Machine learning approach to dynamic risk modeling of mortality in COVID-19: a UK Biobank study DOI Creative Commons
Mohammad A. Dabbah, Angus B. Reed, Adam Booth

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2021, Volume and Issue: unknown

Published: Feb. 10, 2021

Abstract The COVID-19 pandemic has created an urgent need for robust, scalable monitoring tools supporting stratification of high-risk patients. This research aims to develop and validate prediction models, using the UK Biobank, estimate mortality risk in confirmed cases. From 11,245 participants testing positive COVID-19, we a data-driven random forest classification model with excellent performance (AUC: 0.91), baseline characteristics, pre-existing conditions, symptoms, vital signs, such that score could dynamically assess disease deterioration. We also identify several significant novel predictors equivalent or greater predictive value than established comorbidities, as detailed anthropometrics prior acute kidney failure, urinary tract infection, pneumonias. design feature selection enables utility outpatient settings. Possible applications include individual-level profiling progression across patients at-scale, especially hospital-at-home

Language: Английский

Machine learning approach to dynamic risk modeling of mortality in COVID-19: a UK Biobank study DOI Creative Commons
Mohammad A. Dabbah, Angus B. Reed, Adam Booth

et al.

Scientific Reports, Journal Year: 2021, Volume and Issue: 11(1)

Published: Aug. 19, 2021

The COVID-19 pandemic has created an urgent need for robust, scalable monitoring tools supporting stratification of high-risk patients. This research aims to develop and validate prediction models, using the UK Biobank, estimate mortality risk in confirmed cases. From 11,245 participants testing positive COVID-19, we a data-driven random forest classification model with excellent performance (AUC: 0.91), baseline characteristics, pre-existing conditions, symptoms, vital signs, such that score could dynamically assess disease deterioration. We also identify several significant novel predictors equivalent or greater predictive value than established comorbidities, as detailed anthropometrics prior acute kidney failure, urinary tract infection, pneumonias. design feature selection enables utility outpatient settings. Possible applications include individual-level profiling progression across patients at-scale, especially hospital-at-home

Language: Английский

Citations

18
Using decision tree algorithms for estimating ICU admission of COVID-19 patients DOI Creative Commons
Mostafa Shanbehzadeh, Raoof Nopour, Hadi Kazemi-Arpanahi

et al.

Informatics in Medicine Unlocked, Journal Year: 2022, Volume and Issue: 30, P. 100919 - 100919

Published: Jan. 1, 2022

Coronavirus disease 2019 (COVID-19) outbreak has overwhelmed many healthcare systems worldwide and put them at the edge of collapsing. As intensive care unit (ICU) capacities are limited, deciding on proper allocation required resources is crucial. This study aimed to develop compare models for early predicting ICU admission in COVID-19 patients point hospital admission. Using a single-center registry, we studied records 512 patients. First, most important variables were identified using Chi-square test (at p < 0.01) logistic regression (with odds ratio P 0.05). Second, trained seven decision tree (DT) algorithms (decision stump (DS), Hoeffding (HT), LMT, J-48, random forest (RF), (RT) REP-Tree) selected variables. Finally, models' performance was evaluated. Furthermore, used an external dataset validate prediction models. test, 20 identified. Then, 12 model construction regression. Comparing DT methods demonstrated that J-48 (F-score 0.816 AUC 0.845) had best performance. Also, = 80.9% 0.822) gained generalizability dataset. The results can be predict requirements based first time data. Implementing such potential inform clinicians managers adopt policy get prepare during time-sensitive resource-constrained situation.

Language: Английский

Citations

13
Application of Machine Learning in Hospitalized Patients with Severe COVID-19 Treated with Tocilizumab DOI Open Access

Antonio Ramón,

Marta Zaragozá,

Ana María Torres

et al.

Journal of Clinical Medicine, Journal Year: 2022, Volume and Issue: 11(16), P. 4729 - 4729

Published: Aug. 12, 2022

Among the IL-6 inhibitors, tocilizumab is most widely used therapeutic option in patients with SARS-CoV-2-associated severe respiratory failure (SRF). The aim of our study was to provide evidence on predictors poor outcome COVID-19 treated tocilizumab, using machine learning (ML) techniques. We conducted a retrospective study, analyzing clinical, laboratory and sociodemographic data admitted for SRF, tocilizumab. extreme gradient boost (XGB) method had highest balanced accuracy (93.16%). factors associated worse use terms mortality were: baseline situation at start treatment requiring invasive mechanical ventilation (IMV), elevated ferritin, lactate dehydrogenase (LDH) glutamate-pyruvate transaminase (GPT), lymphopenia, low PaFi [ratio between arterial oxygen pressure inspired fraction (PaO2/FiO2)] values. hospital stay IMV or supplemental oxygen, levels glutamate-oxaloacetate (GOT), GPT, C-reactive protein (CRP), LDH, In focused that were weighted strongly predicting clinical status hyperferritinemia.

Language: Английский

Citations

6
Clinical Outcomes and Severity of Acute Respiratory Distress Syndrome in 1154 COVID-19 Patients: An Experience Multicenter Retrospective Cohort Study DOI Creative Commons
Abbas Al Mutair, Saad Alhumaid, Laila Layqah

et al.

COVID, Journal Year: 2022, Volume and Issue: 2(8), P. 1102 - 1115

Published: Aug. 1, 2022

Background: Acute Respiratory Distress Syndrome (ARDS) is caused by non-cardiogenic pulmonary edema and occurs in critically ill patients. It one of the fatal complications observed among severe COVID-19 cases managed intensive care units (ICU). Supportive lung-protective ventilation prone positioning remain mainstay interventions. Purpose: We describe severity ARDS, clinical outcomes, management ICU patients with laboratory-confirmed infection multiple Saudi hospitals. Methods: A multicenter retrospective cohort study was conducted who were admitted to developed ARDS. Results: During our study, 1154 experienced ARDS: 591 (51.2%) severe, 415 (36.0%) moderate, 148 (12.8%) mild The mean sequential organ failure assessment (SOFA) score significantly higher ARDS (6 ± 5, p = 0.006). Kaplan–Meier survival analysis showed had a rate compared (p 0.023). Conclusion: challenging condition complicating infection. carries significant morbidity results elevated mortality. requires protective mechanical other critical supportive measures. associated death

Language: Английский

Citations

5
Machine learning approach to dynamic risk modeling of mortality in COVID-19: a UK Biobank study DOI Creative Commons
Mohammad A. Dabbah, Angus B. Reed, Adam Booth

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2021, Volume and Issue: unknown

Published: Feb. 10, 2021

Abstract The COVID-19 pandemic has created an urgent need for robust, scalable monitoring tools supporting stratification of high-risk patients. This research aims to develop and validate prediction models, using the UK Biobank, estimate mortality risk in confirmed cases. From 11,245 participants testing positive COVID-19, we a data-driven random forest classification model with excellent performance (AUC: 0.91), baseline characteristics, pre-existing conditions, symptoms, vital signs, such that score could dynamically assess disease deterioration. We also identify several significant novel predictors equivalent or greater predictive value than established comorbidities, as detailed anthropometrics prior acute kidney failure, urinary tract infection, pneumonias. design feature selection enables utility outpatient settings. Possible applications include individual-level profiling progression across patients at-scale, especially hospital-at-home

Language: Английский

Citations

3