≡ Menu

Integrating Predictive Analytics and BRM to Improve Health Plan Member Experience


Two gentlemen from Deloitte presented Integrating Predictive Analytics and BRM to Improve Health Plan Member Experience. 80% of healthcare costs are incurred by 20% of members and traditionally the 20% get all the focus. Analytics and data mining get applied to claims, authorization, costs as a result. Segmentation focuses on the unprofitable and unhealthy. Increasingly segmentation and analytics are being applied to managing the people who are not yet sick, though a lack of data and focus is an issue. Business rules management has also been applied in healthcare but almost only in process-centric areas like claims processing or fraud for instance. Some case management and care management is beginning as are connections between different parts of the member lifecycle.

Health insurance companies are facing some major issues. Premium revenue is dropping and less profitable products are becoming more popular. In addition there is a shift to consumerism and individual choice from company coverage. Healthcare costs, meanwhile, have become a more and more significant element of disposable income and this is beginning to force trade-offs between healthcare and other expenditure. Fewer employers are offering healthcare and more people are opting out/becoming uninsured. These changes are creating new “infomediaries” like webMD who are trying to own the information relationship between consumers and health providers, new products from traditional insurers, new competitors as retailers and financial services target healthcare with products for individuals using their more analytic and targeted marketing skills. All this means that health plans need to attract new members and retain existing ones by creating loyal members whose primary medical relationship is to the plan. This requires both predictive analytics to develop insight and business rules to push these insights into production – decision management, in other words.

Part of what is driving the more effective use of predictive analytics and rules in healthcare is the broader base of data available – claims data used to be the main source but this only applies to a small percentage of the members. Using demographic data, lifestyle data, census data and other sources of information about individuals enables much more holistic modeling and segmentation and this data has been shown to be very predictive of future health risk. Taking these new data sources, aggregating and cleaning this data and integrating it with claims data drives new segmentation for members. Rules-based decision making can use this segmentation and models for targeted outreach, incentive programs, compliance programs, personalized customer service and improved disease management. All the consumer-facing decisions throughout the member lifecycle.

An example model is one that predicts the likelihood a member will dis-enroll in the next 6-12 months. 80-100 variables get used to create a model that generates a probability score. This might use some attributes from traditional data but also things like demographics, distance to primary care, active gym memberships etc. The score might be used to group people into Low Touch (not likely to dis-enroll), Average and High Touch (likely to dis-enroll). Rules can be used to ensure minimal outreach to the Low Touch group but instead focus on quality and medical management while also focusing outreach efforts on the High Touch group.

Segmentation can be used to understand how to reach the consumer, what products and services they want, what support they need and their value. This segmentation can be used, with rules, to drive better decision making in sales and marketing, pricing, customer service, medical management – decisions throughout the lifecycle. Some of these decisions are ones familiar to plans while some are new. For instance, member rewards/loyalty decisions can be driven very effectively with these approaches and this is new area for most plans. Medical management is one they always thought about but new data sources can be used to improve the analytics and rules being used to drive these decisions.

End results:

  • Improved acquisition and engagement, retention
  • More efficient allocation of resources
  • Innovation opportunities

This session touched on many of the same issues that came up when I was working with Silverlink. Pretty classic “why use EDM” stuff.


Comments on this entry are closed.

  • Neil Raden November 7, 2008, 11:19 am

    I really question this. You can’t apply predictive modeling to individuals for health insurance. The underlying risk follows two simultaneous curves: one of high frequency and low severity, and one of extremely low frequency and high severity, and it’s the second one that chews up all the dollars. You may buy a car every few years, but you only contract a terminal disease once (hopefully). A premature infant is only premature once. You cannot model that kind of experience over one person, or even a small group.

    Health insurance is not a consumer product and insureds don’t view themselves as consumers.   I do, however, believe that your (or Deloitte’s) suggestion that predictive modeling can be used for disease management and other sorts of outreach, but these programs are often voluntary and sort of gratuitous.   You can’t reach out to someone with an unhealthy lifestyle (based on your data plus 3rd party demographic data from aggregators) and tell them, “We’re your insurance company and we’re here to help.” That is a pipe dream that starts and finishes with an unwarranted belief that technology is the solution to every problem. People didn’t get that way overnight and it will take more than a decision management solution to change anything.

    -NR   twitter nraden