Previously submitted to: JMIR Medical Informatics (no longer under consideration since Oct 23, 2026)
Date Submitted: Oct 22, 2026
Open Peer Review Period: Oct 23, 2026 - Oct 23, 2026
(closed for review but you can still tweet)
NOTE: This is an unreviewed Preprint
Warning: This is a unreviewed preprint (What is a preprint?). Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn (a note "no longer under consideration" will appear above).
Peer review me: Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period (in this case, a "Peer Review Me" button to sign up as reviewer is displayed above). All preprints currently open for review are listed here. Outside of the formal open peer-review period we encourage you to tweet about the preprint.
Citation: Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author).
Final version: If our system detects a final peer-reviewed "version of record" (VoR) published in any journal, a link to that VoR will appear below. Readers are then encourage to cite the VoR instead of this preprint.
Settings: If you are the author, you can login and change the preprint display settings, but the preprint URL/DOI is supposed to be stable and citable, so it should not be removed once posted.
Submit: To post your own preprint, simply submit to any JMIR journal, and choose the appropriate settings to expose your submitted version as preprint.
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Interpretable Machine Learning Models for Early Prediction of Sepsis in Elderly ICU Patients With Urinary Tract Infections: A Multicenter External Validation Study
ABSTRACT
Background:
Sepsis is a frequent and often abrupt complication of urinary tract infection (UTI) in elderly intensive care unit (ICU) patients. Machine learning may support early risk identification if predictors precede the outcome and performance holds in independent cohorts.
Objective:
To develop and externally validate interpretable machine learning models for early prediction of sepsis in elderly ICU patients with UTI, using information from the first 24 hours after UTI documentation.
Methods:
Patients aged 65 years or older with UTI documented during the ICU stay were identified in MIMIC-IV (n=5,310) and the eICU Collaborative Research Database (n=2,902). Patients meeting Sepsis-3 criteria before or during the 24-hour observation window, or with another primary infection source, were excluded. The outcome was incident Sepsis-3 sepsis after this window. MIMIC-IV was split into training (n=3,731) and internal validation (n=1,579) sets, and eICU-CRD served as external validation. Preprocessing, feature filtering (48 predictors retained), and tuning were fitted on the training cohort only. Nine algorithms were compared using ROC-AUC, precision-recall curves, Brier scores, decision curve analysis, and DeLong tests, with Shapley additive explanations (SHAP) for interpretation.
Results:
Sepsis incidence was 64.35%, 66.43%, and 64.33% in the training, internal validation, and external validation sets. Internal validation ROC-AUCs were 0.8754 for extreme gradient boosting (XGBoost), 0.8701 for light gradient boosting machine (LightGBM), and 0.8636 for random forest (RF). External validation ROC-AUCs were 0.8495, 0.8456, and 0.845, respectively, with no significant pairwise differences (P>.05). Brier scores were 0.135-0.141 internally and 0.151-0.157 externally. Net benefit exceeded treat-all and treat-none strategies at thresholds of 0.10-0.40. Serum potassium, vasopressor use, blood urea nitrogen, and creatinine were the leading SHAP contributors.
Conclusions:
The models showed consistent discrimination and acceptable calibration across two independent databases, using routinely available variables from a defined 24-hour window. They are candidates for prospective evaluation rather than deployable decision-support tools. Clinical Trial: Ethics Considerations: This study involved secondary analysis of publicly available, de-identified patient data from the MIMIC-IV and eICU-CRD databases. The use of MIMIC-IV data was approved by the MIT Institutional Review Board (Protocol #0403000206). Due to the de-identification of data in compliance with HIPAA regulations, this study was classified as non-human subjects research and did not require additional IRB approval. The researchers completed the required CITI certification to access MIMIC-IV and eICU-CRD data (CITI Record ID: 65922376).
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.