A Medicare pilot that uses AI and machine learning in prior-authorization reviews has produced long delays, disputed denials and technical failures, according to federal documents obtained by the Electronic Frontier Foundation. WISeR began in January across six states and covers roughly a dozen services, including pain injections, cervical fusion and wound treatments.
Medicare traditionally did not require this preapproval for the affected patients. The pilot promises decisions within 72 hours, yet provider feedback included cases lasting weeks or months; one request was still pending after 83 days. A surgery was reportedly delayed nearly two months, and clinicians described patients waiting in severe pain without a reliable way to reach vendors.
Implementation problems differed by contractor. Innovaccer asked for a delayed launch and temporarily approved every request to avoid a backlog while its rules system was unfinished. Zyter reportedly confused Medicare Part A and Part B data, though it says that issue is resolved. A March 30 report showed Virtix denied 3,233 of 6,096 requests, or 53 percent. Virtix says it later cut average turnaround to about 1.18 days and completed a corrective action plan in August.
The program is scheduled to run through 2031, but lawmakers have sought more records and an end to the pilot. The Government Accountability Office also found that officials did not follow the required process when establishing it, raising questions about its legal basis.