A new Blue Cross study says AI documentation tools are inflating claims. For orthopedic practices, that's a reason to adopt AI now.

By Mark Maloney

On September 24, the Blue Cross Blue Shield Association put a number on something payers have been complaining about all year. Its study estimated that AI tools have raised healthcare spending for insurers by nearly $1 billion over two years, as providers billed for more complex or severe patient care. Compared with 2000, 2023 saw more intense coding, adding $ 942 million in cost, and 653 million of that came from providers billing more often for secondary conditions.  102.7 WBOW102.7 WBOW

The association blamed AI coding tools.  BCBSA said providers used AI to find secondary conditions by scanning existing patient records or through ambient scribes that listen to patient conversations and draft notes. Its argument rests on a gap between coding and care. BCBSA’s Luke Chalker told reporters that the data showed no matching change and the care delivered to supposedly more complex patients. And his words to Reuters: "If patients are truly sicker, we expect to see more treatment.” 102.7 WBOWFierce Healthcare

This study was in hospitals, and they dispute the causality. The American Hospital Association said in August that higher coding intensity can also reflect older, sicker patients and more accurate documentation of existing conditions, and it put the rise in hospital case-mix severity at about 5% from 2019 through 2024. Independent observers have pointed out the study's limits. BCBSA acknowledges it relied on claims data rather than clinical charts, and it released the study during a period of hard contract negotiations between plans and hospitals. The link to AI rests on timing, the types of codes involved, and adoption surveys, not on individual claims traced to specific software. Study Says AI Coding Made Hospital Costs $942M More Expensive | the deep dive +2

Orthopedic surgeons in both hospital and private practice should know what the study did not examine. It covered inpatient billing for major bowel surgery and maternity cases.  It did not examine orthopedics, but it will in both inpatient and outpatient settings. Such scrutiny will also extend outside the hospital to office or clinical settings.  Although relatively nascent, AI Sribes are already making an impact in helping physicians do more efficient coding and combat denials by using the correct language or coding to do that.   AI tools generated nearly $1 billion in extra costs, Blue Cross insurers say | 102.7 WBOW | The Valley's Greatest Hits | Terre Haute, IN +2

The story left out a key fact: payers adopted AI first, and they use it aggressively. BCBSA and its member plans are building detection models to flag unusual coding patterns and are targeting facilities where diagnostic complexity has grown faster than measurable acuity or treatment. On the physician side, as of July 1, Indiana law bars health plans from using an automated tool, including AI, as the only basis for downcoding a claim without reviewing the medical record. Lawmakers in California, Connecticut, Illinois, Maryland, Missouri, and Oregon introduced downcoding bills this year, part of a wave of 2026 state laws on insurer use of AI in prior authorization and claims review. In a May 2026 AMA survey, physicians reported completing an average of 40 prior authorizations a week, and 32% said their requests are often or always denied.   Blue Cross report links hospital AI to $1B rise in costs +3

An orthopedic practice feels the impact of this through high-level E/M visits, MRI and injection authorizations, and surgical pre-certs. The payer's algorithm reads every claim. If no one on the practice's side reads with the same speed and thoroughness, the practice loses.

Your best first move: use an AI Scribe.

The reason is simple.  It will impact the first day with more efficient coding, plus phrasing and language to reduce needless denials. But the best reason is that you, the physician, can control this, assuming you do not need permission or board approval.

Furthermore, the case for ambient scribes starts with physician time, and the evidence here is the strongest of the three tools. In a UCLA randomized trial published in NEJM AI, 238 outpatient physicians across 14 specialties were assigned to DAX Copilot, Nabla, or usual care, and the scribes may reduce documentation time and improve physician well-being, though the authors noted occasional clinically significant inaccuracies. A Stanford emergency department study of more than 10,000 encounters linked scribe use to a 72.6-second reduction in on-shift documentation time per encounter. AI-Based Ambient Scribes May Reduce Physician Documentation Time, Burnout - Hematology Advisor +2

Your next best move: AI-assisted EMR

Ambient notes are most useful when the EMR can do something with them: fill ortho-specific templates, pull prior imaging and therapy history into the note, and surface what a prior-auth criterion requires before the patient leaves the room. This is also where the BCBSA critique applies most directly. A tool that surfaces a documented, treated comorbidity supports accurate coding. A tool that suggests diagnoses no one is treating creates liability.

AI-assisted billing and RCM

On the revenue side, AI mostly works on defense: scrubbing claims before submission, catching gaps in authorization and documentation, and drafting appeals. Practices should treat vendor ROI claims skeptically, but appeals clearly work when practices file them. Fewer than 1% of denied claims are ever appealed. Software that turns a half-day appeal into a short review changes how often a practice can afford to appeal. Mdrevenuegroup

Compliance issues: Doing it right

I run a compliance company, Venops. We have a saying I stole from my Army days.  “You should expect what you inspect.”  Here are four points to consider.

1. Require physicians to review and sign every AI-drafted note.

2. Code only what was assessed and treated.

3. Audit the practice's own E/M level distribution before and after adoption, so any shift can be explained clinically.

4. Choose vendors that give audit trails, not "revenue uplift" projections.

BCBSA's message is a warning that payers can see coding changes.  Getting flagged is not an accusation, but it can feel that way.  The best defense is documentation and tying coding to treatment.  BCBSA's argument was that the increase in coding was not reflected in treatment. 

Having said that, some of these tools may have a cost, but I don’t think it compares with the cost of doing nothing.

Even considering any risks, it is better to be a wolf than a sheep.  Enough said today.

MM