Clinical Scorecard: AI in Surgery: Debate Highlights Benefits, Gaps
At a Glance
Category
Detail
Condition
Surgical workflow and outcomes
Key Mechanisms
AI models for forecasting supply needs, scheduling optimization, automated documentation, perioperative chatbots, postoperative monitoring via computer vision and time-series analysis, and AI-assisted surgical training
Target Population
Surgical patients including diverse demographic groups
Care Setting
Surgical and perioperative care settings
Key Highlights
AI use associated with reductions in operative times (25%), recovery periods (15%), intraoperative complications (30%), and a 40% improvement in surgical precision
Concerns include trust, safety, reliability, underrepresentation of minorities, and risks of algorithmic errors such as hallucinations
Many AI models lack external validation and publicly accessible data sets, raising issues of reproducibility and transparency
Guideline-Based Recommendations
Diagnosis
Use AI tools cautiously given current limitations in validation and data inclusivity
Management
Integrate AI applications in workflow optimization and postoperative monitoring with clinician oversight
Avoid premature adoption to prevent unintended burdens and errors
Monitoring & Follow-up
Implement rigorous validation and continuous monitoring of AI system performance
Ensure inclusive data development to mitigate disparities
Risks
Be aware of potential algorithmic hallucinations and spurious correlations
Recognize risks of perpetuating disparities due to underrepresentation in data sets
Consider risks of added complexity and new error sources similar to early electronic health record implementations
Patient & Prescribing Data
Surgical patients across diverse demographics including underrepresented groups
AI applications show promise in improving surgical precision and reducing complications but require further validation before widespread clinical use
Clinical Best Practices
Engage clinicians actively in AI system development and implementation
Prioritize evidence-based, patient-centered integration of AI
Ensure robust governance and transparency in AI applications
Promote inclusive data collection to address disparities
Validate AI models externally before clinical deployment
Researchers urge caution in interpreting joint replacement predictors, noting that surgery reflects access and decision-making as well as disease biology.
