Evidence
Publications
Peer-reviewed validation is the foundation of clinical AI trust. Our five-paper roadmap documents the methodology, validation, and deployment of Marqi Index.
Papers In Submission
Paper 1
AMI 30-Day Readmission Prediction
Abstract
We present a 15-seed transformer ensemble with an XGBoost out-of-fold meta-learner for 30-day readmission prediction in acute myocardial infarction patients. The model was trained on structured EHR data from the TriNetX federated research network (n=5,153). On temporal validation, the model achieved an AUC of 0.807 (95% CI: 0.771-0.838), representing an improvement of 8-19 points over published LACE and HOSPITAL scores. We report full calibration metrics and disclose a data integrity event caught during development.
Publication Roadmap
A five-paper program documenting Marqi Index development, validation, and deployment.
AMI 30-Day Readmission
npj Digital Medicine
Foundational paper establishing the 15-seed transformer ensemble methodology and validation framework. Includes disclosure of data integrity event and five-control verification protocol.
All-Cause Readmission with LLM Distillation
Target: JAMA Network Open
Extends AMI methodology to all-cause readmission across HRRP conditions. Introduces LLM concept distillation for clinical interpretability. In active validation with academic medical center co-authors.
External Multi-Site Validation
Target: Annals of Internal Medicine
Independent recomputation of Marqi Index on health system partner data. Multi-site external validation with full calibration and discrimination metrics across diverse patient populations.
Cross-Note Longitudinal Extraction
Target: npj Digital Medicine
Introduces Marqi Index v2 with structured signal extraction from clinical notes. Cross-note longitudinal extraction for signals not captured in structured EHR fields.
Clinical Impact Study
Target: TBD
Prospective evaluation of Marqi Index impact on readmission rates and care management efficiency. Outcome-based assessment of clinical deployment.
Research Collaboration
We are actively seeking validation partners for Papers 3-5. If your health system is interested in independent validation of Marqi Index on your patient population, we would welcome a conversation.
Co-authorship
Validation partners receive co-authorship on resulting publications.
Data stays local
Validation can be performed on your infrastructure. Data does not need to leave your environment.
Full transparency
All validation results are published regardless of outcome. We do not suppress unfavorable findings.
Ideal Validation Partners
Academic Medical Centers
With established research infrastructure and IRB processes.
Health Systems
With diverse patient populations and structured EHR data availability.
Researchers
With expertise in clinical informatics, ML validation, or readmission prediction.
Interested in research collaboration?
Schedule a call to discuss validation partnership opportunities for Papers 3-5.