have excellent outcomes, and the annual reduction in cost to the healthcare system resulting from changing prehospital transport strategies. Methods: We performed a case study using the stroke system in Alberta, Canada. The population of each 7 km by 7 km grid section in Alberta was obtained through the ArcGIS GeoEnrichment service. This was combined with the annual stroke incidence and large vessel occlusion (LVO) screening tool (ex. LAMS) sensitivity and positive predictive value to model the expected number of suspected LVO patients transported by ambulance. Using transport to the closest hospital as a baseline, the resultant annual change in hospital volume was determined. This was used to calculate the annual increase in the number of patients that will have excellent outcomes and resultant cost savings to the healthcare system from average post-stroke cost data. Results: The change in annual hospital volumes for Alberta is shown in the Figure. There was an increase of 6.65 patients predicted to have an excellent outcome (90-day mRS 0-1) annually, and an increase of 11.93 patients to have good outcomes (90-day mRS 0-2). This resulted in an annual cost savings of $283,463.86 to the healthcare system. Conclusions: Translating the benefit of an existing conditional probability model into the change in annual hospital volume, increase in the annual number of patients with excellent outcomes, and health system cost avoidance, allows one to fully realize the impact of using a modeled transport protocol to the health system.

Teaching Trauma in Resource-Limited Settings: A Scoping Review of Pediatric Trauma Courses.

Injury remains an important cause of death and disability globally, with 95% of all childhood injury deaths occurring in low- and lower-middle-income countries (LMICs). Pediatric trauma training, tailored to the resources in LMICs, represents an opportunity to improve such outcomes. We explored the nature of course offerings in pediatric trauma in resource-limited settings. Seven databases were interrogated up to June 12, 2020, to retrieve articles examining pediatric trauma training in LMICs, as defined by the World Bank, without language restrictions. Independent authors reviewed and selected abstracts based on set criteria. Data from included studies was extracted and analyzed. An adapted Critical Appraisal Skills Programme checklist designed for cohort studies was used to assess the risk of bias. After screening 3960 articles for eligibility, 16 were included for final analysis. Course delivery methods included didactic modules, simulations, clinical mentorship, small group discussion, audits, assessments, and feedback. Knowledge acquisition was primarily assessed through pre/post-tests, clinical skills assessments, and self-assessment questionnaires. Twelve studies detailed course content, nine of which were based on the WHO Emergency Triage, Assessment and Treatment model, which is not specific to trauma. The other three studies involved locally developed pediatric trauma-focused training courses, including airway management, head trauma and cervical spine management, thoracic and abdominal trauma, orthopedic trauma, burn and wound management, and shock. Despite being essential to decreasing pediatric trauma morbidity and mortality worldwide, educational programs in pediatric trauma are not a widespread reality in low-and-middle-income countries. The development of accessible and efficient pediatric trauma education programs is critical for improving pediatric trauma quality of care.

Ethics, Personal Responsibility and the Pandemic: A New Triage Paradigm

Background: How much of a role should personal responsibility play in triage criteria? Because voluntarily unvaccinated people are not fulfilling their societal obligations during a pandemic, the ethical principle of justice demands that they reap the egalitarian consequences. These consequences could include lower priority for care, an increasing number of employer and government mandates, and restrictions to entering many entertainment venues. Discussion: Voluntarily unvaccinated individuals increase the chance that the COVID-19 virus will mutate and spread, endangering the entire population, but especially those who cannot get vaccinated for medical reasons, children for whom vaccines have yet to be approved, and older adult and immunocompromised people for whom the vaccine is less effective. When voluntarily unvaccinated individuals seek medical treatment for COVID-19 (94% of patients with COVID-19 in U.S. intensive care units), they use resources needed for those with non–COVID-related illnesses. Conclusions: A method to balance resource allocation between those patients who refuse vaccination and patients who need the same health care resources is necessary. An ethical solution is to give those who are voluntarily unvaccinated a lower priority for admission and for the use of other health care resources. Current in-hospital triage models can easily be modified to accomplish this. This substantive change in practice may encourage more people to get vaccinated.

"Playing God": Analyzing the Bioethics of Triage from Hurricane Katrina and Haiti's Earthquake to Covid-19

As the world experiences more mass-casualty incidents, such as extreme weather events or pandemics, health systems are put under pressure to treat those in need, generating the problem of scarcity of vital resources. The allocation of scarce resources is done via triage, the assigning of priority order to persons on the basis of where resources can be best used, are most needed, or are most likely to achieve success. This article analyzes the ethics of three potential triage models—the survival model, social worth model, and lottery model—as a means of assessing the benefits and drawbacks of the models in relation to the United States healthcare system. Hurricane Katrina, the 2010 Haiti earthquake, and COVID-19 are utilized as case studies to present success and failures of implementation of existing triage models, focusing on how models trade off between efficacy and efficiency. While the initial intent of this research was to determine if there is a superior form of triage that should be adopted to maximize health for U.S. communities in the face of future disasters, as a result of analyses conducted with health and ethics experts, it is apparent that superiority is not the answer. Instead, triage should be treated as a complex and changing system, one that should be evaluated on a case-by-case basis by different providers to create the most optimal solutions.

Predictive Models for Emergency Department Triage using Machine Learning: A Review

Background: Recently, many research groups have tried to develop emergency department triage decision support systems based on big volumes of historical clinical data to differentiate and prioritize patients. Machine learning models might improve the predictive capacity of emergency department triage systems. The aim of this review was to assess the performance of recently described machine learning models for patient triage in emergency departments, and to identify future challenges.

Integrating Telepsychiatry Into Rural Primary Care for Upstream Prevention: Feasibility Case Study of the Faith Net Program.

IntroductionFor decades, there has been a deficit of mental health services in rural areas of the United States. Beyond that longstanding need, the COVID-19 pandemic has reportedly increased the prevalence of unmet mental health needs among adults. Presently, many non-critical but urgent mental health concerns are first identified in rural emergency departments. This report describes the results of a 6-month feasibility case study of a program to integrate telepsychiatric triage “upstream” from emergency departments in rural primary care.MethodsAt routine primary care encounters in a single midwestern rural county, patients at risk for moderate-severe or severe depression, expressing thoughts of self-harm, or otherwise presenting in a way that raised clinical concern for mental or behavioral health, were referred to on-site telepsychiatric triage. Patients whose triage indicated further concern were provided six psychiatric and/or social work encounters for stabilization and treatment.Results68 patients were referred to telepsychiatric triage during the pilot study (.85% of the estimated adult population in the county). Of those, only two had a documented mental/behavioral health diagnosis prior to triage, but 46 were diagnosed with at least one psychiatric disorder during the program.ConclusionsThis model of telepsychiatric triage was feasible in rural primary care and may support identification and mitigation of unmet mental health needs.

Accuracy of the Traditional COVID-19 Phone Triaging System and Phone Triage-Driven Deep Learning Model.

Objectives:During the COVID-19 pandemic, a quick and reliable phone-triage system is critical for early care and efficient distribution of hospital resources. The study aimed to assess the accuracy of the traditional phone-triage system and phone triage-driven deep learning model in the prediction of positive COVID-19 patients.Setting:This is a retrospective study conducted at the family medicine department, Cairo University.Methods:The study included a dataset of 943 suspected COVID-19 patients from the phone triage during the first wave of the pandemic. The accuracy of the phone triaging system was assessed. PCR-dependent and phone triage-driven deep learning model for automated classifications of natural human responses was conducted.Results:Based on the RT-PCR results, we found that myalgia, fever, and contact with a case with respiratory symptoms had the highest sensitivity among the symptoms/ risk factors that were asked during the phone calls (86.3%, 77.5%, and 75.1%, respectively). While immunodeficiency, smoking, and loss of smell or taste had the highest specificity (96.9%, 83.6%, and 74.0%, respectively). The positive predictive value (PPV) of phone triage was 48.4%. The classification accuracy achieved by the deep learning model was 66%, while the PPV was 70.5%.Conclusion:Phone triage and deep learning models are feasible and convenient tools for screening COVID-19 patients. Using the deep learning models for symptoms screening will help to provide the proper medical care as early as possible for those at a higher risk of developing severe illness paving the way for a more efficient allocation of the scanty health resources.

Modeling search and rescue, medical disaster team response and transportation of patients in Ishinomaki city after tsunami disaster

The 2011 Great East Japan Earthquake caused massive devastation and numerous affected people. In particular, the Ishinomaki area of the Miyagi Prefecture was one of the hardest hit areas during the event. In addition, due to tsunami inundation, several medical facilities were destroyed or inactive at the time. The Ishinomaki Red Cross Hospital served as the last bastion to respond and attend injuries and illnesses in the disaster. We aim to build an agent-based model of triage and transportation of patients to the hospital. The individual and joint activities of response teams were considered to build the agent architecture and they were incorporated into the model. We model the first three days of response activities within the central Ishinomaki area and evaluated the number of patients transported to the hospital including their evaluated health status. We discuss here the case closest to the 2011 event and evaluate the strategies for search and rescue and patient transportation to speed up medical response in the aftermath of a disaster.

METASTART: A Systematic Review and Meta-Analysis of the Diagnostic Accuracy of the Simple Triage and Rapid Treatment (START) Algorithm for Disaster Triage.

The goal of disaster triage at both the prehospital and in-hospital level is to maximize resources and optimize patient outcomes. Of the disaster-specific triage methods developed to guide health care providers, the Simple Triage and Rapid Treatment (START) algorithm has become the most popular system world-wide. Despite its appeal and global application, the accuracy and effectiveness of the START protocol is not well-known. The purpose of this meta-analysis was two-fold: (1) to estimate overall accuracy, under-triage, and over-triage of the START method when used by providers across a variety of backgrounds; and (2) to obtain specific accuracy for each of the four START categories: red, yellow, green, and black. A systematic review and meta-analysis was conducted that searched Medline (OVID), Embase (OVID), Global Health (OVID), CINAHL (EBSCO), Compendex (Engineering Village), SCOPUS, ProQuest Dissertations and Theses Global, Cochrane Library, and PROSPERO. The results were expanded by hand searching of journals, reference lists, and the grey literature. The search was executed in March 2020. The review considered the participants, interventions, context, and outcome (PICO) framework and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Accuracy outcomes are presented as means with 95% confidence intervals (CI) as calculated using the binomial method. Pooled meta-analyses of accuracy outcomes using fixed and random effects models were calculated and the heterogeneity was assessed using the Q statistic. Thirty-two studies were included in the review, most of which utilized a non-randomized study design (84%). Proportion of victims correctly triaged using START ranged from 0.27 to 0.99 with an overall triage accuracy of 0.73 (95% CI, 0.67 to 0.78). Proportion of over-triage was 0.14 (95% CI, 0.11 to 0.17) while the proportion of under-triage was 0.10 (95% CI, 0.072 to 0.14). There was significant heterogeneity of the studies for all outcomes (P < .0001). This meta-analysis suggests that START is not accurate enough to serve as a reliable disaster triage tool. Although the accuracy of START may be similar to other models of disaster triage, development of a more accurate triage method should be urgently pursued.

Development and Validation of a Simplified Prehospital Triage Model Using Neural Network to Predict Mortality in Trauma Patients: The Ability to Follow Commands, Age, Pulse Rate, Systolic Blood Pressure and Peripheral Oxygen Saturation (CAPSO) Model.

Objective: Most trauma scoring systems with high accuracy are difficult to use quickly in field triage, especially in the case of mass casualty events. We aimed to develop a machine learning model for trauma mortality prediction using variables easy to obtain in the prehospital setting.Methods: This was a retrospective prognostic study using the National Trauma Data Bank (NTDB). Data from 2013 to 2016 were used for model training and internal testing, and data from 2017 were used for validation. A neural network model (NN-CAPSO) was developed using the ability to follow commands (whether GCS-motor was <6), age, pulse rate, systolic blood pressure (SBP) and peripheral oxygen saturation, and a new score (the CAPSO score) was developed based on logistic regression. To achieve further simplification, a neural network model with the SBP variable removed (NN-CAPO) was also developed. The discrimination ability of different models and scores was compared based on the area under the receiver operating characteristic curve (AUROC). Furthermore, a reclassification table with three defined risk groups was used to compare NN-CAPSO and other models or scores.Results: The NN-CAPSO had an AUROC of 0.911(95% confidence interval 0.909 to 0.913) in the validation set, which was higher than the other trauma scores available for prehospital settings (all p < 0.001). The NN-CAPO and CAPSO score both reached the AUROC of 0.904 (95% confidence interval 0.902 to 0.906), and were no worse than other prehospital trauma scores. Compared with the NN-CAPO, CAPSO score, and the other trauma scores in reclassification tables, NN-CAPSO was found to more accurately classify patients to the right risk groups.Conclusions: The newly developed CAPSO system simplifies the method of consciousness assessment and has the potential to accurately predict trauma patient mortality in the prehospital setting.

To triage or not to triage? The history and evidence for this model of care in psychiatry

SUMMARYTriage wards were introduced as a new model of psychiatric in-patient care in 2004. However, there is limited evidence comparing them with the traditional in-patient models of care. This article reviews the history of triage wards, their principles, the evidence for this model (e.g. length of in-patient stay, readmission rates, staff and patient satisfaction) and the development of assessment wards based on the triage model of care. The evidence shows that the triage model has higher rates of rapid discharge, with a greater proportion of ‘acute care’ performed in the community with the support of home treatment teams. This leads to lower bed occupancy in the triage wards without increased rates of readmission or a worse patient experience of in-patient care. However, overall staff experience was better in the traditional model, given that staff satisfaction rates were lower on locality wards in settings with triage systems in place. Future research should explore the potential impact on home treatment teams, and the rates of serious incidents due to the high number of acutely unwell patients on triage wards.

Development and validation of a prognostic model for early triage of patients diagnosed with COVID-19

We developed a tool to guide decision-making for early triage of COVID-19 patients based on a predicted prognosis, using a Korean national cohort of 5,596 patients, and validated the developed tool with an external cohort of 445 patients treated in a single institution. Predictors chosen for our model were older age, male sex, subjective fever, dyspnea, altered consciousness, temperature ≥ 37.5 °C, heart rate ≥ 100 bpm, systolic blood pressure ≥ 160 mmHg, diabetes mellitus, heart disease, chronic kidney disease, cancer, dementia, anemia, leukocytosis, lymphocytopenia, and thrombocytopenia. In the external validation, when age, sex, symptoms, and underlying disease were used as predictors, the AUC used as an evaluation metric for our model’s performance was 0.850 in predicting whether a patient will require at least oxygen therapy and 0.833 in predicting whether a patient will need critical care or die from COVID-19. The AUCs improved to 0.871 and 0.864, respectively, when additional information on vital signs and blood test results were also used. In contrast, the protocols currently recommended in Korea showed AUCs less than 0.75. An application for calculating the prognostic score in COVID-19 patients based on the results of this study is presented on our website (https://nhimc.shinyapps.io/ih-psc/), where the results of the validation ongoing in our institution are periodically updated.

Optimal Triage for COVID-19 Patients Under Limited Health Care Resources With a Parsimonious Machine Learning Prediction Model and Threshold Optimization Using Discrete-Event Simulation: Development Study.

BackgroundThe COVID-19 pandemic has placed an unprecedented burden on health care systems.ObjectiveWe aimed to effectively triage COVID-19 patients within situations of limited data availability and explore optimal thresholds to minimize mortality rates while maintaining health care system capacity.MethodsA nationwide sample of 5601 patients confirmed with COVID-19 until April 2020 was retrospectively reviewed. Extreme gradient boosting (XGBoost) and logistic regression analysis were used to develop prediction models for the maximum clinical severity during hospitalization, classified according to the World Health Organization Ordinal Scale for Clinical Improvement (OSCI). The recursive feature elimination technique was used to evaluate the maintenance of model performance when clinical and laboratory variables were eliminated. Using populations based on hypothetical patient influx scenarios, discrete-event simulation was performed to find an optimal threshold within limited resource environments that minimizes mortality rates.ResultsThe cross-validated area under the receiver operating characteristic curve (AUROC) of the baseline XGBoost model that utilized all 37 variables was 0.965 for OSCI ≥6. Compared to the baseline model’s performance, the AUROC of the feature-eliminated model that utilized 17 variables was maintained at 0.963 with statistical insignificance. Optimal thresholds were found to minimize mortality rates in a hypothetical patient influx scenario. The benefit of utilizing an optimal triage threshold was clear, reducing mortality up to 18.1%, compared with the conventional Youden index.ConclusionsOur adaptive triage model and its threshold optimization capability revealed that COVID-19 management can be achieved via the cooperation of both the medical and health care management sectors for maximum treatment efficacy. The model is available online for clinical implementation.