A state Medicaid plan set a $4MM cost-of-care savings target. Trexin's machine-learning model learned the experts' member-reassignment rules at over 99% accuracy, cutting cycle time an estimated 8–12 weeks and driving more than $1MM of the goal.
A state Medicaid plan targeted $4MM in cost-of-care savings by improving primary-care-physician network performance, which meant reassigning members away from under-performing providers early in the fiscal year, so improvements had time to accumulate. The catch: the reassignment rules were undocumented and lived with a few senior experts who had very little availability.
We designed a machine-learning classification model that learned the reassignment rules from the historical record of past expert decisions, then applied them to new cases. That removed the dependency on scarce experts and compressed the reassignment cycle by an estimated 8–12 weeks.
The model hit over 99% accuracy in tests run by the experts themselves, which drove fast adoption by the teams accountable for reassignments. Deployed within two months of the fiscal-year start, it expanded the savings window from the usual 7–8 months to as much as 10, a 25–42% expansion estimated to drive more than $1MM of the $4MM target.
The right tool for the problem: a focused model that removed a human bottleneck, not an AI moonshot.
case study
case study
event
Tell us what you're trying to do. A senior practitioner will read it.
Talk to us