{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,6,26]],"date-time":"2026-06-26T20:48:01Z","timestamp":1782506881351,"version":"3.54.5"},"posted":{"date-parts":[[2026,6,24]]},"group-title":"Neurology","reference-count":43,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,6,24]],"date-time":"2026-06-24T00:00:00Z","timestamp":1782259200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000050","name":"National Heart Lung and Blood Institute","doi-asserted-by":"crossref","award":["R35HL155466"],"award-info":[{"award-number":["R35HL155466"]}],"id":[{"id":"10.13039\/100000050","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/100000050","name":"National Heart Lung and Blood Institute","doi-asserted-by":"crossref","award":["K24HL133373"],"award-info":[{"award-number":["K24HL133373"]}],"id":[{"id":"10.13039\/100000050","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/100000050","name":"National Heart Lung and Blood Institute","doi-asserted-by":"crossref","award":["K01HL171839"],"award-info":[{"award-number":["K01HL171839"]}],"id":[{"id":"10.13039\/100000050","id-type":"DOI","asserted-by":"crossref"}]},{"name":"The Alabama Genomic Health Initiative"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,6,24]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:sec>\n                  <jats:title>Importance<\/jats:title>\n                  <jats:p>Machine-learning models for ischemic stroke risk prediction are rarely validated across ancestrally distinct cohorts, and the contributions of polygenic risk scores (PRS) and self-reported race in such models remain unclear.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Objective<\/jats:title>\n                  <jats:p>To develop and externally validate a 10-year ischemic stroke risk model and quantify the incremental contributions of laboratory trajectories, PRS, and self-reported race and ethnicity across populations.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Design, Setting, and Participants<\/jats:title>\n                  <jats:p>\n                    Retrospective cohort study with model development in the All of Us (AoU) Research Program (n = 34,987; 1,920 incident strokes) and external validation in the Bio\n                    <jats:italic>Me<\/jats:italic>\n                    Biobank at Mount Sinai (n = 10,693; 107 incident strokes). Adults aged \u226545 years with \u22651 year of pre-baseline electronic health record data were anchored to a January 2010 baseline with 10-year follow-up.\n                  <\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Exposures<\/jats:title>\n                  <jats:p>Three XGBoost model tiers added laboratory feature trajectories (M2) and 20 PRS (M3) to clinical baseline features (M1); evaluated under race-blind and race-aware specifications.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Main Outcomes and Measures<\/jats:title>\n                  <jats:p>First inpatient ischemic stroke within 10 years; discrimination (area under the receiver operating characteristic curve [AUROC]) and calibration (observed-to-expected [O\/E] ratio).<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>\n                    In the AoU test partition (n = 6,998; 384 cases), M3 achieved AUROC 0.813 (95% CI, 0.788\u20130.837), outperforming the Revised Framingham Stroke Risk Profile (\u0394AUROC 0.164) and Pooled Cohort Equations (\u0394AUROC 0.181; both\n                    <jats:italic>P<\/jats:italic>\n                    &lt; 0.001). Discrimination transferred to Bio\n                    <jats:italic>Me<\/jats:italic>\n                    (AUROC 0.745) but predictions were systematically high (aggregate O\/E 0.12 vs. 1.00 in AoU), consistent with intercept-shift miscalibration; Bio\n                    <jats:italic>Me<\/jats:italic>\n                    -fit intercept recalibration restored calibration in African American and Hispanic but not European American participants. PRS contribution was significant only in the Bio\n                    <jats:italic>Me<\/jats:italic>\n                    Hispanic (\u0394AUROC + 0.042;\n                    <jats:italic>P<\/jats:italic>\n                    = 0.003), with no significant within-stratum gain in five other cohort-by-race combinations. Adding self-reported race produced small gains when combined with PRS (Bio\n                    <jats:italic>Me<\/jats:italic>\n                    + 0.022,\n                    <jats:italic>P<\/jats:italic>\n                    = 0.034; AoU + 0.006,\n                    <jats:italic>P<\/jats:italic>\n                    = 0.052) but not when added without PRS.\n                  <\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Conclusions and Relevance<\/jats:title>\n                  <jats:p>A machine-learning ensemble combining clinical, laboratory, and polygenic features outperformed traditional risk scores by 0.16\u20130.18 AUROC and retained discriminative validity in an ancestrally distinct external cohort but required site-specific recalibration of absolute risk. The marginal contribution of self-reported race overlapped with polygenic signal, supporting per-ancestry calibration over universal race-aware model deployment.<\/jats:p>\n                  <jats:fig id=\"ufig1\" position=\"float\" orientation=\"portrait\" fig-type=\"figure\">\n                    <jats:graphic xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"26356280v1_ufig1\" position=\"float\" orientation=\"portrait\"\/>\n                  <\/jats:fig>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Graphical Abstract<\/jats:title>\n                  <jats:p>\n                    Study design and analytic workflow. The All of Us (AoU) Research Program analytic cohort (n = 34,987; 1,920 incident ischemic strokes) was partitioned 60\/20\/20 into train, calibration, and held-out test sets. Two parallel model families were trained on the AoU train partition: full-feature models using all AoU features (evaluated only in AoU), and harmonized models using the intersection of AoU and Bio\n                    <jats:italic>Me<\/jats:italic>\n                    features (evaluated in both cohorts). Each family comprised three nested tiers (M1, clinical baseline; M2, M1 plus laboratory feature trajectories; M3, M2 plus 20 polygenic risk scores). Three specifications of the harmonized models addressed ancestry-related questions: full-cohort race-blind (primary cross-cohort analyses), ancestry-aligned race-blind, and ancestry-aligned race-aware. The Bio\n                    <jats:italic>Me<\/jats:italic>\n                    Biobank cohort at Mount Sinai (n = 10,693; 107 incident ischemic strokes) served as the external validation set under matched eligibility criteria and an identical inpatient ischemic stroke outcome definition. Internal evaluation in the AoU test partition included discrimination, calibration, comparison with the Revised Framingham Stroke Risk Profile (rFSRP) and Pooled Cohort Equations (PCE), Shapley additive explanations (SHAP) feature attribution, and subgroup analyses. Cross-cohort analyses evaluated discrimination, calibration (observed-to-expected [O\/E] ratio), per-ancestry performance, domain-level SHAP attribution, and the marginal contributions of polygenic risk scores and self-reported race.\n                  <\/jats:p>\n                <\/jats:sec>","DOI":"10.64898\/2026.06.22.26356280","type":"posted-content","created":{"date-parts":[[2026,6,25]],"date-time":"2026-06-25T03:30:11Z","timestamp":1782358211000},"source":"Crossref","is-referenced-by-count":0,"title":["Development and External Validation of a Machine Learning Model for 10-Year Ischemic Stroke Risk Prediction in Diverse Populations"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-7253-199X","authenticated-orcid":false,"given":"Ahmed","family":"Khattab","sequence":"first","affiliation":[{"name":"Department of Epidemiology, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhe","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Epidemiology, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA"},{"name":"The Charles Bronfman Institute for Personalized Medicine, The Icahn School of Medicine at Mount Sinai, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Vinodh","family":"Srinivasasainagendra","sequence":"additional","affiliation":[{"name":"Department of Biostatistics, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hemant K.","family":"Tiwari","sequence":"additional","affiliation":[{"name":"Department of Biostatistics, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ruth","family":"Loos","sequence":"additional","affiliation":[{"name":"The Charles Bronfman Institute for Personalized Medicine, The Icahn School of Medicine at Mount Sinai, NY, USA"},{"name":"The Novo Nordisk Foundation Center for Basic Metabolic Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nita","family":"Limdi","sequence":"additional","affiliation":[{"name":"Department of Neurology, Heersink School of Medicine, The University of Alabama at Birmingham, AL, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marguerite Ryan","family":"Irvin","sequence":"additional","affiliation":[{"name":"Department of Epidemiology, School of Public Health, The University of Alabama at Birmingham, Birmingham, AL, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026062613150739000_2026.06.22.26356280v1.1","doi-asserted-by":"publisher","DOI":"10.1177\/17474930241308142"},{"key":"2026062613150739000_2026.06.22.26356280v1.2","doi-asserted-by":"publisher","DOI":"10.1161\/STR.0000000000000475"},{"key":"2026062613150739000_2026.06.22.26356280v1.3","first-page":"23","article-title":"RACIAL DISPARITIES IN STROKE AFFECTING BLACK AMERICANS","volume":"135","year":"2025","journal-title":"Trans Am Clin Climatol Assoc"},{"key":"2026062613150739000_2026.06.22.26356280v1.4","doi-asserted-by":"publisher","DOI":"10.1161\/STROKEAHA.121.036263"},{"key":"2026062613150739000_2026.06.22.26356280v1.5","doi-asserted-by":"publisher","DOI":"10.1161\/CIRCULATIONAHA.115.021275"},{"key":"2026062613150739000_2026.06.22.26356280v1.6","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacc.2013.11.005"},{"key":"2026062613150739000_2026.06.22.26356280v1.7","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-01855-w"},{"key":"2026062613150739000_2026.06.22.26356280v1.8","doi-asserted-by":"publisher","DOI":"10.3389\/fneur.2025.1668420"},{"key":"2026062613150739000_2026.06.22.26356280v1.9","doi-asserted-by":"publisher","DOI":"10.1016\/j.jnrt.2024.100156"},{"key":"2026062613150739000_2026.06.22.26356280v1.10","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijmedinf.2025.106041"},{"key":"2026062613150739000_2026.06.22.26356280v1.11","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMms2004740"},{"key":"2026062613150739000_2026.06.22.26356280v1.12","doi-asserted-by":"publisher","DOI":"10.1089\/heq.2023.0038"},{"key":"2026062613150739000_2026.06.22.26356280v1.13","doi-asserted-by":"publisher","DOI":"10.1186\/s40246-024-00664-y"},{"key":"2026062613150739000_2026.06.22.26356280v1.14","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-019-0379-x"},{"key":"2026062613150739000_2026.06.22.26356280v1.15","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMsr1809937"},{"key":"2026062613150739000_2026.06.22.26356280v1.16","doi-asserted-by":"publisher","DOI":"10.1016\/J.CELL.2021.03.034"},{"key":"2026062613150739000_2026.06.22.26356280v1.17","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-021-00783-5"},{"key":"2026062613150739000_2026.06.22.26356280v1.18","doi-asserted-by":"publisher","DOI":"10.1186\/s12864-025-11693-9"},{"key":"2026062613150739000_2026.06.22.26356280v1.19","doi-asserted-by":"publisher","DOI":"10.1145\/2939672.2939785"},{"key":"2026062613150739000_2026.06.22.26356280v1.20","doi-asserted-by":"publisher","DOI":"10.1145\/3292500.3330701"},{"key":"2026062613150739000_2026.06.22.26356280v1.21","unstructured":"Platt J . Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods. Adv Large Margin Classif. 2000;10."},{"key":"2026062613150739000_2026.06.22.26356280v1.22","first-page":"307","volume-title":"Contributions to the Theory of Games","volume":"II","year":"2016"},{"key":"2026062613150739000_2026.06.22.26356280v1.23","unstructured":"Lundberg SM , Lee SI . A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. NIPS\u201917. Curran Associates Inc.; 2017:4768\u20134777. Accessed May 5, 2026. https:\/\/dl.acm.org\/doi\/10.5555\/3295222.3295230"},{"key":"2026062613150739000_2026.06.22.26356280v1.24","doi-asserted-by":"publisher","DOI":"10.2307\/2531595"},{"key":"2026062613150739000_2026.06.22.26356280v1.25","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0232414"},{"key":"2026062613150739000_2026.06.22.26356280v1.26","doi-asserted-by":"publisher","DOI":"10.3390\/jcdd11070207"},{"key":"2026062613150739000_2026.06.22.26356280v1.27","doi-asserted-by":"publisher","DOI":"10.3389\/fcvm.2022.901240"},{"key":"2026062613150739000_2026.06.22.26356280v1.28","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijnss.2025.10.011"},{"key":"2026062613150739000_2026.06.22.26356280v1.29","doi-asserted-by":"publisher","DOI":"10.1161\/STROKEAHA.125.050961"},{"key":"2026062613150739000_2026.06.22.26356280v1.30","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-021-95244-8"},{"key":"2026062613150739000_2026.06.22.26356280v1.31","doi-asserted-by":"publisher","DOI":"10.1161\/CIRCOUTCOMES.117.004197"},{"key":"2026062613150739000_2026.06.22.26356280v1.32","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-018-36745-x"},{"key":"2026062613150739000_2026.06.22.26356280v1.33","doi-asserted-by":"publisher","DOI":"10.3389\/fneur.2021.784250"},{"key":"2026062613150739000_2026.06.22.26356280v1.34","doi-asserted-by":"publisher","DOI":"10.1186\/s12916-019-1466-7"},{"key":"2026062613150739000_2026.06.22.26356280v1.35","doi-asserted-by":"publisher","DOI":"10.1161\/JAHA.124.040386"},{"key":"2026062613150739000_2026.06.22.26356280v1.36","doi-asserted-by":"publisher","DOI":"10.1093\/eurheartj\/ehac648"},{"key":"2026062613150739000_2026.06.22.26356280v1.37","doi-asserted-by":"publisher","DOI":"10.1161\/STROKEAHA.117.018928"},{"key":"2026062613150739000_2026.06.22.26356280v1.38","doi-asserted-by":"publisher","DOI":"10.1161\/CIRCULATIONAHA.117.028429"},{"key":"2026062613150739000_2026.06.22.26356280v1.39","doi-asserted-by":"publisher","DOI":"10.7326\/M17-3011"},{"key":"2026062613150739000_2026.06.22.26356280v1.40","doi-asserted-by":"publisher","DOI":"10.1161\/JAHA.125.041549"},{"issue":"2","key":"2026062613150739000_2026.06.22.26356280v1.41","first-page":"e27","article-title":"Positive predictive values of ICD-9 codes to identify patients with stroke or TIA","volume":"20","year":"2014","journal-title":"Am J Manag Care"},{"key":"2026062613150739000_2026.06.22.26356280v1.42","doi-asserted-by":"publisher","DOI":"10.1002\/pds.2312"},{"key":"2026062613150739000_2026.06.22.26356280v1.43","doi-asserted-by":"publisher","DOI":"10.1002\/pds.1518"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.06.22.26356280","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,26]],"date-time":"2026-06-26T20:15:18Z","timestamp":1782504918000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.64898\/2026.06.22.26356280"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,24]]},"references-count":43,"URL":"https:\/\/doi.org\/10.64898\/2026.06.22.26356280","relation":{},"subject":[],"published":{"date-parts":[[2026,6,24]]},"subtype":"preprint"}}