Predicting what an insurer should pay on a claim
DashPoint Analytics built an AI model that predicts how much an insurer should pay on a healthcare claim, with more than 95% accuracy, so underpayments are easier to spot and recover.
The problem
To recover money from underpaid claims, you first need to know what each claim should have paid. But the insurers' price lists were incomplete and inconsistent, so that number was often a guess.
What we built
A model that learned from years of past claims what each insurer typically pays, which the client's other systems can ask for a price on any claim. A second tool compares insurers' price lists to find duplicates and group related contracts. Along the way we removed an input that was quietly giving the model the answer. The score dropped a little, but it became one the client could trust.
Outcome
More than 95% accurate: on average, predictions land within about 4.5% of what was actually paid. For the hardest insurer to predict, accuracy rose from about 14% to nearly 90%.