Artificial intelligence assistants could ease the transition to value-based care for primary care practices, a new report by Phyx Primary Care found.
VBC can be administratively burdensome due to its enhanced reporting requirements. Primary care practices report that the transition to VBC is often long and results in a mix of VBC and fee-for-service billing practices.
In a study of 120 physicians who had used an AI scribe for 30 days or more, providers reported a 40% reduction in clinical review time for complex patients and a 32% decrease in physician burnout. The study was conducted by Phyx Primary Care, a nonprofit innovation lab that evaluates emerging technologies and evolving payment models.
Navina’s AI assistant was selected to be used by providers for the duration of the study, and data were collected from January 2024 to January 2025. Providers reported the assistant helped them with summarizing the patient record and surfacing issues. The assistant also helped improve documentation accuracy, identify codes for reimbursement and automate chart review.
For VBC, providers have additional documentation requirements like risk adjustment, utilization management and quality reporting. Phyx found a +0.153 increase in risk adjustment factor and a +1.9 point improvement in star quality ratings when physicians use AI scribes.
The report found that the assistants resulted in better care, lower costs and improved outcomes. It also argues AI assistants “may be essential for thriving” under VBC.
“Practices often manage five or more different models—each requiring its own workflows and reporting infrastructure,” the Phyx report says. “For independent practices with limited staff, this fragmentation creates a significant burden.”
In the study, providers reported high fidelity in the AI assistant recommending diagnoses and finding gaps in care. The adoption metrics measured above 90% for frequency and ease of use and above 90% for trust in the recommendations the assistant made.
The limitations of the study include the lack of a control group that did not use an AI assistant to help complete their work. The study was also completed retrospectively and was based on provider surveys, which can introduce bias, the report authors said.