In 2013-14, the team helped a leading health insurer who wanted to use their deep internal data to develop a model to predict member’s likelihood of using a specific health care service.  This predictive model focused on the sometimes discretionary service of Orthodonture and addressed the question – Are there common characteristics of members who receive Orthodonture that we can observe in our member data or claims-based treatment history and how do these characteristics combine to help us forecast users of Orthodonture services. The project goals included:

  • Mining the insurer’s large internal databases and exploring the types of analysis that can be done to better understand services, incentives, and consumer behavior.
  • Developing a specific predictive model to understand who gets Orthodontic services and what we know about them.
  • Engaging the insurer’s member base more effectively in understanding and maximizing their benefits while developing a rationale to focus marketing and education resources for the highest ROI.

We developed a statistical forecast of the probability that the member will receive Orthodontic services based on demographic data for the insurance company’s members, treatment history, and a few external fields such as income level based on member’s zip code and employer NAICS code.  Through a regression model we were able to explain a relatively large amount of the variance in the data as well as give the insurer a tool to segment their members based on likelihood of receiving Orthodonture.  The insurer was then able to target education materials to help patients choose an Orthodontist and get the most from their insurance to these high probability patients.




Art Credit: Spaarnestad Photo, via Nationaal Archief. via Wikimedia Commons