AI That Pays: Lessons from Revenue Cycle — Nathan Wan, Ensemble Health

Key Concepts

• Revenue Cycle Management (RCM)
• Healthcare Administration
• Patients, Providers, Payers
• Denials (Prior Authorization, Clinical Denials)
• Friction (Inefficiencies in communication)
• Generative AI (GenAI)
• Electronic Medical Records (EMR)
• EIQ Platform
• Multimodal LLMs

The State of Healthcare and the Revenue Cycle

The healthcare system, particularly the revenue cycle, is facing significant challenges. 40% of hospitals operate at a negative margin, not due to clinical costs, but due to broken and manual processes within the revenue cycle. These inefficiencies lead to delays, denials, rework, and lost revenue.

Introduction to Ensemble Health Partners

Nathan, Head of AI at Ensemble Health Partners, introduces the company as an end-to-end RCM solution provider working with hundreds of hospitals and health systems in the US. With 14,000 employees, Ensemble focuses on managing the entire patient financial journey, aiming to identify and prevent inefficiencies.

Nathan's Background and Perspective

Nathan's background includes experience at Google (working on speech recognition, language modeling, and early ambient technology for reducing administrative burden on doctors) and biotech startups (diagnostics for early cancer detection from blood biomarkers and therapeutic startups identifying compounds from microbiome interactions for drug discovery). He highlights that while AI in healthcare is often associated with diagnostics, imaging, and drug discovery, the financial side of healthcare presents a significant opportunity for AI disruption.

The Financial Side of Healthcare

Healthcare administration accounts for a large proportion of the 20% of GDP attributed to the healthcare system. This includes billing, insurance-related activities, eligibility checks, registration, documentation, and medical coding. The number of healthcare administrators has increased 30-fold in the past three decades, while the number of clinicians has barely doubled, highlighting the rapid growth and complexity of this area.

Patients, Providers, and Payers Defined

• Patients: Individuals receiving care.
• Providers: Hospitals, specialty offices, specialists, nurses, and doctors delivering medical care.
• Payers: Insurance companies (private payers) and government institutions like Medicare and Medicaid.

Friction in Healthcare and the Role of Denials

A large amount of healthcare cost is related to "friction," which refers to inefficiencies in communication between payers, providers, and patients. Denials are a major component of this friction, costing providers time and money to manage and appeal. AI can potentially shift resources away from this bureaucracy towards more productive areas like clinical care.

Example of a Denied Claim

The video presents an example of a claim that was denied four times and appealed four times, requiring multiple submissions of documentation through various interfaces. The provider did not receive payment until 200 days after the procedure. Payers are also leveraging AI to increase denial rates, further complicating the process for providers.

Addressing Errors at the Source

The key point is that most denials are due to technical errors in registration or missing data, not necessarily medical disagreements. By addressing these errors upfront, much of the friction can be avoided.

Ensemble's Approach to Solving the Problem

Ensemble leverages its end-to-end RCM position to connect data points from the beginning to the end of the process, aiming to prevent errors before they occur.

Prior Authorization Example

Prior authorization, where payers require permission for certain procedures, is a challenging area. It's often unclear when prior authorization is required, and denials can still occur even after obtaining authorization. Ensemble uses AI to predict and correct denials by analyzing historical data and flagging potential issues (e.g., missing procedures on the original document). AI can also automate the manual process of acquiring prior authorization.

Clinical Denials Case Study

Clinical denials occur when payers and providers disagree about the medical necessity of care. Appealing these denials is time-consuming, requiring clinicians to review patient records (EMRs), clinical guidelines, and payer policies. GenAI can generate appeal letters, but off-the-shelf models are insufficient. Ensemble developed a model and pipeline in collaboration with clinical experts to generate high-quality appeal letters, with the clinical expert making the final decision. This has resulted in a 40% reduction in time, higher overturn rates, and increased volume. The ROI is directly measured and tracked.

Challenges and Opportunities

The healthcare industry is reliant on long-standing processes with inconsistent rules and unstructured data scattered across various systems. Ensemble's EIQ platform brings together multiple data formats within a single platform. While AI is already delivering value, automation alone is not enough. Reasoning and connectivity are crucial for addressing errors upstream.

Conclusion

Ensemble aims to build a smarter, more coordinated system to reduce waste in the revenue cycle process. The company believes it is uniquely positioned to lead this transformation due to its data set, team, and full scope of the RCM process. The speaker encourages the audience to think about AI in healthcare in new ways.