Evaluation of an AI-Assisted Virtual Fracture Clinic System: Common Referral Deficiencies and Outcomes Following Automated Feedback in a Major Trauma Centre.
- 1Orthopaedics and Trauma, Charing Cross Hospital, London, GBR.
Abstract
Introduction
Virtual fracture clinics (VFCs) rely on accurate and complete referral information to facilitate effective clinical review. Deficiencies in documentation or errors in the selection of the referral pathway result in additional clinical work. This study evaluated an artificial intelligence (AI)-assisted VFC referral feedback system that checks referrals against predefined local VFC criteria and provides automated feedback when required information is missing. We discuss the common referral deficiencies identified and improvements to referrals following automated feedback.Methods
A retrospective service evaluation was conducted at Imperial College Healthcare NHS Trust between 18 December 2025 and 26 March 2026. Referrals were made using a cloud-based system called FollowUp.Clinic, which incorporates an AI-assisted referral feedback function. The system assesses referrals against predefined local VFC criteria and generates automated feedback when deficiencies are identified. This allows referrers to amend their submissions before clinical review. No patient-identifiable information was transmitted to OpenAI (San Francisco, CA, USA) through the API.Results
A total of 1,201 referrals were processed. Of these, 901 (75.0%) met the predefined referral criteria without requiring modification, while 300 (25.0%) were flagged for modification. The most frequent deficiency was specific imaging not being requested, identified in 167 referrals (55.7%), followed by missing venous thromboembolism (VTE) documentation in 70 (23.3%), insufficient clinical information in 65 (21.7%), and an incorrect referral pathway in 18 (6.0%). Categories were not mutually exclusive, as referrals could contain multiple deficiencies. Following one round of automated feedback, 282 of 300 initially flagged referrals (94.0%) were amended and subsequently progressed through the VFC pathway. The remaining 18 (6.0%) required redirection to an alternative pathway.Discussion
Most referral deficiencies identified by the system were modifiable, with 94.0% successfully corrected following a single round of automated feedback. Specific imaging not being requested represented the most frequent deficiency. Missing required imaging could add extra steps before treatment. Missing VTE documentation was the second most frequent issue. In the absence of documented VTE assessment, the VFC team had to assume that this had not been completed, with potential implications for patient safety.Conclusion
The AI-assisted referral system identified common, potentially modifiable deficiencies in VFC referrals, with 94.0% of initially flagged referrals successfully corrected following a single feedback cycle.
Related
- Quality, Readability, and Content Deficiencies of Three Large Language Models in Bee Sting Health Information: A Comparative Evaluation with Patient Perspectives
- Microscope-Assisted Excision of CT-Occult Inferior Rectus Adhesions After Pediatric Orbital Floor Trapdoor Fracture Repair: A Case Report
- Fracture and Vertebral Augmentation Utilization Following Androgen Deprivation Therapy in Men With Prostate Cancer: A Propensity Score–matched Cohort Study
- Contralateral atypical femoral fracture following a single dose of denosumab after incomplete teriparatide therapy in a patient with long-term bisphosphonate exposure: A case report
- Artificial Intelligence‐Assisted Automated Optical Coherence Tomography Analysis for Percutaneous Coronary Intervention Planning: From Plaque Analysis to Precision Lesion Preparation and Stent Optimization
BibTeX
@article{ooi2026evaluation,
title = {{Evaluation of an AI-Assisted Virtual Fracture Clinic System: Common Referral Deficiencies and Outcomes Following Automated Feedback in a Major Trauma Centre.}},
author = {Wern Y Ooi},
journal = {Cureus},
year = {2026},
month = sep,
doi = {10.7759/cureus.116556},
url = {https://europepmc.org/article/MED/42899297},
note = {CC BY},
}
RIS
TY - JOUR
TI - Evaluation of an AI-Assisted Virtual Fracture Clinic System: Common Referral Deficiencies and Outcomes Following Automated Feedback in a Major Trauma Centre.
AU - Wern Y Ooi
PY - 2026
DA - 2026/09/19
JO - Cureus
DO - 10.7759/cureus.116556
UR - https://europepmc.org/article/MED/42899297
AB - Introduction Virtual fracture clinics (VFCs) rely on accurate and complete referral information to facilitate effective clinical review. Deficiencies in documentation or errors in the selection of the referral pathway result in additional clinical work. This study evaluated an artificial intelligence (AI)-assisted VFC referral feedback system that checks referrals against predefined local VFC criteria and provides automated feedback when required information is missing. We discuss the common referral deficiencies identified and improvements to referrals following automated feedback.Methods A retrospective service evaluation was conducted at Imperial College Healthcare NHS Trust between 18 December 2025 and 26 March 2026. Referrals were made using a cloud-based system called FollowUp.Clinic, which incorporates an AI-assisted referral feedback function. The system assesses referrals against predefined local VFC criteria and generates automated feedback when deficiencies are identified. This allows referrers to amend their submissions before clinical review. No patient-identifiable information was transmitted to OpenAI (San Francisco, CA, USA) through the API.Results A total of 1,201 referrals were processed. Of these, 901 (75.0%) met the predefined referral criteria without requiring modification, while 300 (25.0%) were flagged for modification. The most frequent deficiency was specific imaging not being requested, identified in 167 referrals (55.7%), followed by missing venous thromboembolism (VTE) documentation in 70 (23.3%), insufficient clinical information in 65 (21.7%), and an incorrect referral pathway in 18 (6.0%). Categories were not mutually exclusive, as referrals could contain multiple deficiencies. Following one round of automated feedback, 282 of 300 initially flagged referrals (94.0%) were amended and subsequently progressed through the VFC pathway. The remaining 18 (6.0%) required redirection to an alternative pathway.Discussion Most referral deficiencies identified by the system were modifiable, with 94.0% successfully corrected following a single round of automated feedback. Specific imaging not being requested represented the most frequent deficiency. Missing required imaging could add extra steps before treatment. Missing VTE documentation was the second most frequent issue. In the absence of documented VTE assessment, the VFC team had to assume that this had not been completed, with potential implications for patient safety.Conclusion The AI-assisted referral system identified common, potentially modifiable deficiencies in VFC referrals, with 94.0% of initially flagged referrals successfully corrected following a single feedback cycle.
SN - 2168-8184
ER -