How AI-Assisted Hospital Coding Was Linked to $942 Million in Added Costs

A Blue Cross Blue Shield Association analysis estimated that rising inpatient coding intensity added $942 million to Blue Cross insurers’ costs in 2024 and 2025 compared with a 2023 baseline. The association connected the trend to hospitals’ use of AI-enabled documentation and coding tools, which can surface diagnoses in records or clinical conversations. The estimate reflects the association’s claims analysis; it does not establish that AI caused every added cost or that every newly coded diagnosis was inappropriate. 116

What the analysis found

Hospitals increasingly coded inpatient admissions as clinically complex. More frequent billing of secondary conditions—diagnoses in addition to the primary reason for a stay—accounted for an estimated $653 million of the $942 million increase, or roughly 70%. 16

Secondary diagnoses can affect how a hospital stay is classified for payment. AI-enabled tools may help identify conditions recorded in clinical notes, lab reports, or conversations and make them easier to include in documentation and claims. The association describes these tools as part of the context for rising coding intensity, but the claims findings alone cannot isolate the effect of AI from other changes in documentation or billing practices. 16

The bowel-surgery example

One example cited in coverage of the analysis involved patients having major bowel surgery: anemia diagnoses reportedly increased, but blood transfusions—the treatment highlighted in that comparison—did not rise in parallel.

The association used gaps between newly documented conditions and expected related care to argue that more complex billing did not necessarily mean patients were sicker or received more treatment. Such comparisons can raise questions about coding patterns, but they do not by themselves show whether an individual diagnosis was clinically valid or whether it warranted the specific treatment used as a comparison. 16

What the $942 million estimate does—and doesn’t—say

The figure is an estimate of additional costs to Blue Cross companies associated with higher coding intensity, measured against 2023. It is not a direct measure of how much AI alone increased healthcare spending. The analysis also concerns the claims examined by the association, so its conclusion should be read as evidence of a billing trend—not proof that all hospitals using these tools overcoded patients. 116

The central issue is the distinction between more diagnoses recorded on a claim and more severe illness requiring more care. The association’s findings suggest those did not always move together in its analysis. They support scrutiny of how AI-assisted documentation affects billing, while leaving room for case-by-case review of whether each code accurately reflects a patient’s condition. 16