Government Disability Program—Data Mining

Summary:  Booz Allen Hamilton saved a U.S. Government agency $270M by applying predictive modeling to a quality review process for a government disability program.

Client Challenge:  The client believed that its disability claims process allowed for a significant portion of erroneous claims to be approved by the agency. Its quality review was not effective in identifying erroneous decisions in the system, resulting in payout to customers who were not eligible for these monies.

How Booz Allen Helped:  Booz Allen recognized that the algorithm used by the client to select cases for its quality review was no better than random sample for locating erroneous decisions. We first studied historical data and determined the characteristics of claims decisions that were likely to be erroneous.  Next, we developed predictive models and data mining algorithms that would flag claims that were probable candidates for approval errors. Then, our analytics experts used the output of the models to create a risk-based inventory prioritization scheme to ensure that review resources were focused on claims decisions that were likely to be incorrect.

Results Achieved:  By creating a prioritization scheme based on predictive models,  Booz Allen helped prevent the client from paying an estimated U.S. $270M in erroneously approved claims and doubled the ROI of the review process.