In March 2016, the Centers for Medicare & Medicaid Services (CMS) concluded in published statements and at a public meeting that risk adjustment programs created under the aegis of the Affordable Care Act (ACA) are largely working according to plan. However, in a post published on the official CMS blog the day before the public meeting, CMS also acknowledged that there’s “always room for improvement.”
To this end the agency is exploring changes to the risk adjustment methodology, described in a lengthy white paper published in advance of the public meeting, which would account for plan differences such as network variability, plan efficiency, effective care coordination and disease management that are not currently included in the model.
My high level takeaways from CMS’ statements and the white paper are:
The proposed changes outlined in the white paper would go into effect starting in 2018 and CMS is asking for feedback on the following specific proposals:
Accounting for partial year enrollment
Initial feedback after 2014, the first year the ACA risk adjustment plan was instituted, indicated various concerns from the public over partial year enrollment, including reports of unexpectedly high claims costs from some issuers as well as belief that the methodology was not accounting for some enrollees with chronic conditions that weren’t diagnosed during their partial year enrollment. CMS is exploring two possible fixes: creating separate models by enrollment duration or a hybrid approach that encompasses both enrollment duration adjustment factors and separate models.
Developing a prescription drug model
The feedback after the initial plan year also recommended adding prescription drug utilization into risk scoring. CMS lays out several benefits to this approach, including the ability to use the information to augment missing diagnostic data; gaining a more complete picture of the severity of illness; access to timelier, standardized data; and a reduced financial incentive to restrict access to expensive medications. They also noted several concerns, primarily the increased vulnerability to gaming the system, i.e., prescribing medically unnecessary drugs in order to trigger additional payments.
High-cost enrollee pooling
CMS acknowledges that even with risk adjustment in place, there is still an incentive for insurers to engage in risk selection by avoiding very high cost enrollees. The agency is exploring ways to remove this incentive by shielding insurers from high-cost members through an update to the risk adjustment methodology.
Finally, the white paper lays out for discussion whether and how CMS should recalibrate the model using data from the individual and small group populations rather than a separate commercial database. Interestingly, they note that this recalibration would allow them to incorporate predictors such as socioeconomic status or other sociodemographic factors, as mentioned above, into the risk adjustment model, an approach they are planning to explore in the longer term.
Talix avidly supports CMS’ direction to improve the risk adjustment program by adding more elements to the methodology for two reasons. First, the electronic medical record and other patient information databases are fully equipped to handle the collection of more patient information, particularly into unstructured text fields. Second, adding complexity to risk adjustment to capture a more refined view of a patient’s condition does NOT need to add more time and effort to factor a patient’s risk score. Appropriately tuned software easily scales in such an environment and can complement and expedite the efforts of those professionals in charge of coding patients.
As experts in risk adjustment and risk-based reimbursement Talix is well-positioned to respond to these changes and, in fact, our semantic Health Taxonomy, which we believe to be the largest of its kind in the world and which is part of the engine that powers our Coding InSight application, already accounts for the majority of the factors being discussed. For example, we already have NDC codes and drug information mapped to conditions in the Health Taxonomy, giving us the ability to quickly and efficiently respond to the potential inclusion of prescription drug data into the model. In the longer term, our existing natural language processing (NLP) technology is optimized for identifying any socioeconomic factors that may be incorporated into the model since that information is most commonly found in unstructured data. As always, Talix is working to improve our products and capabilities to ensure that we’re prepared for whatever the brave new world of healthcare risk adjustment refinement brings.