Previously submitted to: JMIR Medical Informatics (no longer under consideration since Jul 21, 2026)
Date Submitted: May 1, 2026
Open Peer Review Period: May 15, 2026 - Jul 10, 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 Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn (a note "no longer under consideration" will appear above).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Warning: This is an author submission that is not peer-reviewed or edited.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 for dynamic risk prediction of hemodynamic deterioration in acute myocardial infarction with cardiogenic shock: development, external validation, and clinical decision support deployment using the MIMIC-IV and eICU-CRD databases
ABSTRACT
Background:
Current risk stratification in acute myocardial infarction complicated by cardiogenic shock (AMI-CS) relies on static admission-time scores that cannot capture a patient's dynamic response to initial resuscitation. We developed and externally validated a parsimonious, interpretable machine learning model—V20-lite—that predicts hemodynamic deterioration during the critical 24- to 48-hour window using only the physiological trajectory from the first 24 hours of intensive care.
Methods:
We conducted a retrospective cohort study using the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database for model development and internal validation, with the eICU Collaborative Research Database (eICU-CRD, version 2.0) serving as an independent, geographically diverse external validation cohort. Adult patients with AMI-CS were identified using ICD-9/10 codes. The primary outcome was a composite of hemodynamic deterioration occurring between 24 and 48 hours after ICU admission, defined by sustained hypotension, worsening hyperlactatemia, progressive oliguria, acute kidney injury, vasopressor escalation, or death. We engineered 70 candidate features from the 0–24-hour window and trained seven machine learning algorithms. SHapley Additive exPlanations (SHAP) were used to distill an 8-feature composite model (V20-lite) from the best-performing XGBoost algorithm. We assessed discrimination, calibration (Integrated Calibration Index [ICI], Brier score), and clinical utility (decision curve analysis). Logistic recalibration was pre-specified to address calibration drift in external validation. The model was deployed as an open-access Shiny application.
Results:
The derivation cohort included 1,633 patients (mean age 70.5 years, 61.8% male) from MIMIC-IV; 653 patients (mean age 68.1 years, 61.9% male) from eICU-CRD comprised the external validation cohort. The primary outcome occurred in 482 (29.5%) and 246 (37.7%) patients, respectively. In the internal test set (n=327, event rate 30.6%), V20-lite achieved an area under the receiver operating characteristic curve (AUC) of 0.815 (95% CI, 0.763–0.867), significantly outperforming the admission Sequential Organ Failure Assessment (SOFA) score (AUC 0.639; DeLong P<0.001). Sensitivity analysis confirmed that removing trajectory data reduced AUC to 0.588. In external validation, the original model maintained robust discrimination (AUC 0.756; 95% CI, 0.716–0.795). Observed calibration drift (intercept −0.324; ICI 0.144) was fully corrected by logistic recalibration (intercept 0.000; ICI 0.032; Brier score 0.186). SHAP analysis identified the 24-hour creatinine ratio, 24-hour lactate, and lactate-to-MAP ratio as the dominant prognostic drivers. Subgroup analysis revealed reduced performance in patients with initially normal lactate (AUC 0.671; P for interaction <0.001).
Conclusions:
V20-lite provides accurate, interpretable, and dynamically informed risk stratification for early hemodynamic deterioration in AMI-CS using eight routine bedside parameters. The model's calibration is transportable across heterogeneous healthcare systems through simple logistic recalibration, and its open-access deployment facilitates immediate clinical application.
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.