165 Can Artificial Intelligence Successfully Identify CPT Codes for Common Procedures in Neurosurgery?

INTRODUCTION: Coding for Neurosurgical procedures is a complex process that is dynamically changing year to year, through the introduction and removal of codes and modifiers every year. It is an increasing challenge for Neurosurgeons to keep up with the changes and implement them in their practice to ensure accurate coding and maximize billing. The AANS reports only 43% of Neurosurgeons are estimated to have certified professional coders. METHODS: ChatGPT was asked to provide CPT codes for common neurosurgical procedures of the brain and spine with a given indication for the procedure. The responses of ChatGPT were recorded and compared to the accurate CPT codes used for the procedure. The total amount of RVUs provided by Chat GPT CPT codes were compared to the actual amount of RVUs for the given procedure. RESULTS: ChatGPT was asked for four cranial and spinal procedures each. In terms of accuracy, three responses were completely inaccurate, four responses were partially correct, and one response was completely correct. From the total 8 cases, ChatGPT provided CPT codes accounting for 132.88 of 227.33 expected RVUs (58.5%). For spinal procedures ChatGPT provided 81.7 of 107.8 expected RVUs (75.8%). For cranial procedures ChatGPT provided 51.2 of 119.5 expected RVUs (42.8%). CONCLUSIONS: The ability of ChatGPT to successfully code for neurosurgical procedures appears to be limited at this time, but it is not without promise. The information to successfully code for billing for medical procedures is not readily available on the internet. Using limited information ChatGPT is still able to successfully identify codes being used for procedures. Teaching artificial intelligence to use CPT codes for billing could offer neurosurgeons increased efficiency and efficacy in their billing in the future.