Alternative Patient Conditional Count Model Confirmed for PY 2020
It’s time to limber up so you don’t get lost in the upcoming HCC model melee. On April 1st, CMS released the 2020 final call letter, confirming what models will be used moving forward and deciding the direction of MA for the immediate future.
The mixed model took shape with the Advance Notice. In part I, CMS proposed two possible models for 2020. One suggestion was the Patient Condition Count (PCC) model, which consisted of 83 HCCs. The other was the Alternative Patient Condition Count (APCC) model—for all intents and purposes the same as PCC, but with 86 HCCs. Of the two models, CMS chose APCC for the 2020 PY. The APCC model still has an additional coefficient for HCC counts between 4 and 10—this was included in the Final 2020 call letter.
The additional 3 HCCs in the APCC model are: HCC51 Dementia with Complications, HCC52 Dementia, and HCC159 Pressure Ulcer of Skin with Partial Thickness Skin Loss. Last year, in the Advance Notice Part 1, CMS noted how these three HCCs predicted cost well and fit the CMS criteria to be included in the HCC model for Medicare Advantage.
Considering the changes, now is an opportune time to study your population. Look for the additional HCCs reported in prior years to understand if they will be materially impactful in PY 2020 RAF scores. Conducting more informed submissions, with error management and reconciliation that uses blended modeling is a good place to start.
RAPS and EDS 50/50 Split, Two Models Running in Tandem
Not so much complicating, but certainly not simplifying submissions, CMS confirmed RAF scores for PY 2020 will be calculated from RAPS data, leveraging the 2017 HCC model. Meanwhile, EDS data will leverage the new APCC model, at a 50/50 split. The source of the EDS RAF score will include RAPS inpatient data.
Health plans should look at this as an opportunity to study and correct differences between RAPS inpatient data and EDS inpatient data. The continued use of the RAPS inpatient data is indicative that not all plans are reconciling RAPS and EDS inpatient at the same rate as professional data.
Reconciliation between RAPS and EDS overall should continue to be a focus for all plans. CMS indicated in the Advance notice that by PY 2022 the data source for RAF will all come from EDS and the APCC model, dropping one facet of the mixed model arts discipline, but not all.
As you incorporate the changes, take stock of submission performance and ask relevant questions you should always be able to answer. Can you identify on an ongoing basis the top five errors in your inpatient EDS data? Furthermore, can you identify the source of these errors, claims issues, provider, member etc.?
ESRD Members – two MORE models
In the Advance Notice II, CMS indicated that the current model had underpredicted cost for functioning graft ESRD (new enrollee and Long term institutional (LTI)) and over predicted for new enrollee ESRD. The reason given for the lack of precision in the model was the smaller available size of these populations, whereas most functioning graft ESRD are community members.
Subsequently, CMS proposed a new 2020 PY model for ESRD be phased in. For functioning graft, CMS reset the predictive ratio to 1 by simply dividing all coefficients for new enrollee functioning graft segments by 0.806 and all coefficients in the functioning graft LTI segment by 0.836. For new enrollees, since there was over prediction in cost due to the small sample population, CMS set the entire predictive ratio to 1.0 by dividing all coefficients by 1.149.
During the comment period there was ample pushback against leveraging the new proposed model, especially for new enrollees as CMS clarified that cost was 15% overpredicted (1.149). However, in the final notice CMS has shown they will move ahead with the new ESRD models, albeit in phases.
For 2020 PY CMS has also stated that they “will blend 50% of the risk score using the 2019 ESRD models, using diagnoses from RAPS and FFS, summed with 50% of the risk score calculated with the 2020 ESRD models, using diagnoses from encounter data, RAPS inpatient records, and FFS.”
Feel like you’ve Mastered Mixed Model Arts yet?
Lastly, while we need to blend these models to come up with our final RAF score accruals, it’s good practice to build each RAF score separately under each model before you blend. This way you can review YOY as an apples to apples comparison and review any anomalies in your data before you blend the different models. PY 2020 may require new resources and training, but in the end, it’s sure to make you a Master of MMA!