Another AI Side Effect: Hospitals Overbilled Blue Cross And Blue Shield By $1,000,000,000
A lot of big numbers get thrown around in the news, but one billion dollars is still a lot of money, especially when overbilling by that amount can make medical costs rise.
A recent study by the Blue Cross Blue Shield Association (BCBSA) indicates that a significant increase in the number of patients documented as having complex conditions resulted in an additional $942 million in medical care costs for Blue Cross and Blue Shield (BCBS) companies over the two-year period from 2023 to 2025.
The report “raises concerns” that the widespread use of AI coding methods is driving higher healthcare spending in the U.S., which “ultimately puts pressure on premiums and out-of-pocket expenses for families, employers and taxpayers.”
The study attempted to connect the dots. The research looked at secondary diagnoses such as anemia after major bowel surgery.
BCBSA Senior Vice President of Product and Data Science Luke Chalker explained:
“If patients are truly sicker, we’d expect to see more treatment. For example, we’re seeing significantly more anemia diagnoses at these hospitals without a corresponding increase in transfusions.”
“The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients.”
The report indicates that secondary diagnoses caused claims to fall into a higher reimbursement category, which accounted for about $650 million of the increased expenses for BCBS companies over the two-year period.
BCBSA Senior Vice President of External Affairs David Merritt said:
“We know families are frustrated by the rising cost of healthcare, and we are committed to tackling the root causes of these higher prices.”
“As families face higher healthcare costs, this research underscores the urgent need to better understand these AI tools and the role they may play in exacerbating the affordability crisis.”
There is a saying about computers that if you put garbage in, you get garbage out. That said, I believe one problem with AI is that humans have convinced themselves it is so much faster and so much farther advanced that they forget it isn’t perfect.
In some cases, that misguided belief has led to no one overseeing the results the system produces. In this case, BCBSA Senior Vice President of Product and Data Science Luke Chalker said it best:
“The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients.”
Either no one in the hospitals involved thought of checking the billing codes to confirm they were accurate, or they knew of the error and let the billing process proceed anyway.
The lesson is clear: AI systems still need monitoring, and results need close examination for inaccuracies. If this type of error can occur in the billing of medical procedures, it can occur anywhere billing codes are used, which is just about everywhere.
