HFMA Annual 2024: Fathom panel on autonomous coding

At HFMA Annual in Las Vegas (June 24-27, 2024), Fathom hosted a panel with Amy Katnik, COO of ApolloMD, and Nick Rogers, VP Revenue Cycle of Amazon Health Services (One Medical), who have led their organizations through adoption of autonomous medical coding. This recap synthesizes key takeaways and actionable advice from the discussion.

Key takeaways

1. Confirm your autonomous coding vendor is signed up to SLAs that guarantee automation rates that will actually give your team leverage and drive efficiencies.
2. Make sure you select a solution that can truly automate coding versus a CAC that will merely suggest codes that ultimately need to be validated by a human coder.
3. Do not underestimate the total costs associated with clinician coding.

Summary

1. Case for change

Provider organizations have grappled with several long-standing operational challenges, prompting leaders to initiate change toward stronger automation technology such as AI coding. Here are some issues that One Medical and ApolloMD wanted to solve:

Medical coder shortage and off-shore quality concerns: A chronic labor shortage has made coding increasingly expensive while worsening coding backlogs. Organizations sometimes turn to off-shore coding to save costs, but quality and accuracy tend to suffer.
‍
Clinician administrative burdens: Clinician frustration and burnout have hovered around peak levels since the pandemic. When providers shoulder some of the burden of coding, the administrative load detracts from patient care and contributes to clinician burnout.
‍
Compliance and consistency: Achieving consistent, high-quality coding with human coders is challenging, as they must keep up with frequent regulatory updates, such as the 2023 E/M guideline changes. Differential coder experience also leads to inconsistency across providers.
‍
Turnaround times and SLAs: Fast coding is essential for timely payments, but manual coding is typically slow or delayed, resulting in worse service level agreements (SLAs) for providers.
‍
Capacity planning for growth: Many organizations forecast significant expansion, but recruiting and maintaining staff to support this growth can be difficult or costly. AI coding circumvents this obstacle by scaling coding capacity without increasing headcount.
‍

2. Project expectations

Given the case for change, what expectations should leaders set for autonomous coding projects? Clearly outlining goals with measurable outcomes is crucial to spearheading an effective AI implementation. Both ApolloMD and One Medical set explicit expectations around:

Automation rates: The percentage of encounters that AI fully codes, determining all required coding elements for direct-to-bill submission. Anchor to high automation rates upon go-live – 80% to 90%, depending on specialty – to hold the vendor accountable for delivering high ROI.
‍
Accuracy rates: Many provider organizations strive for 95% coding accuracy, but actual baselines vary. First, determine current accuracy. Then, set a goal for the technology to meet or exceed this baseline. This is often a main expectation set for pilot projects.
‍
Clinician impacts: To relieve clinician burdens associated with coding, set goals for clinician time saved, satisfaction score improvement, or other measurable outcomes. For example, One Medical expects to save about 14,000 hours annually by shifting coding from clinicians to AI.
‍
Technological scope: Be clear on technologies in or out of consideration. As ApolloMD and One Medical discussed, computer-assisted coding (CAC) is an intermediate step that still requires extensive manual review. True AI coding offers much stronger results and efficiency.
‍

3. Evaluation and selection

The novelty and fast pace of AI technology can be a misfit with existing procurement processes. Designing a clear evaluation and selection process at the outset makes managing the effort easier. Here are the factors that ApolloMD and One Medical prioritized:

Automation rates: This is the most important factor for determining performance and ultimate ROI of AI in RCM. The higher the automation rate, the greater the volume of work that the technology can handle inexpensively, saving greater costs relative to manual effort. Push vendors on their automation rate definitions, proven results, and SLAs, which may vary by specialty, to compare performance subject to high accuracy of coding results.
‍
Compatibility and integration: AI vendors must be able to integrate with EHR systems to ensure interoperability and data normalization. Choose vendors with proven experience across EHRs, as ApolloMD needed to support systems across their hospital partners.
‍
Configurability and adaptability: Coding guidelines evolve, and every organization has unique rules. Autonomous coding technology must be able to incorporate regular guideline updates and support payer-specific coding rules.
‍
Scalability: Providers often need to scale operations due to increased patient volumes, new service offerings, or added departments. Broad specialty coverage thus offers an advantage for scaling autonomous technology across the enterprise over time.
‍
Vendor support: The strength of a vendor's customer team, including qualifications and behaviors, has an outsized impact on the success of the project. An outstanding team promptly resolves issues, provides regular updates, and proactively partners on improvement opportunities. Ultimately, the vendor’s team – not just the technology – shapes the experience
‍

4. ROI and operational impact

Implementing AI coding delivers significant benefits, enhancing financial performance and operational efficiency in several areas:

Reduced costs: AI coding is less expensive than manual coding, even off-shore, and thus delivers significant savings for providers – typically 30-50% relative to the incumbent solution.
‍
Improved revenue capture: Increased coding accuracy improves E/M acuity, where appropriate, and captures procedures that coders may miss, generally contributing to higher RVUs. This is especially important when physicians are compensated based on RVUs.
‍
Reduced denials: Improved coding accuracy minimizes coding-related denials and frees up resources to manage high-value claims more effectively, further reducing the impact of denials.
‍
Enhanced consistency and continuous improvement: AI coding's consistency – applying the desired rules uniformly to encounters – and continuous improvement over time deliver steady, more predictable outcomes for providers.
‍
Faster turnaround times: Shorter turnaround times – from days or weeks to minutes – accelerate organizations' revenue cycles, resulting in timelier reimbursements, reduced A/R days, and a stronger overall financial position.
‍
Freed-up resources: With AI coding in place, providers free up budgets and resources to invest in other priorities such as clinical quality, internal auditing, and denials management.
‍

5. Clinician and patient experience

Beyond the financial benefits, AI coding helps to boost patient and provider satisfaction:

Increasing satisfaction: From a clinician's perspective, AI coding frees up valuable time that would otherwise be spent on manual coding or responding to inquiries about coding. This extra time allows clinicians to focus more on patients, enabling higher satisfaction levels.
‍
Improving documentation practices: With immediate feedback loops from AI coding, revenue cycle teams can partner with clinicians to strengthen documentation upstream and align on more standardized language or formats, ultimately leading to better data quality.
‍

Real-life impact

If you'd like to learn more about Fathom, schedule a meeting here.