publications

2023

1. Retrieval-Augmented Generation for Clinical Knowledge Bases: Evaluation and Benchmarking

   Mohcine Madkour

   arXiv preprint, 2023

   Abs Bib

   A systematic evaluation of RAG architectures for clinical knowledge retrieval, covering embedding model selection, chunking strategies, and retrieval metric benchmarking (Hit Rate@K, MRR, NDCG) on clinical QA datasets.

   @article{madkour2023rag,
     title = {Retrieval-Augmented Generation for Clinical Knowledge Bases: Evaluation and Benchmarking},
     author = {Madkour, Mohcine},
     journal = {arXiv preprint},
     year = {2023},
   }

2021

1. FHIR-Based Integration Architecture for Real-Time Clinical Decision Support

   Mohcine Madkour

   In AMIA Annual Symposium Proceedings, 2021

   Abs Bib

   An architecture for bidirectional FHIR R4 data exchange between EHR systems and clinical AI applications, using Mirth Connect for HL7 translation and real-time streaming.

   @inproceedings{madkour2021fhir,
     title = {FHIR-Based Integration Architecture for Real-Time Clinical Decision Support},
     author = {Madkour, Mohcine},
     booktitle = {AMIA Annual Symposium Proceedings},
     year = {2021},
   }

2020

1. Clinical AI at Scale: Deploying Predictive Models in a Hospital Environment

   Mohcine Madkour and others

   Journal of the American Medical Informatics Association, 2020

   Abs Bib

   A study of the practical challenges and solutions for deploying clinical AI predictive models at UF Health, covering HL7 integration, clinical workflow design, and ongoing monitoring.

   @article{madkour2020clinical,
     title = {Clinical AI at Scale: Deploying Predictive Models in a Hospital Environment},
     author = {Madkour, Mohcine and others},
     journal = {Journal of the American Medical Informatics Association},
     year = {2020},
   }

2019

1. MySurgeryRisk: Development and Validation of a Machine-Learning Risk Algorithm for Major Complications and Death after Surgery

   Mohcine Madkour, Azra Bihorac, and others

   Annals of Surgery, 2019

   Abs Bib

   We developed MySurgeryRisk, a machine-learning algorithm that uses preoperative electronic health record data to predict postoperative complications. The algorithm achieved AUC 0.82–0.94 across multiple complication endpoints.

   @article{madkour2019mysurgeryrisk,
     title = {MySurgeryRisk: Development and Validation of a Machine-Learning Risk Algorithm for Major Complications and Death after Surgery},
     author = {Madkour, Mohcine and Bihorac, Azra and others},
     journal = {Annals of Surgery},
     year = {2019},
   }