Summary: A medical necessity review checks a requested service against coverage criteria using the clinical record. The work is finding facts in charts, and that is where automation helps. The decision to deny is a different matter: CMS guidance and California law both require a clinician to make it. This post walks through the review, the timeframes now in force, and where the line sits, and points to how MightyBot handles it.
The review, step by step
A request arrives with clinical documentation. A reviewer, usually a nurse, reads the chart and tests it against the criteria that apply: Medicare national and local coverage determinations, the plan’s own policies, or a licensed guideline set. Medicare’s statutory standard excludes services that “are not reasonable and necessary for the diagnosis or treatment of illness or injury.” Requests that meet criteria are approved. The rest go to a physician.
Most of the time goes to finding facts. Lab values, imaging results, failed conservative treatment and dates of service sit in faxed notes and EHR exports, and each criteria branch needs a specific one. Two reviewers can read the same chart and land differently.
Why errors run both ways
Slow approvals delay care; wrong denials do worse. An HHS Office of Inspector General review of Medicare Advantage denials from June 2019 estimated that 13 percent of denied prior authorization requests “met Medicare coverage rules,” and found plans asking for documentation when the case file “was already sufficient to demonstrate medical necessity.”
The timeframes now in force
The CMS Interoperability and Prior Authorization rule set decision timeframes of “72 hours for expedited requests” and “7 calendar days for standard requests,” with compliance dates in 2026, and requires that “the payer must give the provider a specific reason for the denial.” The rule adds that decisions involving AI “must still comply with applicable requirements.”
Where the line sits for AI
CMS’s February 2024 guidance to Medicare Advantage plans says “An algorithm or software tool can be used to assist MA plans in making coverage determinations,” and that the plan stays responsible for compliance. It also says “algorithms or artificial intelligence alone cannot be used as the basis to deny admission or downgrade to an observation stay; the patient’s individual circumstances must be considered.”
The 2024 Medicare Advantage rule requires that the clinician reviewing a request “must have expertise in the field of medicine that is appropriate for the item or service being requested” before the plan issues an adverse medical necessity decision. California’s SB 1120 goes further: an AI tool “shall not deny, delay, or modify health care services based, in whole or in part, on medical necessity.”
So the design that satisfies the rules is a split. Automation reads the record and tests the criteria. A clinician decides anything short of a clear approval.
How the split works in practice
On the MightyBot platform, agents read the full record, extract the clinical facts with a pointer to the page each came from, and test them against the criteria written as explicit rules. The output lists every criterion as met, unmet or missing evidence, for this patient, with the source attached.
Requests that clearly meet criteria can be approved quickly. Anything unmet, ambiguous or high cost routes to a nurse or physician with the evaluation assembled, and the clinician makes the determination. When documentation is missing, the agent drafts a request that names the exact item and the criterion it supports. Providers can run the same check before they submit, which prevents the denial and the appeal that follows.
The buyer’s checklist, including the questions to ask about who decides and how the clock is tracked, is on the medical necessity review page.