Abstract: TH-OR25
Enhancing Nephrology with Artificial Intelligence (AI)-Assisted ICD-10 Coding: Improving Health Care Reimbursement, Patient Care, Research, and Previsit Test Workflow Efficiency through Case Scenarios
Session Information
- Augmented Intelligence and Digital Health Advances
October 24, 2024 | Location: Room 2, Convention Center
Abstract Time: 05:30 PM - 05:40 PM
Category: Augmented Intelligence, Digital Health, and Data Science
- 300 Augmented Intelligence, Digital Health, and Data Science
Authors
- Abdelgadir, Yasir, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Thongprayoon, Charat, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Miao, Jing, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Pham, Justin, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Suppadungsuk, Supawadee, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Craici, Iasmina, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Cheungpasitporn, Wisit, Mayo Clinic Minnesota, Rochester, Minnesota, United States
Background
Accurate ICD-10 coding is crucial for healthcare reimbursement, patient care, and research. Inaccuracies in coding, can lead to issues with reimbursement, compromised patient care and skewed research findings. These challenges stem from the complexity and time-consuming nature of coding tasks, burdening physicians. AI, such as ChatGPT, has the potential to improve accuracy and reduce workload. A recent study showed that large language models (LLMs) performed poorly on medical code querying tasks. However, the assessment of AI-assisted ICD-10 coding in nephrology through case scenarios for pre-visit testing remains unexplored.
Methods
This study involved 100 simulated cases covering various nephrology conditions. Two nephrologists created these cases to mirror typical nephrology diagnoses in inpatient and outpatient settings, incorporating case scenarios and pre-visit testing data. The performance of ChatGPT versions 3.5 and 4.0 was assessed by comparing AI-generated ICD-10 codes against expected correct codes for each case. Assessments were conducted in two trials, two weeks apart, in April 2024.
Results
ChatGPT 3.5 reached an average accuracy of 89% over two trials, with a slight variation between them (91% in the first trial and 87% in the second). ChatGPT 4.0 showed a consistent and higher accuracy of 99% in both trials. This indicates that 94% of the AI-coded diagnoses correctly matched the expected ICD-10 codes. However, specific challenges were noted with ChatGPT 3.5, particularly with repeated inaccuracies in diagnosing conditions like Obstructive uropathy due to BPH and Acute Emphysematous Pyelonephritis. ChatGPT 4.0 demonstrated a more stable performance, with only one repeated misdiagnosis (Bartter Syndrome) across the trials.
Conclusion
The findings suggest that AI, particularly ChatGPT 4.0, can significantly improve the accuracy of ICD-10 coding in nephrology when provided with case scenarios and pre-visit testing data. This approach can enhance healthcare reimbursement, patient care, research, and pre-visit test nephrology workflow efficiency.However, the small percentage of errors highlights the need for ongoing review and improvement of AI diagnostic systems.