Problem
Surgery pre-authorization is increasingly complex. This complexity slows the efficient provision of timely patient care. This challenge is dually driven by payors’ increasing documentation requirements and providers’ insufficient documentation. According to data from the Patient Revenue Management Office (PRMO), the Department of Orthopedics saw a 15% increase in surgery authorization requests from early 2023 to late 2024. Manual processes to resolve these challenges have become labor intensive, especially for the Orthopedic surgeons, administrative staff, and PRMO staff. The peer-to-peer consultation workload (time an orthopedic surgeon must talk directly with an insurance company’s doctor to discuss the medical reasons for a proposed surgery) has increased to several hours per week in 2025, when it was once 1-2 hours per month in 2023. These additional hours increase Duke’s valuable human capital costs while often delaying or removing revenue.
Solution
We implemented an AI solution to streamline the surgery pre-authorization process in the Department of Orthopedics. This AI tool automatically identifies and outlines the necessary documentation criteria for surgery authorizations, reducing the likelihood of missing critical elements, minimizing delays in patient care, preventing administrative errors, and mitigating revenue loss. It streamlines surgical scheduler work and frees up healthcare providers to concentrate on direct patient care.
Our first step was to identify prior authorization criteria for spine and arthroplasty plans and assessments submitted to insurers. The Duke PRMO team reviewed prior authorization denials over five years and determined the most common needs across Spine, Hip, Knee, and other joint surgery notes for generic prior authorization (from most health insurers). Today, when an orthopedic surgeon, revenue manager, or surgery-scheduler selects a signed note for AI application review, it checks the note for completion of ten criteria. The user can quickly read the AI solution’s completed checklist and related reasoning per criterion. Then, a surgical scheduler or care practitioner can improve the note before it is sent to an insurance company. The tool is not meant to make treatment decisions, but rather, evaluate how those treatment decisions were documented.
Process Results
We partnered with orthopedic surgeons, lead administrators, and PRMO analysts to create a gold standard prior authorization result for each of 40 notes with orthopedic surgery plans. We developed ten high-value authorization quality checkpoints and tested how well our AI solution performed compared to expert humans’ gold standard. In this set of 40 complex surgical assessments and convoluted notes, the average result score was 98.5%. After this retrospective step, five Duke Orthopedic surgeons reviewed 100 plans’ AI checklist completion over a period of 3 months. When scaled to 100, that is 1000 questions, the accuracy was 99.8%.
Impact
The impact of this solution is substantial. It can streamline surgery authorization and enhance technology integration, resulting in more efficient processes. This will improve patient care, reduce cancellations, and increase operational efficiency. The collaborative effort involving DIHI, PRMO, Duke Compliance Office, DHIP, and the Department of Orthopaedics promises to streamline healthcare processes, leading to a transformative impact across Duke University Health System.
