Prior authorization, coding and payer AI
Algorithms that insurers and their contractors use to approve, limit or deny coverage, and tools that health systems use to code visits and handle prior authorization paperwork.
Category overviewHide category overview
Insurers use algorithms to speed prior authorization and utilization review. In 2024 guidance, the Centers for Medicare & Medicaid Services (CMS) said Medicare Advantage plans may use AI in coverage determinations but that an algorithm deciding coverage from a larger data set instead of the individual patient's history, physician recommendations or clinical notes would not comply with federal law, according to a legal analysis in npj Digital Medicine. CMS has also said a physician or other appropriate health care professional with expertise must review medical necessity before a plan issues a denial, the analysis says, and it notes California's Senate Bill 1120, which sets requirements for plans that use AI in utilization review.
CMS's WISeR (Wasteful and Inappropriate Service Reduction) model brings AI-assisted prior authorization and prepayment review to traditional Medicare in six states from Jan. 1, 2026, through Dec. 31, 2031; CMS says licensed clinicians will determine nonpayment recommendations and participating companies will receive a share of savings, a commentary in the American Journal of Medicine Open said. A 2026 systematic review found 16 studies of AI in real prior authorization workflows; most reported technical metrics, few showed operational gains, and patient-centered outcomes were rarely evaluated.
nH Predict
Proprietary algorithm that predicts skilled nursing facility length of stay and guides post-acute care coverage decisions for Medicare Advantage insurers.
Full reviewHide full review: nH Predict
naviHealth manages post-acute care for insurers: hospital and skilled nursing facility assessments feed nH Predict, and naviHealth produces predictions of length of stay, readmission risk and functional improvement, according to a researcher's description drawn from naviHealth's promotional materials. A naviHealth physician must review and make the final discharge determination, the same description says.
Why this grade. An independent difference-in-differences study of naviHealth's 2019 partnership with Blue Cross Blue Shield of Michigan measured its effect on care: skilled nursing facility length of stay fell 13%, with no observed change in 90-day readmissions or mortality. No randomized trial was found, so the grade is not an A.
The evidence
In a difference-in-differences study in the Journal of Health Economics, a Brown University researcher examined 284,846 hospitalizations of Michigan residents at 161 hospitals in 2019, comparing Blue Cross Blue Shield of Michigan Medicare Advantage enrollees with traditional Medicare beneficiaries before and after the insurer's naviHealth partnership began on June 1, 2019. Average skilled nursing facility length of stay fell 2.3 days, 13% from a pre-period mean of 18.4 days, and stays over 30 days fell 7.1 percentage points; 90-day readmissions and mortality did not change. Partnerships at two smaller insurers showed similar effects, the author reported.
In a class action against UnitedHealthcare, estates of patients whose post-acute coverage was terminated allege that nH Predict has a 90% error rate and that over 80% of prior authorization denials were reversed on appeal, according to a legal analysis in npj Digital Medicine. U.S. District Judge Tunheim allowed some of the claims to proceed, the analysis says. It reports that UnitedHealthcare's post-acute denial rate rose from 10.0% in 2020 to 22.7% in 2022 while it automated the process, and that a Congressional report tied higher Medicare Advantage denial rates to increased AI use.
- Used by
- naviHealth works with UnitedHealthcare, Humana, Cigna and several Blue Cross Blue Shield carriers, and Optum acquired it in 2020 in a deal valued at $2.5 billion, according to Marr.
- FDA
- No FDA clearance or authorization was found. The JAMA Summit Report on Artificial Intelligence counts business operations tools among health AI tools and says many such tools are outside FDA's regulatory oversight.
- Randomized trial
- None published
- Key study
- Journal of Health Economics, 2025, independent difference-in-differences study, 284,846 hospitalizations
- Limits
- The only outcome study is observational, its main analysis covers one insurer in one state before Optum's 2020 acquisition, and it measures naviHealth's whole managed process rather than the algorithm alone; its author had no access to the actual predictions. It could not measure functional recovery or caregiving burden. The lawsuit's error-rate figures are allegations.
- Cost
- No published price found. checked
- Specialties
- Hospital medicine
Reviews of the published evidence, not medical advice or an endorsement of any product.