Clinical implementation of an AI-based prediction model for decision support for patients undergoing colorectal cancer surgery

Andreas Weinberger Rosen ¹, Ilze Ose ², Mikail Gögenur ³

¹Department of Surgery, Center for Surgical Science, Zealand University Hospital, Køge, Denmark. anwr@regionsjaelland.dk.
²Department of Surgery, Center for Surgical Science, Zealand University Hospital, Køge, Denmark. ilos@regionsjaelland.dk.
³Department of Surgery, Center for Surgical Science, Zealand University Hospital, Køge, Denmark.
⁴Department of Anesthesiology, Zealand University Hospital, Køge, Denmark.
⁵Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark.
⁶Department of Physiotherapy, Zealand University Hospital, Køge, Denmark.
⁷Department of Medicine, Zealand University Hospital, Køge, Denmark.
⁸Department of Sports Science and Clinical Biomechanics, Center for Muscle and Joint Health, University of Southern Denmark, Odense, Denmark.
⁹Department of Physiotherapy and Occupational Therapy, Research and Implementation Unit PROgrez, Næstved-Slagelse-Ringsted Hospitals, Slagelse, Denmark.
¹⁰Department of Public Health, Danish Centre for Health Economics, University of Southern Denmark, Odense, Denmark.
¹¹Danish Colorectal Cancer Group, Copenhagen, Denmark.

⁰Department of Surgery, Center for Surgical Science, Zealand University Hospital, Køge, Denmark. anwr@regionsjaelland.dk.

Nature Medicine +

|

September 18, 2025

View abstract on PubMed

Summary

An artificial intelligence (AI) model predicts mortality risk in colorectal cancer surgery patients. Personalized perioperative treatment based on AI risk assessment significantly reduced complications and mortality, proving cost-effective.

Area Of Science

Surgical Oncology
Artificial Intelligence
Health Informatics

Background

Adverse outcomes in elective cancer surgery reduce survival and increase costs.
Identifying high-risk patients for personalized perioperative interventions in cancer surgery remains challenging.

Purpose Of The Study

To develop, validate, and implement an AI-based risk prediction model for personalized perioperative treatment in colorectal cancer surgery.
To utilize real-world data for a scalable, AI-driven decision support tool.

Main Methods

Developed and validated an AI risk prediction model using data from 18,403 colorectal cancer patients.
Implemented personalized treatment pathways based on predicted 1-year mortality risk.
Conducted a nonrandomized before/after cohort study to compare outcomes.

Main Results

The AI model achieved an area under the ROC curve of 0.79 in the validation set.
Personalized treatment reduced the incidence of comprehensive complication index >20 (19.1% vs. 28.0%) and any medical complication (23.7% vs. 37.3%).
Personalized perioperative treatment was found to be cost-effective.

Conclusions

AI-based risk prediction and personalized perioperative treatment can significantly improve surgical outcomes in cancer patients.
This registry-based, AI-driven approach is a scalable and cost-effective strategy for enhancing clinical practice.
Personalized treatment pathways tailored to individual risk profiles are crucial for optimizing perioperative care in cancer surgery.