{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,7]],"date-time":"2026-03-07T14:52:59Z","timestamp":1772895179791,"version":"3.50.1"},"reference-count":18,"publisher":"Public Library of Science (PLoS)","issue":"3","license":[{"start":{"date-parts":[[2026,3,6]],"date-time":"2026-03-06T00:00:00Z","timestamp":1772755200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100003593","name":"Conselho Nacional de Desenvolvimento Cient\u00c3fico e Tecnol\u00c3\u00b3gico","doi-asserted-by":"publisher","award":["444610\/2024-3"],"award-info":[{"award-number":["444610\/2024-3"]}],"id":[{"id":"10.13039\/501100003593","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"<jats:sec id=\"sec001\">\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Machine learning (ML) algorithms are increasingly used in healthcare to support clinical decision-making. While models with similar overall performance are often considered interchangeable for deployment, they may produce divergent predictions, a phenomenon known as algorithmic multiplicity. In such cases, the choice of algorithm may introduce bias. This study investigates the impacts of algorithmic multiplicity in mortality prediction and assesses the influence of patient characteristics on model decisions.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec002\">\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>A cohort of 4,337 adult patients (\u226518 years) with RT-PCR\u2013confirmed covid-19 from five tertiary care hospitals in Brazil was followed from March to August 2020. Five popular ML models for structured data were trained on demographic and laboratory data collected at early hospital admission to predict in-hospital mortality. Model performance, feature importance, and algorithmic prediction similarity were evaluated. Feature distributions were compared between patients correctly or incorrectly classified by all models using paired t-tests or Mann\u2013Whitney U tests, as applicable, at the 5% significance level. Subgroup performance differences were assessed using 10-fold cross-validation applied to five k-means\u2013delineated clusters, compared by one-way ANOVA. Within-cluster predictive divergence was assessed within a 95% confidence interval.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec003\">\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>All models achieved high overall predictive performance (\u00b5\u2009=\u20090.855, \u03c3\u00b2\u2009=\u20090.0072). However, the comparison of individual-level predictions revealed substantial heterogeneity, with pairwise prediction correlations ranging from R\u00b2\u2009=\u20090.56 to 0.80. Unsupervised k-means clustering identified five clinically distinct patient subgroups with mortality rates ranging from 22% to 80%, within which model performance varied significantly (F\u2009=\u200973.18, p\u2009&lt;\u20090.001). Notably, TabPFN and LightGBM showed superior performance in the \u201cAnemia\u201d cluster, whereas TabPFN underperformed in the \u201cImmunodeficient\u201d cluster (95% CI).<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"sec004\">\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>This study demonstrates that ML models with similar overall performance can yield substantially divergent predictions at both the individual and subgroup levels, and that no single algorithm consistently outperforms others across all patient subgroups. These findings highlight the limitations of relying solely on global performance metrics and underscore the need for context-aware evaluation of ML models in heterogeneous clinical populations.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1371\/journal.pone.0344354","type":"journal-article","created":{"date-parts":[[2026,3,6]],"date-time":"2026-03-06T18:40:37Z","timestamp":1772822437000},"page":"e0344354","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Predictive divergence in machine learning models for clinical mortality risk: A multicohort study of covid-19 patients"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0009-0009-8409-297X","authenticated-orcid":true,"given":"J\u00falia Chaves Neuenschwander","family":"Magalh\u00e3es","sequence":"first","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":[[2026,3,6]]},"reference":[{"issue":"2","key":"pone.0344354.ref001","doi-asserted-by":"crossref","first-page":"94","DOI":"10.7861\/futurehosp.6-2-94","article-title":"The potential for artificial intelligence in healthcare","volume":"6","author":"T Davenport","year":"2019","journal-title":"Future Healthc J"},{"key":"pone.0344354.ref002","doi-asserted-by":"crossref","first-page":"277","DOI":"10.1016\/B978-0-323-98352-5.00010-0","article-title":"Machine learning applications for COVID-19: a state-of-the-art review","author":"F Kamalov","year":"2023","journal-title":"Data Science for Genomics. Elsevier"},{"key":"pone.0344354.ref003","doi-asserted-by":"crossref","first-page":"104791","DOI":"10.1016\/j.ijmedinf.2022.104791","article-title":"The accuracy of machine learning approaches using non-image data for the prediction of COVID-19: a meta-analysis","volume":"164","author":"K-M Kuo","year":"2022","journal-title":"Int J Med Inform"},{"key":"pone.0344354.ref004","author":"N Maslej","year":"2025"},{"issue":"7","key":"pone.0344354.ref005","doi-asserted-by":"crossref","first-page":"1093","DOI":"10.1093\/jamia\/ocaf065","article-title":"Adoption of artificial intelligence in healthcare: survey of health system priorities, successes, and challenges","volume":"32","author":"EG Poon","year":"2025","journal-title":"J Am Med Inform Assoc"},{"key":"pone.0344354.ref006","doi-asserted-by":"crossref","unstructured":"Black E, Raghavan M, Barocas S. Model multiplicity: Opportunities, concerns, and solutions. In: Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency. 2022.","DOI":"10.1145\/3531146.3533149"},{"issue":"3","key":"pone.0344354.ref007","doi-asserted-by":"crossref","DOI":"10.1214\/ss\/1009213726","article-title":"Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author)","volume":"16","author":"L Breiman","year":"2001","journal-title":"Statist Sci"},{"key":"pone.0344354.ref008","author":"C Marx","year":"2020"},{"issue":"9","key":"pone.0344354.ref009","doi-asserted-by":"crossref","first-page":"10306","DOI":"10.1609\/aaai.v37i9.26227","article-title":"Predictive multiplicity in probabilistic classification","volume":"37","author":"J Watson-Daniels","year":"2023","journal-title":"AAAI"},{"key":"pone.0344354.ref010","author":"AP Meyer","year":"2024"},{"issue":"8045","key":"pone.0344354.ref011","doi-asserted-by":"crossref","first-page":"319","DOI":"10.1038\/s41586-024-08328-6","article-title":"Accurate predictions on small data with a tabular foundation model","volume":"637","author":"N Hollmann","year":"2025","journal-title":"Nature"},{"issue":"1","key":"pone.0344354.ref012","doi-asserted-by":"crossref","first-page":"1022","DOI":"10.1038\/s41598-022-26467-6","article-title":"Improving the performance of machine learning algorithms for health outcomes predictions in multicentric cohorts","volume":"13","author":"RM Wichmann","year":"2023","journal-title":"Sci Rep"},{"issue":"12","key":"pone.0344354.ref013","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pdig.0000699","article-title":"Multicenter comparative analysis of local and aggregated data training strategies in COVID-19 outcome prediction with Machine learning","volume":"3","author":"C Savalli","year":"2024","journal-title":"PLOS Digit Health"},{"key":"pone.0344354.ref014","doi-asserted-by":"crossref","first-page":"114832","DOI":"10.1016\/j.eswa.2021.114832","article-title":"Identifying mortality factors from Machine Learning using Shapley values - a case of COVID19","volume":"176","author":"M Smith","year":"2021","journal-title":"Expert Syst Appl"},{"key":"pone.0344354.ref015","doi-asserted-by":"crossref","unstructured":"Hsu H, Calmon F. Rashomon capacity: A metric for predictive multiplicity in classification. In: Advances in Neural Information Processing Systems. 2022;28988\u20139000.","DOI":"10.52202\/068431-2101"},{"key":"pone.0344354.ref016","doi-asserted-by":"crossref","first-page":"111","DOI":"10.1038\/s43856-022-00165-w","article-title":"Subpopulation-specific machine learning prognosis for underrepresented patients with double prioritized bias correction","volume":"2","author":"S Afrose","year":"2022","journal-title":"Commun Med (Lond)"},{"issue":"4","key":"pone.0344354.ref017","doi-asserted-by":"crossref","first-page":"848","DOI":"10.1109\/TEVC.2022.3209544","article-title":"Mitigating Unfairness via Evolutionary Multiobjective Ensemble Learning","volume":"27","author":"Q Zhang","year":"2023","journal-title":"IEEE Trans Evol Computat"},{"issue":"7","key":"pone.0344354.ref018","doi-asserted-by":"crossref","first-page":"4763","DOI":"10.1109\/TPAMI.2024.3357847","article-title":"Advances and challenges in meta-learning: a technical review","volume":"46","author":"A Vettoruzzo","year":"2024","journal-title":"IEEE Trans Pattern Anal Mach Intell"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0344354","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,6]],"date-time":"2026-03-06T18:40:41Z","timestamp":1772822441000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0344354"}},"subtitle":[],"editor":[{"given":"Marcela","family":"Pagano","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,3,6]]},"references-count":18,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2026,3,6]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0344354","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,6]]}}}