AI Coding Tools and Rising Hospital Costs: What the BCBSA Analysis Means for Coding Compliance
Why clinical validation, documentation integrity and independent audits matter more than ever in the age of AI-assisted coding
Hinfoma Global LLC | Coding Compliance Insights | September 2026
Artificial intelligence is now part of everyday hospital coding. Computer-assisted coding engines, documentation “nudges” and automated query tools promise faster turnaround, fewer missed diagnoses and fewer denials. But a new analysis from the Blue Cross Blue Shield Association (BCBSA) raises a hard question for every hospital, coding team and compliance officer: when coding becomes more complex, is the care actually more complex too?
This article summarizes what the BCBSA analysis found, how hospitals view AI coding tools, and the practical steps organizations can take to make sure every code they report is supported by documentation and clinical evidence.
What Did the BCBSA Analysis Find?
On September 24, 2026, BCBSA released a claims analysis examining hospital inpatient billing for Blue plan members from early 2023 through the end of 2025, a period in which AI-powered coding tools spread rapidly across hospital revenue cycles. The key findings reported were:
| Finding | Reported Figure |
| Share of inpatient cases billed as medically complex | Rose from 37% (early 2023) to 40% (end of 2025) |
| Estimated additional cost to Blue plans over two years | $942 million |
| Portion attributed to secondary diagnoses | $653 million (about $11,000 per excess complex case) |
| Share of the coding intensity increase driven by secondary diagnoses moving cases to higher-severity DRGs | About 70%, across more than 55,000 cases |
| Major bowel procedures: claims in the highest complexity level | Rose from 10.2% to 22.7% (about $61 million in added cost) |
| Healthcare organizations using AI in revenue cycle workflows (industry survey) | About 63% |
The central argument of the analysis is what BCBSA called a disconnect between coding and care. Hospitals in the top quartile for complex DRG coding showed similar or lower treatment intensity than their peers, including ICU use, transfusions and reoperations. One example involved acute posthemorrhagic anemia: top-quartile hospitals reported the diagnosis more often (13.7% vs. 9.9%), yet transfused patients less often (16.9% vs. 19.3%).
BCBSA’s Luke Chalker summarized the finding this way: coding “has materially changed,” but the association found “no evidence of corresponding change in care.” BCBSA also published a separate analysis of maternity admissions pointing to a similar pattern with postpartum anemia diagnoses.
Why Do Secondary Diagnoses Have Such a Big Financial Impact?
Under the MS-DRG system used by Medicare and many commercial payers, many inpatient cases are grouped into three severity levels: without a complication or comorbidity (CC), with a CC, or with a major CC (MCC). A single secondary diagnosis that qualifies as a CC or MCC can move a case to a higher-paying DRG.
That is why conditions such as acute blood loss anemia, malnutrition, acute respiratory failure, encephalopathy and sepsis receive close attention from payers and auditors. When supported by the record, they reflect real patient severity. When they are coded without clinical support, they inflate cost and create compliance exposure.
What Is the Hospital Perspective on AI Coding Tools?
It is important to read the analysis in context. It was produced by a payer association and, as BCBSA acknowledged, it relies on claims data rather than a review of the actual medical records. Hospitals and their representatives have long argued that:
- AI tools help capture conditions that were already present and treated but previously went undocumented or uncoded.
- Patient acuity has genuinely increased in many settings since the pandemic.
- Hospitals adopted automation partly in response to rising payer denials, prior authorization burden and staffing shortages.
Both sides agree on one point: the code should reflect the care. The question each organization must answer for itself is whether its own data can prove that.
What Do the Official Coding Rules Say About Reporting Secondary Diagnoses?
AI does not change the rules. Under the ICD-10-CM Official Guidelines for Coding and Reporting (Section III, Reporting Additional Diagnoses), an additional diagnosis is reportable for inpatient care only when it affects patient care through at least one of the following:
- Clinical evaluation
- Therapeutic treatment
- Diagnostic procedures
- Extended length of hospital stay
- Increased nursing care and/or monitoring
A diagnosis that appears in the record but meets none of these criteria should not be reported, no matter how confidently a software tool suggests it. In addition, a documented diagnosis may still be challenged through clinical validation if the clinical indicators in the record do not support it. The anemia example in the BCBSA report is exactly the type of pattern a clinical validation review would examine.
Where Do AI Coding Tools Create Compliance Risk?
- Suggestion acceptance without review. Coders who accept AI-suggested codes in bulk, especially under productivity pressure, can report conditions that were not clinically significant for that stay.
- Copy-forward and templated documentation. AI may pick up diagnoses carried forward from prior notes or problem lists that were not addressed during the current admission.
- Leading or automated queries. Queries generated by software must still follow compliant, non-leading query practice.
- Model tuning toward revenue. Tools optimized to find “missed revenue” rather than accurate coding can shift coding patterns across an entire hospital, which is exactly what claims analytics detect.
- Weak audit trails. If an organization cannot show who accepted a code, on what evidence, it will struggle to defend that code in a payer or government audit.
How Can Hospitals Use AI Coding Tools Responsibly?
AI can improve coding accuracy. The goal is not to abandon it but to govern it. A practical framework includes:
- Keep a qualified human in the loop. Every AI-suggested diagnosis should be confirmed by a credentialed coder against the record and the Official Guidelines.
- Benchmark your own data. Track CC/MCC capture rates, case mix index and high-risk secondary diagnoses before and after AI implementation, and compare against peers.
- Match coding to treatment. For high-risk diagnoses such as acute blood loss anemia, malnutrition and acute respiratory failure, confirm that the record shows the monitoring, treatment or evaluation that makes the condition reportable.
- Run clinical validation reviews. Pair coding audits with clinical review so that documented diagnoses are supported by clinical indicators.
- Audit the tool, not just the coder. Review vendor logic, acceptance rates and override patterns on a regular schedule.
- Build AI into your compliance program. Include AI coding tools in risk assessments, policies, training and internal monitoring, consistent with OIG compliance program guidance.
What Should Coding Professionals Take Away from This?
For coders, CDI specialists and auditors, the message is clear: technology can suggest a code, but only professional judgment can validate it. The value of a skilled coder is shifting from finding codes to proving codes, which means reading the full record, applying the guidelines, recognizing when clinical support is missing and knowing when to query.
Payers are now using large-scale data analytics to spot coding patterns that do not match care. Organizations that can show accurate, defensible coding will be in a far stronger position than those relying on software output alone.
Frequently Asked Questions
What is coding intensity?
Coding intensity refers to how many and how severe the diagnoses reported on claims are. An increase in coding intensity is not wrong by itself; it becomes a concern when it is not supported by documentation and the care delivered.
Did the BCBSA analysis prove hospitals were upcoding?
No. The analysis identified patterns in claims data that BCBSA believes point to technology-enabled coding changes. It did not review medical records, and hospitals may have other explanations for the trends.
Is it acceptable for hospitals to use AI for coding?
Yes. AI tools are widely used and can improve accuracy and efficiency. The responsibility for accurate coding remains with the organization, so AI output must be reviewed and supported by documentation.
Which diagnoses are most at risk of audit?
Secondary diagnoses that act as CCs or MCCs, such as acute blood loss anemia, malnutrition, acute respiratory failure, encephalopathy and sepsis, are frequent targets of payer and clinical validation reviews.
How often should AI-assisted coding be audited?
Many organizations audit at go-live, again at 90 days, and then at least quarterly for high-risk DRGs and diagnoses, with more frequent reviews when data shows unusual shifts.
How Hinfoma Global Can Help
Hinfoma Global LLC provides independent HIM, medical coding and compliance services for U.S. hospitals, physician groups and health systems. Our team helps organizations adopt AI-assisted coding with confidence through:
- Independent inpatient coding and DRG validation audits
- Clinical validation reviews for high-risk secondary diagnoses
- AI coding tool performance and governance reviews
- Clinical documentation improvement (CDI) and compliant query support
- Coding compliance program assessments and staff education
Our approach is simple: every code must be supported by the provider’s documentation and the care the patient actually received. To schedule an AI coding compliance review, visit www.hinfomaglobal.com.
References
- Blue Cross Blue Shield Association. BCBSA Analysis: How AI Coding Tools Affect Healthcare Costs (September 2026). https://www.bcbs.com/about-us/association-news/bcbsa-analysis-ai-coding-tools-affects-healthcare-costs
- Blue Cross Blue Shield Association. New BCBSA Research Shows AI Billing Raises Health Care Costs. https://www.bcbs.com/about-us/association-news/new-bcbsa-research-on-ai-hospital-billing-driving-higher-health-care-costs
- Fierce Healthcare. Hospitals’ use of AI coding tools cost BCBSA plans $942M more for similar care: analysis. https://www.fiercehealthcare.com/finance/hospitals-use-ai-coding-tools-cost-bcbsa-plans-942m-more-similar-care-analysis
- CDC/NCHS. ICD-10-CM Official Guidelines for Coding and Reporting and code files. https://www.cdc.gov/nchs/icd/icd-10-cm/files.html
- MS-DRG Classifications and Software. https://www.cms.gov/medicare/payment/prospective-payment-systems/acute-inpatient-pps/ms-drg-classifications-and-software
- HHS Office of Inspector General. General Compliance Program Guidance. https://oig.hhs.gov/compliance/general-compliance-program-guidance/
Disclaimer
This article is provided for general educational and informational purposes only and does not constitute legal, reimbursement, compliance or consulting advice. Figures and statements attributed to the Blue Cross Blue Shield Association and news outlets are summarized from publicly available sources listed above; readers should consult the original sources for complete details. Code assignment must always be based on complete provider documentation and the ICD-10-CM Official Guidelines for Coding and Reporting in effect for the date of service. Hinfoma Global LLC is an independent organization and is not affiliated with, endorsed by or sponsored by the Blue Cross Blue Shield Association, any Blue Cross Blue Shield plan, CMS, CDC/NCHS or the HHS Office of Inspector General. All trademarks and registered names are the property of their respective owners.
