tetano
Editor, Senior Moderator
Clin Transl Sci
. 2025 Nov;18(11):e70403.
doi: 10.1111/cts.70403. Integrative Machine Learning and Bayesian Analysis Reveals Atrial Fibrillation as a Key Predictor of Severe COVID-19 Outcomes
Young Sook Ku[SUP] 1 2 [/SUP], Go Woon Lee[SUP] 1 [/SUP], Hyundeok Seo[SUP] 3 [/SUP], Jin Yeon Gil[SUP] 1 [/SUP], Kyung Hyun Min[SUP] 1 [/SUP], Jun Hyeob Kim[SUP] 1 [/SUP], Jun Hyuk Park[SUP] 1 [/SUP], Woorim Kim[SUP] 3 [/SUP], Kyung Eun Lee[SUP] 1 [/SUP]
Affiliations
This study aimed to identify predictors of critical outcomes, including mortality, in hospitalized COVID-19 patients treated with remdesivir, using statistical, machine learning, and Bayesian methods. A retrospective multicenter cohort of 1628 patients hospitalized between January 2021 and August 2022 was analyzed. Clinical data were collected from electronic medical records. Multivariable logistic regression, machine learning models (LightGBM, Elastic Net) with SHapley Additive exPlanations (SHAP), and Bayesian logistic regression were applied. Among the cohort, 14.5% experienced critical outcomes or death. Advanced age (≥ 65 years; aOR 3.950), atrial fibrillation (aOR 4.087), and kidney disease (aOR 1.939) were identified as significant predictors. Machine learning models achieved moderate predictive performance (AUROC: LightGBM 0.705, Elastic Net 0.698), with SHAP highlighting atrial fibrillation and age as key contributors. Bayesian analysis confirmed a strong association between atrial fibrillation and adverse outcomes (adjusted OR 5.121). Atrial fibrillation emerged as a consistent and strong predictor, underscoring its relevance in clinical risk assessment.
Keywords: Bayesian analysis; COVID‐19; atrial fibrillation; machine learning; risk factor.
. 2025 Nov;18(11):e70403.
doi: 10.1111/cts.70403. Integrative Machine Learning and Bayesian Analysis Reveals Atrial Fibrillation as a Key Predictor of Severe COVID-19 Outcomes
Young Sook Ku[SUP] 1 2 [/SUP], Go Woon Lee[SUP] 1 [/SUP], Hyundeok Seo[SUP] 3 [/SUP], Jin Yeon Gil[SUP] 1 [/SUP], Kyung Hyun Min[SUP] 1 [/SUP], Jun Hyeob Kim[SUP] 1 [/SUP], Jun Hyuk Park[SUP] 1 [/SUP], Woorim Kim[SUP] 3 [/SUP], Kyung Eun Lee[SUP] 1 [/SUP]
Affiliations
- PMID: 41243757
- DOI: 10.1111/cts.70403
This study aimed to identify predictors of critical outcomes, including mortality, in hospitalized COVID-19 patients treated with remdesivir, using statistical, machine learning, and Bayesian methods. A retrospective multicenter cohort of 1628 patients hospitalized between January 2021 and August 2022 was analyzed. Clinical data were collected from electronic medical records. Multivariable logistic regression, machine learning models (LightGBM, Elastic Net) with SHapley Additive exPlanations (SHAP), and Bayesian logistic regression were applied. Among the cohort, 14.5% experienced critical outcomes or death. Advanced age (≥ 65 years; aOR 3.950), atrial fibrillation (aOR 4.087), and kidney disease (aOR 1.939) were identified as significant predictors. Machine learning models achieved moderate predictive performance (AUROC: LightGBM 0.705, Elastic Net 0.698), with SHAP highlighting atrial fibrillation and age as key contributors. Bayesian analysis confirmed a strong association between atrial fibrillation and adverse outcomes (adjusted OR 5.121). Atrial fibrillation emerged as a consistent and strong predictor, underscoring its relevance in clinical risk assessment.
Keywords: Bayesian analysis; COVID‐19; atrial fibrillation; machine learning; risk factor.