Digital Twin Models for Personalized Surgical Risk Assessment

Abstract

Population-derived surgical risk calculators such as the ACS NSQIP tool estimate a patient's complication risk from the aggregated outcomes of similar patients, while digital twin modeling proposes instead building a continuously updated, patient-specific computational representation of an individual's physiology to simulate how that patient will respond to a proposed operation. This review examines where that ambition currently stands in perioperative and surgical practice, from the architectural proposals for a multi-organ perioperative human digital twin to its most mature clinical application in cardiac and structural heart surgery. Digital-twin-guided modeling of patient-specific atrial substrate has been used to stratify and prevent postoperative atrial fibrillation after cardiac surgery, and patient-specific anatomic simulation now informs device selection and procedural rehearsal in transcatheter and aortic interventions, representing the field's most externally grounded applications to date. Extension to trauma, general, and noncardiac surgery currently takes the form of continuously updating multimodal predictive analytics rather than the fully mechanistic multi-organ twins described in foundational proposals, and is constrained by the difficulty of acquiring longitudinal multimodal patient data, the computational cost of clinically timed simulation, and the absence of validation and regulatory pathways suited to continuously adapting models. We conclude that digital twin methods are a promising complement to, rather than yet a replacement for, externally validated population risk calculators in most surgical settings, with cardiac and structural heart applications furthest along the path toward routine clinical use.