Can a Novel Natural Language Processing Model and Artificial Intelligence Automatically Generate Billing Codes From...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Spine Surgery Coding

Background:

• Billing and coding administrative tasks represent a significant healthcare expenditure in the US.
• Accurate CPT code assignment is crucial for reimbursement and healthcare system efficiency.

Purpose of the Study:

• To evaluate the efficacy of a Natural Language Processing (NLP) machine learning algorithm, XLNet, in automating the generation of CPT codes from operative notes.
• To assess the performance of XLNet for anterior cervical discectomy and fusion (ACDF), posterior cervical discectomy and fusion (PCDF), and cervical disc arthroplasty (CDA) procedures.

Main Methods:

• A retrospective cohort of 922 operative notes from ACDF, PCDF, and CDA procedures (2015-2020) was analyzed.
• The XLNet model was trained on these notes, with CPT codes from the billing department serving as the ground truth.
• Model performance was evaluated using Area Under the Receiver Operating Characteristic Curve (AUROC) and Area Under the Precision-Recall Curve (AUPRC).

Main Results:

• XLNet demonstrated high performance, approaching human accuracy in CPT code generation.
• Across different procedural combinations, AUROC values ranged from 0.82 to 0.95, and AUPRC values ranged from 0.70 to 0.91.
• Class-by-class accuracy reached up to 88% for combined procedures (ACDF, PCDF, CDA).

Conclusions:

• The XLNet model is effective in generating CPT billing codes from orthopedic operative notes.
• Advancements in NLP and AI can significantly enhance the medical billing process, improving accuracy and standardization.