Improved interpretable machine learning emergency department triage tool addressing class imbalance

Objective

The Score for Emergency Risk Prediction (SERP) is a novel mortality risk prediction score which leverages machine learning in supporting triage decisions. In its derivation study, SERP-2d, SERP-7d and SERP-30d demonstrated good predictive performance for 2-day, 7-day and 30-day mortality. However, the dataset used had significant class imbalance. This study aimed to determine if addressing class imbalance can improve SERP's performance, ultimately improving triage accuracy.

Methods

The Singapore General Hospital (SGH) emergency department (ED) dataset was used, which contains 1,833,908 ED records between 2008 and 2020. Records between 2008 and 2017 were randomly split into a training set (80%) and validation set (20%). The 2019 and 2020 records were used as test sets. To address class imbalance, we used random oversampling and random undersampling in the AutoScore-Imbalance framework to develop SERP+-2d, SERP+-7d, and SERP+-30d scores. The performance of SERP+, SERP, and the commonly used triage risk scores was compared.

Results

The developed SERP+ scores had five to six variables. The AUC of SERP+ scores (0.874 to 0.905) was higher than that of the corresponding SERP scores (0.859 to 0.894) on both test sets. This superior performance was statistically significant for SERP+-7d (2019: Z = -5.843, p < 0.001, 2020: Z = -4.548, p < 0.001) and SERP+-30d (2019: Z = -3.063, p = 0.002, 2020: Z = -3.256, p = 0.001). SERP+ outperformed SERP marginally on sensitivity, specificity, balanced accuracy, and positive predictive value measures. Negative predictive value was the same for SERP+ and SERP. Additionally, SERP+ showed better performance compared to the commonly used triage risk scores.

Conclusions

Accounting for class imbalance during training improved score performance for SERP+. Better stratification of even a small number of patients can be meaningful in the context of the ED triage. Our findings reiterate the potential of machine learning-based scores like SERP+ in supporting accurate, data-driven triage decisions at the ED.

Electronuclear Transition into a Spatially Modulated Magnetic State in YbRh_{2}Si_{2}

The nature of the antiferromagnetic order in the heavy fermion metal YbRh_{2}Si_{2}, its quantum criticality, and superconductivity, which appears at low mK temperatures, remain open questions. We report measurements of the heat capacity over the wide temperature range 180  μK-80  mK, using current sensing noise thermometry. In zero magnetic field we observe a remarkably sharp heat capacity anomaly at 1.5 mK, which we identify as an electronuclear transition into a state with spatially modulated electronic magnetic order of maximum amplitude 0.1  μ_{B}. We also report results of measurements in magnetic fields in the range 0 to 70 mT, applied perpendicular to the c axis, which show eventual suppression of this order. These results demonstrate a coexistence of a large moment antiferromagnet with putative superconductivity.

Impact of COVID-19 on barriers to dispatcher-assisted cardiopulmonary resuscitation in adult out-of-hospital cardiac arrests in Singapore

OBJECTIVE

Fewer out-of-hospital cardiac arrest (OHCA) patients received bystander cardiopulmonary resuscitation during the COVID-19 pandemic in Singapore. We investigated the impact of COVID-19 on barriers to dispatcher-assisted cardiopulmonary resuscitation (DA-CPR).

METHODS

We reviewed audio recordings of all calls to our national ambulance service call centre during the pandemic (January-June 2020) and pre-pandemic (January-June 2019) periods. Our primary outcome was the presence of barriers to DA-CPR. Multivariable logistic regression was used to assess the effect of COVID-19 on the likelihood of barriers to and performance of DA-CPR, adjusting for patient and event characteristics.

RESULTS

There were 1241 and 1118 OHCA who were eligible for DA-CPR during the pandemic (median age 74 years, 61.6 % males) and pre-pandemic (median age 73 years, 61.1 % males) periods, respectively. Compared to pre-pandemic, there were more residential and witnessed OHCA during the pandemic (87 % vs 84.9 % and 54 % vs 38.1 %, respectively); rates of DA-CPR were unchanged (57.3 % vs 61.1 %). COVID-19 increased the likelihood of barriers to DA-CPR (aOR 1.47, 95 % CI: 1.25-1.74) but not performance of DA-CPR (aOR 0.86, 95 % CI: 0.73 - 1.02). Barriers such as 'patient status changed' and 'caller not with patient' increased during COVID-19 pandemic. 'Afraid to do CPR' markedly decreased during the pandemic; fear of COVID-19 transmission made up 0.5 % of the barriers.

CONCLUSION

Barriers to DA-CPR were encountered more frequently during the COVID-19 pandemic but did not affect callers' willingness to perform DA-CPR. Distancing measures led to more residential arrests with increases in certain barriers, highlighting opportunities for public education and intervention.

An essential review of Singapore's response to out-of-hospital cardiac arrests: improvements over a ten-year period

Care for patients who experience out-of-hospital cardiac arrest (OHCA) has rapidly evolved in the past decade. Increased sophistication of care in the community, emergency medical services (EMS) and hospital setting is associated with improved patient-centred outcomes. Notably, Utstein survival doubled from 11.6% to 23.1% between 2011 and 2016. These achievements involved collaboration between policymakers, clinicians and researchers, and were made possible by a strategic interplay of policy, research and implementation. We review the development and current state of OHCA in Singapore using primary population-based data from the Pan-Asian Resuscitation Outcomes Study and an unstructured search of research databases. We discuss the roles of important milestones in policy, community, dispatch, EMS and hospital interventions. Finally, we relate these interventions to relevant processes and outcomes, such as the relationship between the strategic implementation of bystander cardiopulmonary resuscitation and placement of automated external defibrillator with return of spontaneous circulation, survival to discharge and survival with favourable neurological outcomes.

Effect of coronavirus infection on the human heart: A scoping review

BACKGROUND

The global coronavirus disease 2019 pandemic has highlighted the importance of understanding the cardiovascular implications of coronavirus infections, with more severe disease in those with cardiovascular co-morbidities, and resulting cardiac manifestations such as myocardial injury, arrhythmias, and heart failure.

DESIGN

A systematic review of the current knowledge on the effects of coronavirus infection on the cardiovascular system in humans was performed and results were summarized.

METHODS

Databases such as MEDLINE, EMBASE, CENTRAL, Scopus, Web of Science, ClinicalTrials.gov, Chinese Knowledge Resource Integrated Database and Chinese Clinical Trial Registry were searched on 20 March 2020.

RESULTS

In total, 135 studies were included, involving severe acute respiratory syndrome, Middle East respiratory syndrome, coronavirus disease 2019 and other coronaviruses. Most were case reports, case series and cohort studies of poor to fair quality. In post-mortem examinations of subjects who died from infection, around half had virus identified in heart tissues in severe acute respiratory syndrome, but none in Middle East respiratory syndrome and coronavirus disease 2019. Cardiac manifestations reported include tachycardia, bradycardia, arrhythmias, and myocardial injury, secondary to both systemic infection and treatment. Cardiac injury and arrhythmias are more prevalent in coronavirus disease 2019, and elevated cardiac markers are associated with intensive care unit admission and death. In severe acute respiratory syndrome, Middle East respiratory syndrome, and coronavirus disease 2019, comorbidities such as hypertension, diabetes mellitus, and heart disease are associated with intensive care unit admission, mechanical ventilation, and mortality. There were cases of misdiagnosis due to overlapping presentations of cardiovascular diseases and coronavirus infections, leading to hospital spread and delayed management of life-threatening conditions.

CONCLUSION

This review highlighted the ways in which coronaviruses affect cardiovascular function and interacts with pre-existing cardiovascular diseases.

Nation-Wide Observational Study of Cardiac Arrests Occurring in Nursing Homes and Nursing Facilities in Singapore

Introduction: Nursing home (NH) residents with out-of-hospital cardiac arrests (OHCA) have unique resuscitation priorities. This study aimed to describe OHCA characteristics in NH residents and identify independent predictors of survival. Materials and Methods: OHCA cases between 2010–16 in the Pan-Asian Resuscitation Outcomes Study were retrospectively analysed. Patients aged <18 years old and non-emergency cases were excluded. Primary outcome was survival at discharge or 30 days. Good neurological outcome was defined as a cerebral performance score between 1–2. Results: A total of 12,112 cases were included. Of these, 449 (3.7%) were NH residents who were older (median age 79 years, range 69–87 years) and more likely to have a history of stroke, heart and respiratory diseases. Fewer NH OHCA had presumed cardiac aetiology (62% vs 70%, P <0.01) and initial shockable rhythm (8.9% vs 18%, P <0.01), but had higher incidence of bystander cardiopulmonary resuscitation (74% vs 43%, P <0.01) and defibrillator use (8.5% vs 2.8%, P <0.01). Non-NH (2.8%) residents had better neurological outcomes than NH (0.9%) residents (P <0.05). Factors associated with survival for cardiac aetiology included age <65 years old, witnessed arrest, bystander defibrillator use and initial shockable rhythm; for non-cardiac aetiology, these included witnessed arrest (adjusted odds ratio [AOR] 3.8, P <0.001) and initial shockable rhythm (AOR 5.7, P <0.001). Conclusion: Neurological outcomes were poorer in NH survivors of OHCA. These findings should inform health policies on termination of resuscitation, advance care directives and do-not-resuscitate orders in this population. Ann Acad Med Singapore 2020;49:285–93 Key words: Advance care directives, Do-not-resuscitate orders, Geriatrics, Out-of- hospital, Palliative care

Impact of Air Pollution and Trans-Boundary Haze on Nation-Wide Emergency Department Visits and Hospital Admissions in Singapore

Introduction: Air pollution is associated with adverse health outcomes. However, its impact on emergency health services is less well understood. We investigated the impact of air pollution on nation-wide emergency department (ED) visits and hospital admissions to public hospitals in Singapore. Materials and Methods: Anonymised administrative and clinical data of all ED visits to public hospitals in Singapore from January 2010 to December 2015 were retrieved and analysed. Primary and secondary outcomes were defined as ED visits and hospital admissions, respectively. Conditional Poisson regression was used to model the effect of Pollutant Standards Index (PSI) on each outcome. Both outcomes were stratified according to subgroups defined a priori based on age, diagnosis, gender, patient acuity and time of day. Results: There were 5,791,945 ED visits, of which 1,552,187 resulted in hospital admissions. No significant association between PSI and total ED visits (Relative risk [RR], 1.002; 99.2% confidence interval [CI], 0.995–1.008; P = 0.509) or hospital admissions (RR, 1.005; 99.2% CI, 0.996–1.014; P = 0.112) was found. However, for every 30-unit increase in PSI, significant increases in ED visits (RR, 1.023; 99.2% CI, 1.011–1.036; P = 1.24 × 10˗6) and hospital admissions (RR, 1.027; 99.2% CI, 1.010–1.043; P = 2.02 × 10˗5) for respiratory conditions were found. Conclusion: Increased PSI was not associated with increase in total ED visits and hospital admissions, but was associated with increased ED visits and hospital admissions for respiratory conditions in Singapore. Key words: Epidemiology, Healthcare utilisation, PSI, Public health, Time series

Abstract TP211: Long-Term Trends in Ischemic Stroke Risk Factors and Outcomes in a Multiethnic Population

Introduction: Stroke medicine has evolved over time, with changing demographics and improving outcomes. We aimed to evaluate trends in incidence, risk factors, treatment and case fatality of acute ischemic stroke (AIS) in a multiethnic population.

Methods: Data was obtained from the Singapore Stroke Registry. 60,325 AIS cases that presented to Singapore’s public hospitals from 2005-16 were studied. Trends were analyzed with Poisson regression, logistic regression and Cox regression.

Results: The AIS age-standardized incidence rate has decreased by 11.6% (126.5 to 111.8 per 100,000 person-years from 2005-07 to 2014-16, p < 0.001). This was observed in all gender and ethnic groups except in those younger than 65, which saw a 2.5% increase (51.0 to 52.4 per 100,000 person-years, p = 0.017). In terms of risk factors, hypertension, hyperlipidemia, and atrial fibrillation have increased in prevalence, while history of stroke, ischemic heart disease, diabetes, and smoking have decreased (Figure A). Treatment-wise, thrombolysis utilization rates have increased significantly from 0.8% in 2005-2007 to 6.7% in 2014-2016 (p < 0.001). While atrial fibrillation (AF) has become more prevalent (15.8% to 25.2%, p < 0.001), anticoagulant prescription has not increased by the same magnitude (13.2% to 14.4%, p = 0.014). Case fatality has reduced by 25.6% (8.2% to 6.1%, p < 0.001).

Conclusion: In a multi-ethnic population, whilst we found that AIS age-standardized incidence rate has been declining, we demonstrate a temporal trend of changing risk factors. Further research is needed to investigate the causes and implications of these trends, so as to better strategize prevention efforts.

Chronic remote ischemic conditioning for cardiovascular protection

New treatments are needed to prevent adverse left ventricular remodelling following acute myocardial infarction (AMI), in order to prevent heart failure and improve clinical outcomes following AMI. Remote ischemic conditioning (RIC) using transient limb ischemia and reperfusion has been reported to reduce myocardial infarct (MI) size in AMI patients treated by primary percutaneous coronary intervention, and whether it can improve clinical outcomes is currently being investigated. Interestingly, repeated daily episode of limb RIC (termed 'chronic remote ischemic conditioning', or CRIC) has been shown in experimental and clinical studies to confer beneficial effects on post-AMI cardiac remodelling and chronic heart failure. In addition, the beneficial effects of CRIC extend to vascular function, peripheral arterial disease and stroke. In this review article, we focus on the therapeutic potential of CRIC as a strategy for cardiovascular protection and for improving clinical outcomes in patients with cardiovascular disease.

Remote ischemic conditioning in ST-segment elevation myocardial infarction - an update

Acute myocardial infarction (AMI) and the heart failure (HF) that often results are among the leading causes of death and disability in the world. As such, novel strategies are required to protect the heart against the detrimental effects of acute ischemia/reperfusion injury (IRI), in order to reduce myocardial infarct (MI) size and prevent the onset of HF. The endogenous cardioprotective strategy of remote ischemic conditioning (RIC), in which cycles of brief ischemia and reperfusion are applied to a tissue or organ away from the heart, has been reported in experimental studies to reduce MI size in animal models of acute IRI. In the clinical setting, RIC can be induced by simply inflating and deflating a cuff placed on the upper arm or thigh to induce brief cycles of ischemia and reperfusion, a strategy which has been shown to reduce MI size in ST-segment elevation myocardial infarction (STEMI) patients undergoing primary percutaneous coronary intervention (PPCI). The results of the ongoing CONDI2/ERIC-PPCI trial are eagerly awaited, and will provide definitive answers with regards to the cardioprotective effect and clinical outcome benefits of RIC in STEMI.

Dynamics of thermalization and decoherence of a nanoscale system

We study the decoherence and thermalization dynamics of a nanoscale system coupled nonperturbatively to a fully quantum-mechanical bath. The system is prepared out of equilibrium in a pure state of the complete system. We propose a random matrix model and show analytically that there are two robust temporal regimes in the approach of the system to equilibrium-an initial Gaussian decay followed by an exponential tail, consistent with numerical results on small interacting lattices [S. Genway, A. F. Ho, and D. K. K. Lee, Phys. Rev. Lett. 105, 260402 (2010)]. Furthermore, the system decays towards a Gibbs ensemble in accordance with the eigenstate thermalization hypothesis.

Dynamics of thermalization in small Hubbard-model systems

We study numerically the thermalization and temporal evolution of a two-site subsystem of a fermionic Hubbard model prepared far from equilibrium at a definite energy. Even for very small systems near quantum degeneracy, the subsystem can reach a steady state resembling equilibrium. This occurs for a nonperturbative coupling between the subsystem and the rest of the lattice where relaxation to equilibrium is Gaussian in time, in sharp contrast to perturbative results. We find similar results for random couplings, suggesting such behavior is generic for small systems.

Two-component Fermi gas on internal-state-dependent optical lattices

We study the phase diagram of a one-dimensional, two-component (i.e., pseudo-"spin"-(1/2)) ultracold atomic Fermi gas. The two atom species can have different hopping or mass. A very rich phase diagram for equal densities of the species is found, containing Mott insulators and superfluids. We also discuss coupling such 1D systems and the experimental signatures of the phases. In particular, we compute the spin-structure factor at small momentum, which should reveal a spin gap.

Deconfinement in a 2D optical lattice of coupled 1D boson systems

We show that a two-dimensional (2D) array of 1D interacting boson tubes has a deconfinement transition between a 1D Mott insulator and a 3D superfluid for commensurate fillings and a dimensional crossover for the incommensurate case. We determine the phase diagram and excitations of this system and discuss the consequences for Bose condensates loaded in 2D optical lattices.

Instabilities in binary mixtures of one-dimensional quantum degenerate gases

We show that one-dimensional binary mixtures of bosons or of a boson and a spin-polarized fermion are Luttinger liquids with the following instabilities: (i) For different particle densities, strong attraction between the mixture components leads to collapse, while strong repulsion leads to demixing, and (ii) For a low-density mixture of two gases of impenetrable bosons (or a spin-polarized fermion and an impenetrable boson) of equal densities, the system develops a gap and exhibits enhanced pairing fluctuations when there is attraction between the components. In the boson-fermion mixture, the pairing fluctuations occur at finite momentum. Our conclusions apply to mixtures both on the continuum and on optical lattices away from integer or fractional commensurability.