scholarly journals Prognostic Assessment of COVID-19 in the Intensive Care Unit by Machine Learning Methods: Model Development and Validation

10.2196/23128 ◽  
2020 ◽  
Vol 22 (11) ◽  
pp. e23128
Author(s):  
Pan Pan ◽  
Yichao Li ◽  
Yongjiu Xiao ◽  
Bingchao Han ◽  
Longxiang Su ◽  
...  
Keyword(s):  
Machine Learning ◽  
Risk Factors ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Prediction Model ◽  
Risk Prediction ◽  
Risk Prediction Model ◽  
Data Set ◽  
Icu Patients ◽  
Potential Risk Factors

Background Patients with COVID-19 in the intensive care unit (ICU) have a high mortality rate, and methods to assess patients’ prognosis early and administer precise treatment are of great significance. Objective The aim of this study was to use machine learning to construct a model for the analysis of risk factors and prediction of mortality among ICU patients with COVID-19. Methods In this study, 123 patients with COVID-19 in the ICU of Vulcan Hill Hospital were retrospectively selected from the database, and the data were randomly divided into a training data set (n=98) and test data set (n=25) with a 4:1 ratio. Significance tests, correlation analysis, and factor analysis were used to screen 100 potential risk factors individually. Conventional logistic regression methods and four machine learning algorithms were used to construct the risk prediction model for the prognosis of patients with COVID-19 in the ICU. The performance of these machine learning models was measured by the area under the receiver operating characteristic curve (AUC). Interpretation and evaluation of the risk prediction model were performed using calibration curves, SHapley Additive exPlanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), etc, to ensure its stability and reliability. The outcome was based on the ICU deaths recorded from the database. Results Layer-by-layer screening of 100 potential risk factors finally revealed 8 important risk factors that were included in the risk prediction model: lymphocyte percentage, prothrombin time, lactate dehydrogenase, total bilirubin, eosinophil percentage, creatinine, neutrophil percentage, and albumin level. Finally, an eXtreme Gradient Boosting (XGBoost) model established with the 8 important risk factors showed the best recognition ability in the training set of 5-fold cross validation (AUC=0.86) and the verification queue (AUC=0.92). The calibration curve showed that the risk predicted by the model was in good agreement with the actual risk. In addition, using the SHAP and LIME algorithms, feature interpretation and sample prediction interpretation algorithms of the XGBoost black box model were implemented. Additionally, the model was translated into a web-based risk calculator that is freely available for public usage. Conclusions The 8-factor XGBoost model predicts risk of death in ICU patients with COVID-19 well; it initially demonstrates stability and can be used effectively to predict COVID-19 prognosis in ICU patients.

Prognostic Assessment of COVID-19 in the Intensive Care Unit by Machine Learning Methods: Model Development and Validation (Preprint)

2020 ◽  
Author(s):  
Pan Pan ◽  
Yichao Li ◽  
Yongjiu Xiao ◽  
Bingchao Han ◽  
Longxiang Su ◽  
...  
Keyword(s):  
Machine Learning ◽  
Risk Factors ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Prediction Model ◽  
Risk Prediction ◽  
Risk Prediction Model ◽  
Data Set ◽  
Icu Patients ◽  
Potential Risk Factors

BACKGROUND Patients with COVID-19 in the intensive care unit (ICU) have a high mortality rate, and methods to assess patients’ prognosis early and administer precise treatment are of great significance. OBJECTIVE The aim of this study was to use machine learning to construct a model for the analysis of risk factors and prediction of mortality among ICU patients with COVID-19. METHODS In this study, 123 patients with COVID-19 in the ICU of Vulcan Hill Hospital were retrospectively selected from the database, and the data were randomly divided into a training data set (n=98) and test data set (n=25) with a 4:1 ratio. Significance tests, correlation analysis, and factor analysis were used to screen 100 potential risk factors individually. Conventional logistic regression methods and four machine learning algorithms were used to construct the risk prediction model for the prognosis of patients with COVID-19 in the ICU. The performance of these machine learning models was measured by the area under the receiver operating characteristic curve (AUC). Interpretation and evaluation of the risk prediction model were performed using calibration curves, SHapley Additive exPlanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), etc, to ensure its stability and reliability. The outcome was based on the ICU deaths recorded from the database. RESULTS Layer-by-layer screening of 100 potential risk factors finally revealed 8 important risk factors that were included in the risk prediction model: lymphocyte percentage, prothrombin time, lactate dehydrogenase, total bilirubin, eosinophil percentage, creatinine, neutrophil percentage, and albumin level. Finally, an eXtreme Gradient Boosting (XGBoost) model established with the 8 important risk factors showed the best recognition ability in the training set of 5-fold cross validation (AUC=0.86) and the verification queue (AUC=0.92). The calibration curve showed that the risk predicted by the model was in good agreement with the actual risk. In addition, using the SHAP and LIME algorithms, feature interpretation and sample prediction interpretation algorithms of the XGBoost black box model were implemented. Additionally, the model was translated into a web-based risk calculator that is freely available for public usage. CONCLUSIONS The 8-factor XGBoost model predicts risk of death in ICU patients with COVID-19 well; it initially demonstrates stability and can be used effectively to predict COVID-19 prognosis in ICU patients.

scholarly journals Pressure Ulcers Prevalence and Potential Risk Factors Among Intensive Care Unit Patients in Governmental Hospitals in Palestine: A Cross-sectional Study

2019 ◽  
Vol 12 (1) ◽  
pp. 121-126 ◽  
Author(s):  
Jamal A.S. Qaddumi ◽  
Omar Almahmoud
Keyword(s):  
Risk Factors ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Potential Risk ◽  
Pressure Ulcers ◽  
Advisory Panel ◽  
Cross Sectional ◽  
Data Set ◽  
Potential Risk Factors ◽  
The Government

Aim: To determine the prevalence rate and the potential risk factors of pressure ulcers (PUs) among patients in the intensive care unit (ICU) departments of the government hospitals in Palestine. Methods: A quantitative, cross-sectional, descriptive analytical study was carried out in five government hospital intensive care units in four different Palestinian cities between September 27, 2017, and October 27, 2017. The data of 109 out of 115 (94.78%) inpatients were analyzed. The Minimum Data Set (MDS) recommended by the European Pressure Ulcer Advisory Panel (EPUAP) was used to collect inpatients’ information. Results: The result of the analysis showed that the prevalence of pressure ulcers in the ICU departments was 33%, and the prevalence of PUs when excluding stage one was 7.3%. The common stage for pressure ulcers was stage one. It was also determined that the most common risk factors for the development of pressure ulcers were the number of days in the hospital, moisture, and friction. Conclusion: According to the recent studies in the Asian States, the prevalence of pressure ulcers in Palestine is considerably higher than in China and Jordan. However, it is still lower than the prevalence reported in comparable published studies in Western Europe. Increasing the staff’s knowledge about PUs screening and preventive measures is highly recommended in order to decrease the burden of PUs.

Abstract 15875: Prediction of Intensive Care Unit Admission Among Patients Hospitalized for COVID-19: The Intermountain Coronavirus ICU Risk Model (CORONA-ICU)

Circulation ◽  
2020 ◽  
Vol 142 (Suppl_3) ◽  
Author(s):  
Heidi T May ◽  
Joseph B Muhlestein ◽  
Benjamin D Horne ◽  
Kirk U Knowlton ◽  
Tami L Bair ◽  
...  
Keyword(s):  
Risk Factors ◽  
Intensive Care Unit ◽  
Intensive Care ◽  
Risk Prediction ◽  
Prediction Models ◽  
Risk Model ◽  
Beta Blockers ◽  
Risk Scores ◽  
Channel Blockers ◽  
Icu Patients

Background: Treatment for COVID-19 has created surges in hospitalizations, intensive care unit (ICU) admissions, and the need for advanced medical therapy and equipment, including ventilators. Identifying patients early on who are at risk for more intensive hospital resource use and poor outcomes could result in shorter hospital stays, lower costs, and improved outcomes. Therefore, we created clinical risk scores (CORONA-ICU and -ICU+) to predict ICU admission among patients hospitalized for COVID-19. Methods: Intermountain Healthcare patients who tested positive for SARS-CoV-2 and were hospitalized between March 4, 2020 and June 8, 2020 were studied. Derivation of CORONA-ICU risk score models used weightings of commonly collected risk factors and medicines. The primary outcome was admission to the ICU during hospitalization, and secondary outcomes included death and ventilator use. Results: A total of 451 patients were hospitalized for a SARS-CoV-2 positive infection, and 191 (42.4%) required admission to the ICU. Patients admitted to the ICU were older (58.2 vs. 53.6 years), more often male (61.3% vs. 48.5%), and had higher rates of hyperlipidemia, hypertension, diabetes, and peripheral arterial disease. ICU patients more often took ACE inhibitors, beta-blockers, calcium channel blockers, diuretics, and statins. Table 1 shows variables that were evaluated and included in the CORONA-ICU risk prediction models. Models adding medications (CORONA-ICU+) improved risk-prediction. Though not created to predict death and ventilator use, these models did so with high accuracy (Table 2). Conclusion: The CORONA-ICU and -ICU+ models, composed of commonly collected risk factors without or with medications, were shown to be highly predictive of ICU admissions, death, and ventilator use. These models can be efficiently derived and effectively identify high-risk patients who require more careful observation and increased use of advanced medical therapies.

PROPER: Development of an Early Pediatric Intensive Care Unit Readmission Risk Prediction Tool

2016 ◽  
Vol 33 (1) ◽  
pp. 29-36 ◽  
Author(s):  
Harsheen Kaur ◽  
James M. Naessens ◽  
Andrew C. Hanson ◽  
Karen Fryer ◽  
Michael E. Nemergut ◽  
...  
Keyword(s):  
Intensive Care Unit ◽  
Intensive Care ◽  
Prediction Model ◽  
Pediatric Intensive Care Unit ◽  
Risk Prediction ◽  
Pediatric Population ◽  
Cerebral Performance Category ◽  
Pediatric Intensive Care ◽  
Risk Prediction Model ◽  
Retrospective Case

Objective: No risk prediction model is currently available to measure patient’s probability for readmission to the pediatric intensive care unit (PICU). This retrospective case–control study was designed to assess the applicability of an adult risk prediction score (Stability and Workload Index for Transfer [SWIFT]) and to create a pediatric version (PRediction Of PICU Early Readmissions [PROPER]). Design: Eighty-six unplanned early (<48 hours) PICU readmissions from January 07, 2007, to June 30, 2014, were compared with 170 random controls. Patient- and disease-specific data and PICU workload factors were compared across the 2 groups. Factors statistically significant on multivariate analysis were included in the creation of the risk prediction model. The SWIFT scores were calculated for cases and controls and compared for validation. Results: Readmitted patients were younger, weighed less, and were more likely to be admitted from the emergency department. There were no differences in gender, race, or admission Pediatric Index of Mortality scores. A higher proportion of patients in the readmission group had a Pediatric Cerebral Performance Category in the moderate to severe disability category. Cases and controls did not differ with respect to staff workload at discharge or discharge day of the week; there was a much higher proportion of patients on supplemental oxygen in the readmission group. Only 2 of 5 categories in the SWIFT model were significantly different, and although the median SWIFT score was significantly higher in the readmissions group, the model discriminated poorly between cases and controls (area under the curve: 0.613). A 7-category PROPER score was created based on a multiple logistic regression model. Sensitivity of this model (score ≥12) for the detection of readmission was 81% with a positive predictive value of 0.50. Conclusion: We have created a preliminary model for predicting patients at risk of early readmissions to the PICU from the hospital floor. The SWIFT score is not applicable for predicting the risk for pediatric population.

scholarly journals Prediction Model of Anastomotic Leakage Among Esophageal Cancer Patients After Receiving an Esophagectomy: Machine Learning Approach

10.2196/27110 ◽  
2021 ◽  
Vol 9 (7) ◽  
pp. e27110
Author(s):  
Ziran Zhao ◽  
Xi Cheng ◽  
Xiao Sun ◽  
Shanrui Ma ◽  
Hao Feng ◽  
...  
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Prediction Model ◽  
Anastomotic Leakage ◽  
Cancer Patients ◽  
Risk Prediction ◽  
Risk Prediction Model ◽  
Training Data ◽  
Data Set ◽  
Selection Operator

Background Anastomotic leakage (AL) is one of the severe postoperative adverse events (5%-30%), and it is related to increased medical costs in cancer patients who undergo esophagectomies. Machine learning (ML) methods show good performance at predicting risk for AL. However, AL risk prediction based on ML models among the Chinese population is unavailable. Objective This study uses ML techniques to develop and validate a risk prediction model to screen patients with emerging AL risk factors. Methods Analyses were performed using medical records from 710 patients who underwent esophagectomies at the National Clinical Research Center for Cancer between January 2010 and May 2015. We randomly split (9:1) the data set into a training data set of 639 patients and a testing data set of 71 patients using a computer algorithm. We assessed multiple classification tools to create a multivariate risk prediction model. Our ML algorithms contained decision tree, random forest, naive Bayes, and logistic regression with least absolute shrinkage and selection operator. The optimal AL prediction model was selected based on model evaluation metrics. Results The final risk panel included 36 independent risk features. Of those, 10 features were significantly identified by the logistic model, including aortic calcification (OR 2.77, 95% CI 1.32-5.81), celiac trunk calcification (OR 2.79, 95% CI 1.20-6.48), forced expiratory volume 1% (OR 0.51, 95% CI 0.30-0.89); TLco (OR 0.56, 95% CI 0.27-1.18), peripheral vascular disease (OR 4.97, 95% CI 1.44-17.07), laparoscope (OR 3.92, 95% CI 1.23-12.51), postoperative length of hospital stay (OR 1.17, 95% CI 1.13-1.21), vascular permeability activity (OR 0.46, 95% CI 0.14-1.48), and fat liquefaction of incisions (OR 4.36, 95% CI 1.86-10.21). Logistic regression with least absolute shrinkage and selection operator offered the highest prediction quality with an area under the receiver operator characteristic of 72% in the training data set. The testing model also achieved similar high performance. Conclusions Our model offered a prediction of AL with high accuracy, assisting in AL prevention and treatment. A personalized ML prediction model with a purely data-driven selection of features is feasible and effective in predicting AL in patients who underwent esophagectomy.

Sign in / Sign up

Export Citation Format

Share Document