Clinical coding—the translation of clinical documentation into coded data—drives hospital funding. Under Activity Based Funding (ABF), what gets coded determines what gets paid. Accurate coding matters.

AI is increasingly being applied to clinical coding. The applications range from coding assistance to automated coding to revenue integrity. Here’s what’s actually working and what to consider.

The Coding Challenge

Clinical coding is complex work:

• Coders must understand clinical documentation
• They must translate that documentation into ICD-10-AM codes and Australian Refined Diagnosis Related Groups (AR-DRGs)
• Accuracy affects funding, reporting, and patient classification
• Volume is high and timelines are tight
• Workforce supply is constrained—there aren’t enough qualified coders

These factors create significant AI opportunity.

AI Coding Applications

Several AI applications address clinical coding:

Natural Language Processing for Code Suggestion

AI that reads clinical documentation and suggests appropriate codes:

• Diagnoses suggested from discharge summaries
• Procedures identified from operation reports
• Conditions extracted from various clinical notes

The coder reviews and confirms rather than identifying codes from scratch. This can significantly speed up coding while maintaining human oversight.

Automated Pre-coding

AI that automatically assigns preliminary codes before human review:

• Routine cases coded automatically
• Complex cases flagged for detailed human review
• Coding exceptions identified for attention

More aggressive than suggestion, this approaches automated coding with human verification.

Code Audit and Quality Assurance

AI that reviews coding for:

• Errors and inconsistencies
• Missed opportunities (conditions documented but not coded)
• Over-coding risks (codes not supported by documentation)
• DRG optimisation opportunities

Quality assurance rather than primary coding.

Clinical Documentation Improvement (CDI)

AI that identifies where clinical documentation doesn’t support optimal coding:

• Missing severity indicators
• Unclear principal diagnosis
• Incomplete procedure documentation
• Queries that should go to clinicians

CDI AI works upstream of coding to improve documentation quality.

Revenue Integrity Analytics

AI that analyses coding patterns to identify:

• Under-coding compared to expected patterns
• Variation across coders or facilities
• Potential revenue leakage
• Audit risks from over-coding

Strategic analytics rather than operational coding.

What’s Working in Australian Hospitals

In Australian contexts, I observe:

NLP-based code suggestion is increasingly deployed. Several vendors offer products that integrate with coding workflows and suggest codes based on documentation. Coders generally find these helpful when accuracy is high.

Pre-coding for routine cases is emerging in some high-volume settings. Works best for straightforward admissions where documentation follows predictable patterns.

CDI applications are growing. Particularly in larger facilities trying to capture appropriate case complexity. Often paired with clinical documentation specialists who follow up on AI-identified issues.

Revenue integrity analytics are used by finance and revenue teams. Less about individual case coding, more about pattern analysis and strategic decision-making.

Fully automated coding without human review remains rare. Most organisations maintain human verification given the funding, audit, and accuracy implications.

Implementation Considerations

For organisations considering coding AI:

Accuracy Is Non-Negotiable

Coding errors have consequences:

• Incorrect funding (potentially both under and over-payment)
• Audit risks and potential recoveries
• Inaccurate data for planning and research
• Professional and regulatory implications

AI that introduces errors is worse than no AI. Rigorous accuracy validation is essential before deployment.

Integration Matters

Coding AI needs to integrate with:

• Patient administration systems (PAS)
• Clinical documentation sources
• Coding workflow systems
• DRG groupers and validation tools

Poor integration creates workflow friction that undermines productivity gains.

Coder Acceptance

Coders are skilled professionals. AI that seems to threaten their role or undermine their expertise will face resistance.

Position AI as a tool that enhances coder productivity and handles routine work, freeing coders for complex cases requiring judgment. Involve coders in implementation design.

Compliance Considerations

Coding is subject to:

• Australian Coding Standards
• State/territory funding rules
• Audit requirements
• Professional standards

AI must produce compliant outputs. Vendor claims about automation need verification against compliance requirements. AI consultants Sydney recommend including compliance verification as an explicit step in any coding AI evaluation.

Vendor Due Diligence

The coding AI vendor market is evolving. Due diligence should include:

• Australian coding standards compliance (not just international standards)
• Integration capability with your systems
• Reference sites in similar Australian contexts
• Support and maintenance arrangements
• Roadmap and vendor viability

Measure Outcomes

Track whether AI delivers expected benefits:

• Coding productivity (time per episode)
• Accuracy (error rates before and after)
• Revenue impact (average case weight changes)
• Coder satisfaction
• Audit outcomes

If benefits don’t materialise, know quickly.

Revenue Integrity Perspective

Beyond operational coding, AI enables revenue integrity functions:

Identifying missed revenue. AI can identify cases where conditions were documented but not coded, or where coding didn’t capture appropriate complexity. This is legitimate revenue recovery, not gaming.

Detecting over-coding risk. AI can also identify where coding may exceed documentation support. Catching this before audit is valuable.

Benchmarking performance. Comparing coding patterns against expected benchmarks, peer facilities, or historical trends. Understanding where variation exists and why.

Prioritising CDI efforts. Identifying where documentation improvement would have greatest revenue impact. Focusing CDI resources on high-value opportunities.

These applications work at analytical level, informing strategy rather than operational coding decisions.

Workforce Implications

AI in coding affects the coding workforce:

• Routine coding may become more automated
• Complex cases continue requiring skilled human judgment
• New roles emerge around AI oversight and quality assurance
• Productivity expectations may change

For coding managers, workforce planning should account for these shifts. For coders, developing skills for complex cases and AI oversight matters for career development.

Practical Path Forward

My recommendations for organisations pursuing coding AI:

Start with code suggestion tools. Relatively mature technology, clear productivity benefits, maintains human oversight. Lower risk entry point.

Build CDI capability. AI-supported CDI can improve both documentation quality and coding accuracy. Often produces measurable revenue improvement.

Develop revenue integrity analytics. Use AI to understand coding patterns and identify opportunities. Strategic value even if operational coding isn’t AI-enhanced.

Approach automation cautiously. Fully automated coding raises accuracy and compliance questions. Most organisations should maintain human verification.

Involve coding professionals. Coders should shape implementation, not just receive it. Their expertise improves outcomes and their buy-in enables adoption.

For organisations navigating coding AI options, external guidance can help. AI consultants Brisbane and revenue integrity specialists can assist with vendor evaluation and implementation planning, though coding-specific expertise matters.

The Bigger Picture

Clinical coding AI is really about capturing appropriate funding for care delivered. When done right, it:

• Ensures hospitals are funded for the work they do
• Improves data accuracy for planning and research
• Reduces burden on skilled coding professionals
• Strengthens revenue integrity and compliance

The goal isn’t to game funding or inflate revenue—it’s accurate translation of clinical care into coded data. AI can help achieve that goal when implemented thoughtfully.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.