By RaeAnn Grossman

Data scientists and health informatics teams are increasingly using predictive analytics to tackle high cost of care challenges such as patient readmission, non-emergent ER and ICU utilization, and high-risk pregnancy.  CMS, HHS, and state Medicaid agencies are progressively leaning on risk adjustment to help predict the cost of care for the current and future calendar years.

 

 

The main challenges with risk adjustment predictive modeling are:

  • Purpose: People mix up risk stratification and risk adjustment.
  • Coding: It is based on selected types of physicians and their ICD-10 coding for conditions and health status.
  • Conditions: Risk adjustment modules for Medicare Advantage impede predictive cost for terminal illness or palliative care because of coverage changes.
  • Connectivity: System connectivity and field limitations in data transmitted from EHRs to clearinghouses to claims payment.

As the industry strives to create better analytics and more actionable insights, we must continue to address and interpret challenges and create weights and rules that quiet false positives, outliers, and erroneous relationships that some of the math would push us toward. The simplest answer is typically correct with fewer clicks and over interpretation.