On September 27, 2026, Beijing time, the Blue Cross Blue Shield Association of the United States (BCBSA) released an analysis report stating that between 2023 and 2025, the growth of complex coding will increase medical expenditures by approximately US$942 million for member companies. The association pointed out that AI coding tools may have contributed to this change, but did not directly attribute the entire increase in costs to the application of artificial intelligence technology. This conclusion is based on its analysis of de-identified claims data and is not a single cause-and-effect relationship proven by a randomized trial. The report was explained by Luke Chalker, senior vice president of product and data science at BCBSA.
What changes does claims data cover?
The main body of this analysis is the American Blue Cross Blue Shield Association, covering claims changes from 2023 to 2025. Together, its member companies provide health care coverage to about one-third of Americans, and claims data covers every zip code across the United States. BCBSA uses de-identified data to observe changes in diagnoses, billing categories and some treatment records, but does not publish hospital-level lists or specific provider shares. The analysis was used to compare changes in coding complexity, billing, and observable treatments, and cannot prove that AI alone is responsible for rising spending.
Hospitalization reimbursement is affected by diagnosis-related grouping and case severity. When the system detects and records secondary diagnoses, cases may move into higher-reimbursed billing categories and bill amounts may change, even if observable treatment items do not increase simultaneously. BCBSA views this payment mechanism as a key context for this analysis, but fully documenting true medical conditions and inappropriately increasing charges are not the same thing. Coding changes in claims data can change payment outcomes, but category changes alone cannot determine whether a diagnosis is necessary or accurate.
How secondary diagnostics are changing payments
According to BCBSA estimates, about 70% of new medical expenditures, or more than $650 million, are related to secondary diagnoses pushing cases into higher severity billing categories. This amount comes from de-identified claims data of member companies, and the calculation caliber and explanation are provided by the association. BCBSA linked some of these changes to the popularity of AI coding tools, but did not attribute all of the new costs directly to AI, nor did it define the number as a total increase in the U.S. healthcare system.
According to the association, more than 60% of hospital systems have begun using AI coding technology that can read laboratory results, medical records and physician records. These tools can search for secondary diagnoses from large volumes of records and then assist hospitals in classifying and coding cases. Secondary diagnoses that alter case severity may push the same hospitalization into a diagnosis-related grouping with higher compensation. The official announcement does not list the specific hospitals, suppliers or market shares that use the tool, so the public information cannot compare the identification accuracy and charging differences of different systems.
Anemia Examples and Method Limitations
The BCBSA cited anemia after major intestinal surgery as an example: there was a significant increase in anemia diagnoses in claims records, but there was no corresponding increase in blood transfusions. The association believes that coding changes may be faster than changes in observable treatments, and uses this misalignment as a clue to analyze the impact of AI-assisted coding. This example comes from payer claims data, not a randomized trial or patient case audit; it can support further verification that coding and treatment match, but it cannot independently confirm that each new diagnosis is not true.
De-identified claims data can show relationships between diagnoses, charge categories and partial treatment records, but is not a substitute for a full medical record review. The BCBSA public announcement did not provide a hospital-level list, specific AI supplier shares, case sampling audit results, nor did it explain how all potential confounding factors were handled. Because those details aren't public yet, outside researchers can't currently recalculate the $942 million estimate by hospital or tool, or tell whether the same patterns appear across regions and providers.
Title attribution and issues to be verified
TechCrunch's report title on September 26 used a strong statement that AI has increased medical costs, and the BCBSA official announcement used a cautious approach that the use of AI coding tools may lead to higher expenditures. The causal strength of the two expressions is not the same: media headlines directly highlight AI and rising costs, while association materials retain observational correlations and uncertainty. This article follows the official definition and does not write all US$942 million as expenditures directly caused by AI, nor does it automatically equate new secondary diagnoses with improper charges.
BCBSA is the national association of health care payers with a clear interest in controlling reimbursement spending. No hospital associations, hospital systems or coding tool vendors have publicly responded to the new analysis, and the $942 million is not an industry-wide total confirmed by government audits or peer reviews. Highlights for follow-up verification include complete methodology, hospital and case sampling, diagnostic accuracy, actual treatment changes, and whether independent investigators can replicate the results using comparable data.