Characterising risk of in-hospital mortality following cardiac arrest using machine learning: A retrospective international registry study

BACKGROUND In most of Taiwan’s medical institutions, congestion is a serious problem for emergency departments. Due to a lack of beds, patients spend more time in emergency retention zones, which make it difficult to detect cardiac arrest (CA). OBJECTIVE We seek to develop a Drug Early Warning System Model (DEWSM), it included drug injections and vital signs as this research important features. We use it to predict cardiac arrest in emergency departments via drug classification and medical expert suggestion. METHODS We propose this new model for detecting cardiac arrest via drug classification and by using a sliding window; we apply learning-based algorithms to time-series data for a DEWSM. By treating drug features as a dynamic time-series factor for cardiopulmonary resuscitation (CPR) patients, we increase sensitivity, reduce false alarm rates and mortality, and increase the model’s accuracy. To evaluate the proposed model, we use the area under the receiver operating characteristic curve (AUROC). RESULTS Four important findings are as follows: (1) We identify the most important drug predictors: bits (intravenous therapy), and replenishers and regulators of water and electrolytes (fluid and electrolyte supplement). The best AUROC of bits is 85%, it means the medical expert suggest the drug features: bits, it will affect the vital signs, and then the evaluate this model correctly classified patients with CPR reach 85%; that of replenishers and regulators of water and electrolytes is 86%. These two features are the most influential of the drug features in the task. (2) We verify feature selection, in which accounting for drugs improve the accuracy: In Task 1, the best AUROC of vital signs is 77%, and that of all features is 86%. In Task 2, the best AUROC of all features is 85%, which demonstrates that thus accounting for the drugs significantly affects prediction. (3) We use a better model: For traditional machine learning, this study adds a new AI technology: the long short-term memory (LSTM) model with the best time-series accuracy, comparable to the traditional random forest (RF) model; the two AUROC measures are 85%. It can be seen that the use of new AI technology will achieve better results, currently comparable to the accuracy of traditional common RF, and the LSTM model can be adjusted in the future to obtain better results. (4) We determine whether the event can be predicted beforehand: The best classifier is still an RF model, in which the observational starting time is 4 hours before the CPR event. Although the accuracy is impaired, the predictive accuracy still reaches 70%. Therefore, we believe that CPR events can be predicted four hours before the event. CONCLUSIONS This paper uses a sliding window to account for dynamic time-series data consisting of the patient’s vital signs and drug injections. The National Early Warning Score (NEWS) only focuses on the score of vital signs, and does not include factors related to drug injections. In this study, the experimental results of adding the drug injections are better than only vital signs. In a comparison with NEWS, we improve predictive accuracy via feature selection, which includes drugs as features. In addition, we use traditional machine learning methods and deep learning (using LSTM method as the main processing time series data) as the basis for comparison of this research. The proposed DEWSM, which offers 4-hour predictions, is better than the NEWS in the literature. This also confirms that the doctor’s heuristic rules are consistent with the results found by machine learning algorithms.

Download Full-text

Characteristics and outcomes in acute patients receiving mechanical ventilation in a Coronary Care Unit (CCU)

European Heart Journal Acute Cardiovascular Care ◽

10.1093/ehjacc/zuab020.165 ◽

2021 ◽

Vol 10 (Supplement_1) ◽

Author(s):

M Rivadeneira Ruiz ◽

DF Arroyo Monino ◽

T Seoane Garcia ◽

MP Ruiz Garcia ◽

JC Garcia Rubira

Keyword(s):

Mechanical Ventilation ◽

Cardiac Arrest ◽

Hospital Mortality ◽

Coronary Care Unit ◽

Chronic Obstructive ◽

Obstructive Pulmonary Disease ◽

Funding Sources ◽

Copd Patients ◽

The Mean ◽

Coronary Care

Abstract Funding Acknowledgements Type of funding sources: None. Objectives Mechanical ventilation is the short-term technical support most widely used and cardiac arrest its main indication in a Coronary Care Unit (CCU). However, the knowledge about the specific moment and ventilator mode of onset to avoid the acute lung injury is still equivocal. Our objective is to determine the survival rate and the prognostic factors in patients supported by mechanical ventilation. Methods We conducted a retrospective cohort study of adult patients admitted to the CCU between January 2018 and November 2020 that received mechanical ventilation during the hospital stay. Results We collected 94 patients, 28% females with a median age of 68 ± 11,9. 43% were diabetics and almost one quarter of them had some degree of chronic obstructive pulmonary disease (COPD). Ischemic cardiopathy (33%) and heart failure (31%) were frequent pathologies as well as renal injury (29% patients a filtration rate below 45 mL/min/1,73m2). The reason for initiating mechanical ventilation was cardiac arrest in the half of the patients. Volume-controlled ventilation (73%) was the initial setting mode in most cases. The support with vasoactive drugs were highly necessary in these patients (Infection rate of 48%). In the subgroup analysis, we realized that the number of reintubations and the necessity of non-invasive ventilation were higher in the COPD group (p = 0,01), as well as tracheostomy (p = 0,03). COPD patients also needed higher maintaining PEEP, though this was not statistically significant. The mean length of stay in the intensive care unit of our cohort was 11 days (range: 1-78 days; median: 8 days) and the mean length of mechanical ventilation 6 days (range: 1-64 days; median: 3 days). The in-hospital mortality was 41,4%. Conclusions Cardiac arrest is the most common reason of mechanical ventilation support. Our study showed that COPD patients presented more complications during the weaning and the period after extubation. In-hospital mortality remains high in intubated patients.

Download Full-text

LEVERAGING MACHINE LEARNING MODELS TO PREDICT IN-HOSPITAL MORTALITY AFTER MITRACLIP

Journal of the American College of Cardiology ◽

10.1016/s0735-1097(21)02299-3 ◽

2021 ◽

Vol 77 (18) ◽

pp. 940

Author(s):

Chayakrit Krittanawong ◽

Hafeez Ul Hassan Virk ◽

Joshua Hahn ◽

Fu'ad Al-Azzam ◽

Kevin Greason ◽

...

Keyword(s):

Machine Learning ◽

Hospital Mortality ◽

Learning Models ◽

Machine Learning Models

Download Full-text

The in-hospital outcomes of the use of microaxial left ventricular assist device vs intra-aortic ballon pump in acute myocardial infarction complicated by cardiogenic shock

European Heart Journal ◽

10.1093/ehjci/ehaa946.1850 ◽

2020 ◽

Vol 41 (Supplement_2) ◽

Author(s):

C Jin ◽

Y Yang ◽

B Liu

Keyword(s):

Cardiac Arrest ◽

Blood Transfusion ◽

Length Of Stay ◽

Propensity Score ◽

Hospital Mortality ◽

Cardiogenic Shock ◽

Left Ventricular ◽

Bleeding Complications ◽

Categorical Variables ◽

Cardiac Complications

Abstract Purpose To compare the outcomes of patients with AMI underwent percutaneous coronary intervention (PCI) complicated by cardiogenic shock treated with IABP vs MLVAD. Methods The Nationwide Inpatient Sample (NIS) database is the largest inpatient registry in the U.S. We used NIS year 2009–2014 to identify adult patients admitted for AMI, who received PCI and complicated by cardiogenic shock. Based on the use of IABP or MLVAD, the study population was divided into 2 groups. To reduce selection bias, we performed propensity score matching using Kernell method. Patient characteristics, hospital characteristics, and comorbidities were matched. Logistic regression was used for categorical variables including in-hospital mortality, requirement of blood transfusion, sepsis, cardiac arrest and cardiac complications (including iatrogenic complications, hemopericardium, and cardiac tamponade). Poisson regression was used for continuous variables including length of stay and total cost. Results A total of 49837 patients were identified. With propensity score match, 34132 patients in IABP group were matched to 1430 patients in MLVAD group. Compared with MLVAD group, the IABP group had lower in-hospital mortality rates (28.29% vs 40.36%, OR 0.58 (0.42–0.81), p=0.002), lower rate of blood transfusion (9.63% vs 11.50%, OR 0.49 (0.27–0.88), p=0.017), and lower cost (47167 vs 70429 USD, p<0.001). IABP and MLVAD group had similar length of stay (8.9 versus 9.3 days, p=0.882), rates of cardiac complication (6.50% vs 7.24%, OR 0.56 (0.26–1.19), p=0.134), rates of sepsis (9.30% vs 14.98%, OR 0.66 (0.38–1.14), p=0.133), and rates of cardiac arrest (37.84% vs 41.05%, OR 0.70 (0.45–1.10), p=0.123). Conclusion In patients with AMI underwent PCI and complicated by cardiogenic shock, MLAVD compared with IABP was associated with higher risk of in-hospital mortality, requirement of blood transfusion indicating presence of major bleeding complications, and cost, although study interpretation is limited by retrospective observational design. Further research is warranted to elucidate the optimal MCSD in these patients. Funding Acknowledgement Type of funding source: None

Download Full-text

Prediction of cardiac arrest in critically ill patients presenting to the emergency department using a machine learning score incorporating heart rate variability compared with the modified early warning score

Critical Care ◽

10.1186/cc11396 ◽

2012 ◽

Vol 16 (3) ◽

pp. R108 ◽

Cited By ~ 44

Author(s):

Marcus Eng Hock Ong ◽

Christina Hui Lee Ng ◽

Ken Goh ◽

Nan Liu ◽

Zhi Koh ◽

...

Keyword(s):

Machine Learning ◽

Emergency Department ◽

Heart Rate ◽

Heart Rate Variability ◽

Cardiac Arrest ◽

Critically Ill ◽

Early Warning ◽

Critically Ill Patients ◽

Early Warning Score ◽

Learning Score

Download Full-text

In-hospital mortality after out-of-hospital cardiac arrest

The Lancet ◽

10.1016/s0140-6736(95)92784-0 ◽

1995 ◽

Vol 346 (8972) ◽

pp. 417-421 ◽

Cited By ~ 69

Author(s):

N.R Grubb ◽

K.A.A Fox ◽

R.A Elton

Keyword(s):

Cardiac Arrest ◽

Hospital Mortality ◽

Hospital Cardiac Arrest

Download Full-text

B-011-22 MACHINE LEARNING TO PREDICT RECURRENT EVENTS FOLLOWING UNEXPLAINED CARDIAC ARREST

Heart Rhythm ◽

10.1016/j.hrthm.2021.07.035 ◽

2021 ◽

Vol 18 (8) ◽

pp. S476-S477

Author(s):

Christopher C. Cheung ◽

Brianna Davies ◽

Jason D. Roberts ◽

Rafik Tadros ◽

Martin S. Green ◽

...

Keyword(s):

Machine Learning ◽

Cardiac Arrest ◽

Recurrent Events

Download Full-text

Pediatric In-Hospital Cardiac Arrest International Registry (PACHIN): protocol for a prospective international multicenter register of cardiac arrest in children

BMC Cardiovascular Disorders ◽

10.1186/s12872-021-02173-5 ◽

2021 ◽

Vol 21 (1) ◽

Author(s):

Jimena del Castillo ◽

Débora Sanz ◽

Laura Herrera ◽

Jesús López-Herce ◽

Cristina Calvo ◽

...

Keyword(s):

Cardiac Arrest ◽

Public Health Problem ◽

Neurological Status ◽

Social Contexts ◽

Secondary Outcome ◽

Major Public Health Problem ◽

International Registry ◽

Neurological Prognosis ◽

Model Database ◽

Hospital Pediatric

Abstract Background and aims Cardiac arrest (CA) in children is a major public health problem. Thanks to advances in cardiopulmonary resuscitation (CPR) guidelines and teaching skills, results in children have improved. However, pediatric CA has a very high mortality. In the treatment of in-hospital CA there are still multiple controversies. The objective of this study is to develop a multicenter and international registry of in-hospital pediatric cardiac arrest including the diversity of management in different clinical and social contexts. Participation in this register will enable the evaluation of the diagnosis of CA, CPR and post-resuscitation care and its influence in survival and neurological prognosis. Methods An intrahospital CA data recording protocol has been designed following the Utstein model. Database is hosted according to European legislation regarding patient data protection. It is drafted in English and Spanish. Invitation to participate has been sent to Spanish, European and Latinamerican hospitals. Variables included, asses hospital characteristics, the resuscitation team, patient’s demographics and background, CPR, post-resuscitation care, mortality, survival and long-term evolution. Survival at hospital discharge will be evaluated as a primary outcome and survival with good neurological status as a secondary outcome, analyzing the different factors involved in them. The study design is prospective, observational registry of a cohort of pediatric CA. Conclusions This study represents the development of a registry of in-hospital CA in childhood. Its development will provide access to CPR data in different hospital settings and will allow the analysis of current controversies in the treatment of pediatric CA and post-resuscitation care. The results may contribute to the development of further international recommendations. Trial register: ClinicalTrials.gov Identifier: NCT04675918. Registered 19 December 2020 – Retrospectively registered, https://clinicaltrials.gov/ct2/show/record/NCT04675918?cond=pediatric+cardiac+arrest&draw=2&rank=10

Download Full-text

Abstract 052: Improved Hospital Outcomes With Hypokalemia on Admission in Cardiac Arrest Survivors Treated With Therapeutic Hypothermia

Circulation Cardiovascular Quality and Outcomes ◽

10.1161/circoutcomes.10.suppl_3.052 ◽

2017 ◽

Vol 10 (suppl_3) ◽

Author(s):

Angelo de la Rosa ◽

Manuel Tapia ◽

Yong Ji ◽

Basil Saour ◽

Mikhail Torosoff

Keyword(s):

Logistic Regression ◽

Cardiac Arrest ◽

Ventricular Fibrillation ◽

Hospital Mortality ◽

Electrical Activity ◽

Therapeutic Hypothermia ◽

Multivariate Logistic Regression Analysis ◽

Pulseless Electrical Activity ◽

Older Age ◽

Chi Square

Purpose: We hypothesized that advanced circulatory compromise, as manifested by acidosis and hyperkalemia should be associated with worsened clinical outcomes in cardiac arrest patients treated with therapeutic hypothermia. Methods: Results of initial admission laboratory studies, medical history, and echocardiogram in 203 consecutive cardiac arrest patients (59 females, 59+/- 15 years old) undergoing therapeutic hypothermia were reviewed. Mortality was ascertained through hospital records. ANOVA, chi-square, Kaplan-Meier, and logistic regression analyses were used. The study was approved by the institutional IRB. Results: Increased mortality was noted with older age, decreased admission pH, elevated admission lactate, lower admission hemoglobin, and pulseless electrical activity or asystole as presenting rhythms (Table). Admission hypokalemia and ventricular fibrillation/tachycardia were associated with improved hospital mortality (Table). Potassium was significantly lower in patients admitted with ventricular fibrillation/tachycardia (3.897+/-0.92) as compared to patients with asystole (4.674+/-1.377) or pulseless electrical activity (4.491+/-1.055 mEq/dL, p<0.0001). In multivariate logistic regression analysis, independent predictors of increased hospital mortality included increased admission potassium (OR 2.0, 95%CI 1.291-3.170, p=0.002)), older age (OR 1.04, 95%CI 1.007-1.071, p=0.017), admission PEA (OR 3.7, 95%CI 1.358-10.282, p=0.011 when compared to ventricular fibrillation/tachycardia) or asystole (OR 17.2, 95%CI 4.423-66.810, p<0.001 when compared to ventricular fibrillation/tachycardia); while decreased mortality was associated with higher hemoglobin (OR 0.8, 95%CI 0.665-0.997, p=0.047). Conclusions: Hyperkalemia, pulseless electrical activity, and asystole are predictive of increased hospital mortality in survivors of cardiac arrest. An association between low or low-normal potassium, observed VT-VF, and better outcomes is unexpected and may be used for prognostic purposes. More prospective investigations of mortality predictors in these critically ill patients are needed.

Download Full-text