• LinkedIn
  • Twitter
  • Facebook
  • Google Plus
  • Digg
  • RSS
  • Email
  • Print

lightbulb-predictive-modeling-solutionsToo often the conversation around predictive modeling begins and ends with the process of designing and building a model or series of models. The problem with this approach is that predictive modeling alone doesn’t save money, reduce litigation, shorten claim durations, etc. unless it is packaged as part of a solution.

The solution utilizes predictive modeling as the first of three components to identify and refine the opportunity. The second component, deployment, translates and transfers the results of the predictive model as actionable information into the hands of the third component, the intervention. The intervention then leverages the information toward action that will impact outcomes. Together, the three components form a solution that drives efficiency and increases effectiveness by targeting interventions in a timely manner.

The second miss in the conversation around predictive modeling is the separation between the accuracy of the model(s) and the effectiveness of the intervention. The intervention’s ability to leverage an opportunity is a function of the design of the intervention. Intervention design should coincide with model design, if not in time, in focus. For instance, the use of a medically based intervention suggests the need for modeling that is looking at medical management opportunities. A litigation or attorney involvement model would suggest the need for an intervention targeting the avoidance of circumstances that promote litigation or attorney involvement. Sounds logical, but too often concepts are developed in parallel without the appropriate coordination, resulting in a disconnect that will translate into a lack of impact.

Lastly, if intervention design and model design are in sync, it is imperative that an evaluation be done on the effectiveness of the intervention in addition to periodic assessments of the modeling. Every intervention should follow the traditional baseline, monitoring and outcomes assessment process of program evaluation.

  • Baseline – the starting point
  • Monitoring – consistent implementation
  • Outcomes assessment – goal attainment

Predictive modeling is a powerful tool when utilized appropriately. But modeling is not a standalone silver bullet. Focusing as much attention on the design and evaluation of the intervention as on the predictive modeling will create a solution that adds real value to your organization or partnerships.

Are your partners thinking in terms of solutions?

Keith Higdon, SVP, Decision Support Services

Tagged with →  
Share →

3 Responses to Predictive modeling alone does not a solution make…

  1. I have written several articles and blogs concerning predictive modeling. I am sorry to hear that predictive modeling is being perceived as a direct tool for reducing costs. Predictive modeling is an awesome tool for giving end users more information to show where improvements are needed and, like any tool, it should be used in context.

    Predictive modeling is also a strong tool for insurers for marketing, underwriting, pricing and other purposes.

    All of this said, employers need to be aware that their insurer is likely using predictive modeling and since it is more precise, it can change their premiums. Self-insured, self-administered employers can use their own data to enjoy many of the same benefits as insurers do.

  2. Keith Higdon says:

    Annmarie,
    I think we are on the same page. The utilization of predictive modeling to find or refine opportunities is crucial to the industry, and your comments broadening its use to marketing underwriting, pricing, etc. is absolutely correct. My comments around a “solution” is to stress the importance of getting the information that comes out of predictive modeling into the correct hands (deployment) and ensuring that the information is leveraged toward a desired outcome (intervention).

  3. Jim Galletly says:

    Excellent article Keith. Predictive models are tools, created with an defined end goal . When implemented effectively, the variable model outcomes (scores) are connected to strategic actions correlated to that score or outcome. Some outcomes may be a specific intervention or referral to an expert, others may confirm that NO action appears warranted. Further, the model should be calibrated and maintained, as post implementation there are invariably improvements and fine tuning that is needed to obtain optimal value from the investment in building and implementing the model.

    One analogy is a race car. Sitting on the track after manufacture, it has no value. Learn to drive it, then modify the car suspension, weight distribution, wind drag, etc. thru improvements and maybe you can win the race with it. Now there is an ROI!

Leave a Reply

Your email address will not be published. Required fields are marked *