AI is one of five potential drivers of health costs climbing up to 9% in 2027—matching this year’s rate, the highest since 2010–11—per PwC. The key reason: AI note-taking tools are documenting more specifics about diagnoses and medical complications that a rushed human clinician might have lumped into one broad “code”—a standardized billing label that tells insurers what to pay. Those extra details can justify a higher severity (read: higher paying) code, even if the actual care a patient receives is the same as before.
Read more Roche and Nestlé fear talent crunch as Switzerland puts population cap to the polls
One Blue Cross Blue Shield analysis found that some hospitals saw the billing code for acute posthemorrhagic anemia in new mothers jump from 4% to 12.3% of maternity admissions between 2022 and 2025. The number of blood transfusions (a common treatment for this condition), meanwhile, barely budged. An audit of the hospital system with the steepest rise in this code found that fewer than 20% of the cases actually met the clinical criteria for a diagnosis. The rise in higher-intensity coding coincides with hospitals’ growing use of AI for billing. According to BCBS, “coding intensity” added $22 million to maternity spending at the hospitals studied in three years.
AI is the report’s top-ranked new pressure, but it’s not the biggest driver of costs overall—old standbys like labor and supply costs still account for more of the increase, one of the report’s authors told Healthcare Dive. And AI tools could eventually push the other way, driving down costs by automating hospital administrative work or catching diagnoses earlier.
AI is often pitched as a way to optimize whatever industry it touches—trimming waste and making systems faster and cheaper. But in healthcare, one of the first things it has optimized is how to charge you more. As one health insurance exec put it: Companies “will take AI and say, ‘How can I use this to further my self-interest?’” —WK
This report was originally published by Tech Brew.
About the Author
