home email us! sindicaci;ón

Predictive Analytics in Population Health Management

by Scott MacStravic

Predictive analytics (PA), the science of collecting, analyzing, and applying information about customers to promote the success of firms that use the information successfully, is a rapidly growing element in marketing and sales, as well as customer service efforts.  It includes two main aims:

  1. to identify the potential worth of different customers in order to tailor the level and costs of effort to the relative value of each, and thereby improve the ROI from investments in customer acquisition and retention; and
  2. to tailor the kinds of communications and experiences delivered to customers to optimize the ROI from acquisition, “development” (usually means up- and cross-selling/marketing), and retention investments

[J. Tsai “Predictive Analytics Foresees Change in the Future” DestinationCRM.com, Oct 30, 2007]

While these applications have had a long history in marketing applications, they are only recently being used in population health management (PHM), of commercially or governmentally insured, and particularly employee populations.  The application of the concept of predictive analysis in PHM has been slow to develop, and its applications still fall far short of those in customer marketing, sales, and service, though it is moving in the direction of more sophisticated and comprehensive efforts.

Initially, PA focused primarily, if not solely, on identifying those members of a given population who had the highest level of current or past expenditures, and perhaps the greatest risk of future expenditures.  This usually meant identifying people with expensive chronic diseases, whose conditions had recently been diagnosed, involved a significant crisis or complication.  Because they had already generated high expenditures, this often meant closing the barn door well after the horse had departed.  It also added to the widespread early tendency for PHM results to be exaggerated by ignoring the tendency for high-cost individuals to “regress to the mean” in the following year, whether or not any PHM intervention had been applied to them.

Gradually, PA has grown to identify those at risk for high expenditures in future, including those at risk of contracting new diseases, both acute and chronic.  By adding those at risk, a far larger population of potential PHM participants can be included in PHM efforts, and a far broader range of expenditures can be avoided or reduced.  Where chronic disease management may target 5-30% of the population, increasing with the average age thereof, it usually engages only a minority of those targeted, meaning that only tiny minority of the total population is involved, and many long after they have generated their highest levels of costs.

By adding in not only disease risks, in employee populations, but productivity/performance impairment factors, PHM can also greatly expand the measured costs and benefits of PHM investments.  Typically, the costs and savings related to improving health and thereby reducing absenteeism, presenteeism, and performance impairment, as well as creating positive impacts on both, is from two to five times as great as are “direct” healthcare, workers compensation and disability expenses alone.  And these are usually improved far more through addressing impairment factors such as general emotional problems, lack of sleep, stress, poor nutrition and hydration, etc. than through disease risk factors, alone.

PA really adds value when it is also applied to identifying which members of a given population are not only most expensive, at risk or impaired, but also most likely to respond to PHM interventions.  Analysis of individuals’ “stages of change”, “internal vs. external locus of control”, self-efficacy, motivation, perceived personal benefit, etc. can differentiate those who are likely to make the necessary behavior changes to succeed vs. not can make PHM interventions more effective and efficient.

For example, HealthMedia, Inc. Ann Arbor, Michigan employs an online health risk assessment (HRA) that includes not only questions used to estimate productivity impairment, but to identify which individuals are the best prospects and what kinds of support will have the best impact on them.  American Healthways, Nashville, Tennessee rely on their personal nurse coaches to learn about and incorporate individual characteristics, attitudes and barriers — not merely in general, but at each coaching session – to customize their coaching accordingly, in addition to using HRA analysis upfront.

Healthcare organizations are already using simple tools to identify individual characteristics and facts that enable them to customize sickness care experiences.  St. Jude Medical Center in Fullerton, California, for exa