Medicaid fraud‑detection tool generates leads; officials stress manual review and tooling limits

Ashley D'Orisio, director of fraud, waste and abuse at Colorado Medicaid, described the department’s use of an analytic SaaS platform (contracted through Allevia Analytics) to run statistical models that identify unusual billing patterns.

"The tool takes absolutely no action whatsoever," D'Orisio said. "These leads come out in a statistical mathematical sense. We then have our data analysts as well as our investigation and case managers who take a look at those leads to then determine if there's some additional action that needs to be taken." She emphasized that models are developed by human analytic staff (no generative AI or LLMs in the modeling step) and that the platform runs in Olivia's AWS environment with certifications and NIST controls.

D'Orisio said the platform runs about 150 analytic models, roughly 75 of which HCPCS staff have verified as accurate for Colorado data. She reported the tool generated 82 new leads for review and that prepayment review of four providers produced approximately $200,000 in savings over six weeks. Senator Ball asked how many of the 82 leads turned into confirmed fraud; D'Orisio said she would provide a breakdown and cautioned that 'fraud' implies criminal intent and many leads are waste or abuse matters.