Optimising vitrectomy operation note coding with machine learning

• DOI: 10.1111/ceo.14257
• Publication Date: 2023-05-24T02:34:14Z
• Authors: Yong Min Lee; Stephen Bacchi; David Sia; Robert J. Casson; WengOnn Chan
• Keywords: Current Procedural Terminology; Reimbursement; Medical classification

The accurate encoding of operation notes is essential for activity-based funding and workforce planning. aim this project was to evaluate the procedural coding accuracy vitrectomy develop machine learning, natural language processing (NLP) models that may assist with task.This retrospective cohort study involved between a 21-month period at Royal Adelaide Hospital. Coding procedures were based on Medicare Benefits Schedule (MBS)-the Australian equivalent Current Procedural Terminology (CPT®) codes used in United States. Manual conducted all reviewed by two vitreoretinal consultants. XGBoost, random forest logistic regression developed classification experiments. A cost-based analysis subsequently conducted.There total 1724 individual performed within 617 totalling $1 528 086.60 after manual review. 1147 (66.5%) missed original amounted $736 539.20 (48.2%). Our XGBoost model had highest (94.6%) multi-label five most common procedures. successful identifying or more missing an AUC 0.87 (95% CI 0.80-0.92).Machine learning has been note encoding. We recommend combined human approach clinical as automation facilitate reimbursement enable surgeons prioritise higher quality care.