mCoder surpasses all known prior peer-reviewed results across both the 1K-constrained and full-label MDACE settings, demonstrating state-of-the-art performance in evidence-grounded autonomous medical coding.

BOSTON, May 14, 2026 /PRNewswire/ — Minset today announced that mCoder, its autonomous medical coding system, has achieved state-of-the-art performance on MDACE, a leading public benchmark for explainable and verifiable automated medical coding.

Introduced in 2023 by 3M Health Information Systems and Carnegie Mellon University researchers, MDACE (MIMIC Documents Annotated with Code Evidence) is widely recognized as a leading public benchmark for explainable automated medical coding. It includes both inpatient and professional-fee coding scenarios, with professional coder annotations linking codes to supporting evidence spans in clinical notes.

On the inpatient side, MDACE evaluates facility coding across real-world hospital admissions using discharge summaries, radiology and procedural reports, and physician notes. On the professional-fee side, MDACE evaluates physician billing scenarios across heterogeneous hospital documentation, including physician and intensivist notes, radiology and imaging reports, respiratory documentation, ECG reports, and discharge summaries.

Together, these settings make MDACE a uniquely demanding benchmark for testing whether automated coding systems can operate across complex, real-world clinical documentation and produce evidence-grounded outputs.

Minset mCoder achieved state-of-the-art results on MDACE across both major evaluation settings:

• 1K-constrained setting: where models are evaluated against the approximately 1,000 ICD codes represented in MDACE.
• Full-label setting: where models must select from the full ICD-10 code space of 70,000+ possible codes, rather than the MDACE-constrained label set, making it a more challenging and clinically realistic coding task.

Across both settings, mCoder surpassed all known prior results reported in peer-reviewed papers.

The full-label setting is the most difficult and important evaluation because it extends to the full ICD-10 coding problem used in real-world healthcare scenarios. Many systems report results on limited subsets of codes that simplify the task and do not reflect production realities. Real-world clinical coding requires operating across the full ICD domain, where scale, complexity, and variation materially increase the difficulty of the problem.

“We believe this represents an important step toward fully autonomous, production-grade coding systems that can operate reliably at scale across inpatient and professional-fee workflows, with the generality required for broader real-world coding environments,” said Matt Scott, CTO of Minset.

mCoder is part of Minset’s broader intelligent revenue cycle platform, which includes mDenials for automated denials management and prevention, and m360 for patient engagement. Together, these capabilities operate on a shared reasoning framework, enabling a closed-loop system in which coding decisions, denial outcomes, and financial workflows continuously inform and improve one another over time, replacing fragmented tools and manual processes with a unified, AI-driven approach.

With this milestone, Minset is establishing autonomous coding as a viable, enterprise-scale standard, unlocking the next phase of automation across the revenue cycle.

Implications for Revenue Cycle Operations

State-of-the-art performance on MDACE means more than achieving a higher benchmark score. It means mCoder has demonstrated superior ability to perform one of the hardest tasks in healthcare AI: assigning medical codes accurately while grounding those codes in the clinical documentation that supports them.

That distinction matters because healthcare organizations need coding systems that are not only accurate , but also:

• Transparent
• Auditable
• And Defensible

In real-world revenue cycle operations, a code without evidence is not enough. Coders, auditors, clinicians, and compliance teams need to understand why a code was assigned and where the supporting documentation appears in the chart.

Minset is currently working with select partners to evaluate mCoder in production environments. Healthcare organizations interested in evaluating mCoder can contact Minset at contact@minset.ai or visit www.minset.ai for more information.

Media Contact: Minset Communications, media@minset.ai, https://www.minset.ai 

About Minset

Minset is building the system for how healthcare revenue cycle work gets done. Its purpose-built AI autonomously performs core work across coding and denials, replacing manual processes across the revenue lifecycle.

As demonstrated in this announcement, Minset operates at real-world scale, improving accuracy, accelerating cash flow, and reducing reliance on human labor.

Through solutions like mCoder, mDenials, and m360, Minset delivers a unified platform that executes revenue cycle work and connects it directly to patient engagement and financial outcomes.

Minset was founded by leaders from Microsoft Research, Google, Optum, and Salesforce. Learn more at www.minset.ai.

View original content to download multimedia:https://www.prnewswire.com/news-releases/minset-announces-its-autonomous-medical-coding-ai-mcoder-achieves-state-of-the-art-performance-on-mdace-a-leading-public-benchmark-for-automated-medical-coding-302771746.html

SOURCE Minset AI