A Deep Learning-Based Prediction Model for Mortality in Severe COVID-19 Patients Requiring Intensive Care Unit Admission: A Multicenter Registry Analysis

BACKGROUND: Coronavirus disease 2019 (COVID-19) has been associated with substantial global mortality. Although artificial intelligence (AI)-based models have shown promise for mortality prediction, evidence focused specifically on critically ill patients remains limited. In addition, meaningful clinical implementation requires models that rely on a minimal set of readily available variables to enable rapid risk stratification. This study aimed to develop a deep learning-based model for predicting mortality in critically ill COVID-19 patients requiring intensive care unit (ICU) admission, with an emphasis on efficiency and clinical applicability. METHODS: This retrospective multicenter study included 900 critically ill patients with COVID-19 admitted to ICUs across 22 centers in Korea between January 1, 2020, and August 31, 2021. A deep learning-based tabular model was constructed, and systematic feature reduction strategies were applied to optimize predictive performance while enhancing practicality. RESULTS: < 0.05). The initial model incorporating 66 variables achieved an area under the curve (AUC) of 0.8500. Following stepwise feature selection, the model was reduced to seven key variables (age, C-reactive protein, lactate dehydrogenase, pro-B-type natriuretic peptide, CXR infiltration, CXR consolidation, and ROX index), resulting in a simplified model with an AUC of 0.8247. CONCLUSION: A deep learning-based AI model was developed to support clinical decision-making in critically ill patients with COVID-19, offering an efficient and clinically practical approach to mortality risk assessment.

Authors

Institutions

Publication Details

Journal
Journal of Korean Medical Science
Published
2026-11-16
DOI
https://doi.org/10.3346/jkms.2026.41.e275
Primary Topic
COVID-19 Clinical Research Studies
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Deep Learning-Based Prediction Model for Mortality in Severe COVID-19 Patients Requiring Intensive Care Unit Admission: A Multicenter Registry Analysis

Tae-Ok Kim, Kang-Moon Park, Onyu Park, Su Hwan Lee et al.
1 links
Journal of Korean Medical Science
COVID-19 Clinical Research Studies
article

A Deep Learning-Based Prediction Model for Mortality in Severe COVID-19 Patients Requiring Intensive Care Unit Admission: A Multicenter Registry Analysis

Tae-Ok Kim, Kang-Moon Park, Onyu Park, Su Hwan Lee, Bo‐Gun Kho, Hye Ju Yeo, Chul Park, Woo Hyun Cho, Sunghoon Park, Joon Hwang, Hong‐Joon Shin, Chang-Seok Yoon, TAEHWA KIM, On behalf of Korean Intensive Care Study Group
article en

Abstract

BACKGROUND: Coronavirus disease 2019 (COVID-19) has been associated with substantial global mortality. Although artificial intelligence (AI)-based models have shown promise for mortality prediction, evidence focused specifically on critically ill patients remains limited. In addition, meaningful clinical implementation requires models that rely on a minimal set of readily available variables to enable rapid risk stratification. This study aimed to develop a deep learning-based model for predicting mortality in critically ill COVID-19 patients requiring intensive care unit (ICU) admission, with an emphasis on efficiency and clinical applicability. METHODS: This retrospective multicenter study included 900 critically ill patients with COVID-19 admitted to ICUs across 22 centers in Korea between January 1, 2020, and August 31, 2021. A deep learning-based tabular model was constructed, and systematic feature reduction strategies were applied to optimize predictive performance while enhancing practicality. RESULTS: < 0.05). The initial model incorporating 66 variables achieved an area under the curve (AUC) of 0.8500. Following stepwise feature selection, the model was reduced to seven key variables (age, C-reactive protein, lactate dehydrogenase, pro-B-type natriuretic peptide, CXR infiltration, CXR consolidation, and ROX index), resulting in a simplified model with an AUC of 0.8247. CONCLUSION: A deep learning-based AI model was developed to support clinical decision-making in critically ill patients with COVID-19, offering an efficient and clinically practical approach to mortality risk assessment.

Journal of Korean Medical ScienceVol. 41(44)
Chonnam National University (KR), Korea National University of Transportation (KR), Yonsei University (KR), Severance Hospital (KR), Chonnam National University Hospital (KR), Ulsan University Hospital (KR), Pusan National University Yangsan Hospital (KR), Hallym University Sacred Heart Hospital (KR), Pusan National University (KR)
Chonnam National University, National Research Foundation of Korea
Good health and well-being
COVID-19 Clinical Research Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon