{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,24]],"date-time":"2025-11-24T19:48:43Z","timestamp":1764013723759,"version":"3.45.0"},"reference-count":18,"publisher":"Public Library of Science (PLoS)","issue":"11","license":[{"start":{"date-parts":[[2025,11,24]],"date-time":"2025-11-24T00:00:00Z","timestamp":1763942400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"CNPQ","award":["444610\/2024-3"],"award-info":[{"award-number":["444610\/2024-3"]}]}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"\n We evaluated Federated Learning (FL) strategies for predicting COVID-19 mortality using a multicenter sample of 17,022 patients from 21 diverse Brazilian hospitals. We tested horizontal FL architectures employing Logistic Regression (LR) and a Multi-Layer Perceptron (MLP) via parameter aggregation, alongside a novel Federated Random Forest (RF) using ensemble aggregation. Performance gain (\n \u0394<\/jats:italic>\n AUC, calculated as AUC\n \n \n \n \n \n \n \u2000<\/mml:mi>\n \n federated<\/mml:mtext>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n minus AUC\n \n \n \n \n \n \n \u2000<\/mml:mi>\n \n local<\/mml:mtext>\n <\/mml:mrow>\n <\/mml:msub>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n ) was quantified using bootstrap analysis to determine 95% confidence intervals. FL models demonstrated a beneficial collaborative effect. The average\n \u0394<\/jats:italic>\n AUC across the network was +0.0018 for LR, +0.0599 for MLP, and +0.0528 for RF. Crucially, the gain\u2019s magnitude and statistical significance showed a strong inverse correlation with local patient volume (N). Substantial and statistically significant gains concentrated in data-limited institutions (N < 500). For example, the smallest hospital (N=86) achieved a remarkable\n \u0394<\/jats:italic>\n AUC of 0.3682 (95% CI [0.0908, 0.6307]) with the RF model. However, interpreting these benefits requires caution because the 95% CIs for\n \u0394<\/jats:italic>\n AUC crossed zero for the majority of hospitals, suggesting the collaborative model\u2019s statistical advantage is not universally certain at every site. This trade-off was particularly evident with the MLP model which, despite achieving the highest average\n \u0394<\/jats:italic>\n AUC, was the most volatile algorithm, registering the maximum performance degradation in the network (\n \u0394<\/jats:italic>\n AUC = \u20130.0884, 95% CI [\u20130.1527, \u20130.0273]) due to its high sensitivity to local data distribution disparities (non-IID). This study validates FL as an equity-enabling mechanism that effectively enhances predictive capacity where local data scarcity is highest. Our findings underscore that maximizing the most statistically certain benefits of FL requires continuous monitoring and local validation for successful clinical deployment across diverse settings.\n <\/jats:p>","DOI":"10.1371\/journal.pcbi.1013695","type":"journal-article","created":{"date-parts":[[2025,11,24]],"date-time":"2025-11-24T18:35:26Z","timestamp":1764009326000},"page":"e1013695","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Federated learning for COVID-19 mortality prediction in a multicentric sample of 21 hospitals"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4951-9327","authenticated-orcid":true,"given":"Roberta Moreira","family":"Wichmann","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Murilo Afonso","family":"Robiati Bigoto","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Alexandre Dias Porto","family":"Chiavegatto Filho","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"340","published-online":{"date-parts":[[2025,11,24]]},"reference":[{"key":"pcbi.1013695.ref001","doi-asserted-by":"crossref","first-page":"119","DOI":"10.1038\/s41746-020-00323-1","article-title":"The future of digital health with federated learning","volume":"3","author":"N Rieke","year":"2020","journal-title":"NPJ Digit Med."},{"key":"pcbi.1013695.ref002","first-page":"429","article-title":"Federated optimization in heterogeneous networks","volume":"2","author":"T Li","year":"2020","journal-title":"Proc Mach Learn Syst."},{"issue":"4","key":"pcbi.1013695.ref003","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3501813","article-title":"Federated learning for healthcare: Systematic review and architecture proposal","volume":"13","author":"RS Antunes","year":"2022","journal-title":"ACM Trans Intell Syst Technol."},{"key":"pcbi.1013695.ref004","doi-asserted-by":"crossref","first-page":"106854","DOI":"10.1016\/j.cie.2020.106854","article-title":"A review of applications in federated learning","volume":"149","author":"L Li","year":"2020","journal-title":"Comput Ind Eng."},{"key":"pcbi.1013695.ref005","first-page":"887","article-title":"Federated tensor factorization for computational phenotyping","volume":"2017","author":"Y Kim","year":"2017","journal-title":"KDD."},{"key":"pcbi.1013695.ref006","unstructured":"Pfohl SR, Dai AM, Heller K. Federated and differentially private learning for electronic health records. arXiv preprint. 2019. https:\/\/doi.org\/10.48550\/arXiv.1911.05861"},{"key":"pcbi.1013695.ref007","doi-asserted-by":"crossref","first-page":"103291","DOI":"10.1016\/j.jbi.2019.103291","article-title":"Patient clustering improves efficiency of federated machine learning to predict mortality and hospital stay time using distributed electronic medical records","volume":"99","author":"L Huang","year":"2019","journal-title":"J Biomed Inform."},{"issue":"2","key":"pcbi.1013695.ref008","article-title":"Privacy-preserving patient similarity learning in a federated environment: Development and analysis","volume":"6","author":"J Lee","year":"2018","journal-title":"JMIR Med Inform."},{"key":"pcbi.1013695.ref009","doi-asserted-by":"crossref","unstructured":"Brisimi TS, Chen R, Mela T, Olshevsky A, Paschalidis IC, Shi W. Federated learning of predictive models from federated electronic health records. Int J Med Inform. 2018.","DOI":"10.1016\/j.ijmedinf.2018.01.007"},{"issue":"8","key":"pcbi.1013695.ref010","doi-asserted-by":"crossref","first-page":"2278","DOI":"10.1093\/bioinformatics\/btac065","article-title":"Federated random forests can improve local performance of predictive models for various healthcare applications","volume":"38","author":"A-C Hauschild","year":"2022","journal-title":"Bioinformatics."},{"key":"pcbi.1013695.ref011","unstructured":"Beutel DJ, Topal T, Mathur A, Qiu X, Fernandez-Marques J, Gao Y. Flower: A friendly federated learning research framework. 2020. https:\/\/arxiv.org\/abs\/2007.14390"},{"issue":"2","key":"pcbi.1013695.ref012","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3298981","article-title":"Federated machine learning","volume":"10","author":"Q Yang","year":"2019","journal-title":"ACM Trans Intell Syst Technol."},{"issue":"5","key":"pcbi.1013695.ref013","doi-asserted-by":"crossref","first-page":"1333","DOI":"10.1109\/TIFS.2017.2787987","article-title":"Privacy-preserving deep learning via additively homomorphic encryption","volume":"13","author":"LT Phong","year":"2018","journal-title":"IEEE TransInformForensic Secur."},{"key":"pcbi.1013695.ref014","unstructured":"McMahan B, Moore E, Ramage D, Hampson S, Arcas BA. Communication-efficient learning of deep networks from decentralized data. In: Artif Intell Stat; 2017. p. 1273\u201382."},{"key":"pcbi.1013695.ref015","doi-asserted-by":"crossref","unstructured":"Asad M, Moustafa A, Ito T. Federated learning versus classical machine learning: A convergence comparison. arXiv preprint. 2021. https:\/\/doi.org\/10.48550\/arXiv.2107.10976","DOI":"10.22541\/au.162074596.66890690\/v1"},{"key":"pcbi.1013695.ref016","article-title":"Gradient-less federated gradient boosting trees with learnable learning rates","author":"C Ma","year":"2023","journal-title":"arXiv preprint"},{"issue":"2","key":"pcbi.1013695.ref017","article-title":"A scalable federated learning solution for secondary care using low-cost microcomputing: privacy-preserving development and evaluation of a COVID-19 screening test in UK hospitals","volume":"6","author":"AAS Soltan","year":"2024","journal-title":"Lancet Digit Health."},{"key":"pcbi.1013695.ref018","unstructured":"scikit-learn developers. sklearn.linear_model.LogisticRegression \u2014 scikit-learn 0.21.2 documentation [Internet]; 2014. Available from: https:\/\/scikit-learn.org\/stable\/modules\/generated\/sklearn.linear_model.LogisticRegression.html"}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013695","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,24]],"date-time":"2025-11-24T18:35:30Z","timestamp":1764009330000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013695"}},"subtitle":[],"editor":[{"given":"Samuel V.","family":"Scarpino","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2025,11,24]]},"references-count":18,"journal-issue":{"issue":"11","published-online":{"date-parts":[[2025,11,24]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1013695","relation":{},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,24]]}}}