From Pilot to 5x Volume:
How Brault Is Scaling Autonomous Coding Without Compromising Quality
Brault, one of the most established RCM organizations in emergency medicine, needed autonomous coding that could maintain its quality standards across complex ED sites.
After an earlier AI deployment fell short in production, Brault deployed CombineHealth — achieving 98%+ accuracy, <12-hour turnaround, and a path to 5× autonomous coding volume.
Brault Achieved Expert-Level Accuracy While Scaling Autonomous Coding
Brault's Requirement: Autonomous Coding Without Compromising Quality
Brault supports large health systems and academic medical centers—the kind of complex care environments represented by Mayo Clinic, Cleveland Clinic, and Mass General Brigham—and processes millions of patient encounters annually.
At that scale, Brault saw autonomous coding as a way to handle growing and fluctuating chart volumes more efficiently without scaling its coding workforce at the same rate.
But emergency medicine is difficult to standardize. Brault codes across multiple ED sites, each with its own requirements spanning E/M, diagnoses, procedures, modifiers, MIPS, CDI, and site-specific payer and billing rules. When those rules change, Brault retrains its coding workforce and ensures the updates are applied consistently across thousands of encounters.
Volume also changes abruptly. EMR outages, holidays, or backlogs cause two or three times the normal number of charts to arrive together—forcing a human operation to balance staffing, turnaround time, and coding quality.
Brault had already tested autonomous coding with another established vendor. While pre-go-live results were encouraging, accuracy did not hold consistently in production. After nearly a year of auditing and remediation, Brault discontinued the solution.
The next system had to prove not simply that AI could code a chart, but that it could meet Brault’s standards consistently in live production.
CombineHealth Adapted Autonomous Coding to the Rules of Each Brault ED Site
CombineHealth deployed its autonomous coding solution around Brault’s existing coding operation. The platform reads the complete encounter, generates the required coding outputs, and applies the rules specific to each ED site.
CombineHealth’s solution reads the complete encounter and generates coding across E/M, CPT, ICD, modifiers, MIPS, CDI, and provider assignment.
Each site’s coding rules are configured directly into the platform—from payer-specific CPT exclusions and modifier sequencing to observation and non-billable scenarios. New sites or rule changes can therefore be configured centrally instead of retraining an entire coding workforce.
CombineHealth’s solution processes charts against Brault’s <12-hour turnaround requirement, including periods when incoming volume rises two or three times above normal.
For Brault, site-specific knowledge that previously had to be propagated across a human coding workforce can now be encoded into the autonomous workflow and applied consistently across encounters.
Brault Continuously Audited Production Coding Before Trusting It to Scale
Given its previous AI experience, Brault did not treat strong pre-go-live results as sufficient proof.
CombineHealth first had to clear Brault’s thresholds through batch audits. After go-live, Brault continued auditing production coding on a recurring basis, increasing audit frequency whenever results required closer review.
The bar was stringent: 96%+ accuracy across individual coding dimensions, rather than a blended overall score, with a 98% threshold for MIPS. This gave Brault an ongoing mechanism to verify that performance remained consistent as autonomous coding moved deeper into production.
Brault’s Medical Coding Accuracy Stayed Above Threshold From Pilot to Scale
Across live deployments, CombineHealth has continued to clear Brault’s production thresholds while maintaining turnaround through fluctuating volumes and shortening the path to bring additional sites live.
Accuracy across key coding dimensions
Primary ICD 99.1%, CPT 98.4%, E/M 98.2%, and MIPS 98.7%
Coding turnaround even through volume spikes
CombineHealth maintained SLA even through volume fluctuations
Go-live time for recent site deployments
The time required to bring a site live and reach Brault’s accuracy thresholds
Planned autonomous coding volume
With production performance consistently exceeding the threshold, Brault plans to increase autonomous coding volume fivefold.
See what it took for autonomous coding to earn Brault’s trust in production
The full case study goes deeper into Brault’s previous AI deployment, its production auditing methodology, performance across eight coding dimensions, and how its deployment model evolved as more ED sites came online.
