44% fewer coding denials: how a US health system strengthened clinical documentation

The organization aimed to strengthen its clinical documentation integrity (CDI) program and make the most of its AI-powered tools—without compromising compliance. They wanted to improve coding accuracy across complex service lines, reduce denials and undercoding that led to missed revenue, and ease the documentation burden on physicians. The system needed an approach that could flag incomplete or unclear documentation, capture overlooked diagnoses, and adapt to specialty-specific medical language while blending seamlessly with existing workflows.

Even with AI tools in place, documentation gaps persisted:

• Missed diagnoses: NLP engines missed complex diagnoses in cardiac surgery and radiology
• Hidden procedures: Billable procedures hidden in free-text notes went unrecognized
• Inconsistent reviews: Manual reviews were inconsistent across coders and service lines
• Documentation burden: Physicians spent ~70% of their time on documentation, leading to burnout
• Financial losses: Millions were lost annually to denials, write-offs, and undercoding
• Competitive gap: Competitors using integrated CDI programs achieved 2x higher claim accuracy

Existing CDI tools couldn't adapt to specialty-specific medical language or provide physicians with clear, actionable guidance. The system needed expert-trained models that could capture nuanced clinical context while reducing documentation burden.

CleverX combined specialist expertise with AI optimization, strengthening both documentation quality and system reliability.

Expert recruitment:

• 19 clinical documentation experts including certified coders, nurses, and ex-Medicare auditors
• Average 12 years of experience in complex specialties like cardiology, neurosurgery, and critical care
• Trained in risk adjustment and compliance with CMS and HIPAA standards

Training data built by experts:

• Created specialty-specific data sets to improve NLP model accuracy
• Annotated real-world documentation patterns capturing subtle details
• Enhanced AI's ability to detect overlooked diagnoses and secondary conditions

Smarter physician queries:

• Designed 450 evidence-based, specialty-tailored query templates
• Clearer, targeted prompts improved physician engagement
• Reduced repetitive back-and-forth between physicians and coders

Human-in-the-loop validation:

• All AI outputs reviewed by expert coders for accuracy
• Monthly calibration sessions across specialties for quality assurance
• Built HIPAA-compliant validation pipeline with audit-ready documentation

Within six weeks, the health system achieved measurable improvements in both documentation quality and operational performance:

• Experts reviewed 20,000+ discharge summaries across 24 specialties
• AI models became more precise in sequencing diagnoses and flagging gaps
• Physicians received clearer, specialty-specific CDI queries, reducing time spent on documentation
• Revenue cycle teams experienced faster bill drops and fewer denials

The program transformed how the health system approached clinical documentation, creating a sustainable framework for ongoing improvement while maintaining compliance with regulatory standards.

The CDI program transformation delivered comprehensive improvements across efficiency, quality, and financial performance:

Efficiency gains:

Documentation time for physicians cut by 38%, and billing turnaround time halved, improving cash flow by $4.1M.

Quality improvements:

Coding-related denials dropped by 44%, Case Mix Index improved from 0.92 to 1.08, and AI models aligned with specialty-specific standards.

Financial impact:

The system recovered $3.8M in missed procedures and saved $1.7M through reduced write-offs.

Strategic recognition:

Earned HIMSS Davies Award for excellence in health IT implementation, with the CDI program recognized as a national benchmark for expert-driven documentation.