Introduction
In 2022, the shift to encounter data as the sole source of diagnosis information used to risk-adjust payments to Medicare Advantage (MA) plans will be fully implemented. Diagnosis codes are used by the Centers for Medicare & Medicaid Services (CMS) to determine risk scores for each MA enrollee (see prior blog on how risk scores are calculated). Historically, CMS has used diagnoses submitted into its Risk Adjustment Processing System (RAPS) by MA plans. Beginning in 2015, CMS began shifting the source of data to the diagnosis codes on encounters submitted as part of the Encounter Data System (EDS). Below, differences between encounter and RAPS data are highlighted, as well as potential impacts of this shift to MA plans and a look towards what else might be on the horizon.
Key Differences Between EDS and RAPS
The first difference is that an encounter submission includes more information from the patient encounter than the RAPS submission. The information available in an encounter submission more closely aligns with a traditional Medicare fee-for-service (FFS) claim, including all data elements from the ANSI 837 v5010 claim format. In contrast, a RAPS data submission is limited to the diagnoses, date of service, beneficiary ID, and provider type.
Another key difference is that under RAPS, the submission of diagnoses for risk adjustment was at the discretion of the MA plan. Plans were not required to submit every diagnosis from every patient encounter. The plan determined which diagnoses best reflected their enrollees’ conditions. Under the new system, in accordance with CFR 42 422.310(d)(1), MA plans are now required to submit an encounter record for every patient encounter. CMS evaluates the encounter submission and makes the determination as to which diagnoses will be submitted for risk adjustment.
To evaluate diagnoses on encounter submissions for risk adjustment, CMS has developed a filtering logic based upon procedure codes using a subset of Current Procedural Terminology (CPT) and Healthcare Common Procedure Coding System (HCPCS) codes. In essence, an encounter submission would need to be associated with one of the filtering procedure codes in order for that record to be viewed as an “acceptable source of diagnoses to be considered for risk adjustment.” CMS developed the procedure code list to help ensure that only diagnoses from eligible health care providers (e.g., procedures from technicians are excluded) and visits (e.g., face-to-face encounters between a health care provider and patient are generally included). Other pieces of information from the encounter records are also used, such as the dates of service to ensure that the service occurred during the appropriate time period to be included in the risk adjustment process. Details on how the procedure code list was developed are available here.
There is also a difference in the frequency with which the data is reported. Encounter data is reported by plans at least monthly, while RAPS data is reported at least quarterly.
Changes to How Risk Scores Are Calculated
From 2015 through 2021, risk scores have been calculated using a blend of both encounter data and RAPS. The weight of the encounters in the blended risk score has generally increased over time. For example, in 2021, the risk score blend was 75% based on encounters and 25% on RAPS. For 2022, 100% of the risk score will be based on encounter data. As part of this shift, plans will also no longer be able to supplement encounter data with RAPS information, as was allowed from 2019 through 2021 with diagnosis data from inpatient records submitted to RAPS.
The move to encounter data coincides with changes to the MA risk adjustment model which were required by the 21st Century Cures Act and are also scheduled to be fully implemented for 2022. These include changes in which diagnoses are considered for risk adjustment, as well as other changes that impact the relative value associated with diagnoses in risk score calculations (more details on the MA risk adjustment modeling changes are available in prior blogs).
Potential Impacts to MA Plans
Health plans have expressed concern regarding the shift to encounter data, principally around the quality of the data. Plans have noted that conditions on encounter data are frequently inaccurate or incomplete, which could result in artificially lower risk scores and, consequently, lower payments to the plan. One major association, America’s Health Insurance Plans (AHIP), points to studies from two research firms that risk scores calculated based on encounter data were 4 percent to 16 percent lower on average in comparison to using RAPS data.
Concerns about the quality of the encounter data and its suitability for program payment have also been raised by the federal government, including the Department of Health and Humans Services Office of Inspector General and the U.S. Government Accountability Office. Presumably, these concerns factored into the decision to phase in the use of encounter data over several years. The phase-in would allow more time for plans and CMS to ensure better quality in the encounter data.
Given the concerns raised, it will be critical for plans, as well as the federal government, to continue to monitor the accuracy and completeness of the encounter data. Thus, while plans will no longer be required to maintain two separate systems for the submission of risk adjustment data, it is not clear that their operational and oversight costs will decrease. In fact, validation costs may increase for some plans as compared to using solely RAPS data, as the volume and increased frequency of encounter data presents additional operational burdens. Also, some plans may want to continue maintaining their RAPS processes to help ensure the quality of diagnosis reporting, at least until they become comfortable with the encounter process.
In addition, because of the shift in control over which diagnoses are submitted for risk adjustment from the MA plan to CMS, the burden of identifying supporting documentation for submitted diagnoses would increase. Under RAPS, plans could identify supporting documentation aligned with the patient encounter they sourced for the submission. Under EDS, because all encounters are submitted, they lose that ability. The scope of identifying supporting documentation is now expanded to every submitted encounter and downstream diagnosis accepted for risk adjustment. This has burden implications for audit and the requirements of the “Overpayment Rule,” which requires insurers to send a refund to CMS within 60 days of becoming aware of an overpayment.
The EDS return reports provide plans with information on the disposition of the submitted encounter, including whether that submission was accepted for risk adjustment. The MA plan is, however, still left with the burden of identifying which of the accepted encounters are supported by medical record documentation. Ideally, if the plan had access to a solution reconciling support in medical record documentation with the diagnoses on the encounter, this burden could be mitigated. For example, there are services available that reconcile encounter diagnoses with all available supporting medical record documentation, as well as identifying missed revenue opportunities in diagnoses supported in medical record documentation, but unsubmitted on an encounter.
With the transition to encounters as the sole source of diagnosis data for risk adjustment, provider education is critical. In many cases, accurate diagnosis coding is not necessarily required for providers to be reimbursed for their services. Plans should ensure that providers understand that complete coding in the encounter is essential to accurate risk-adjusted payment.
Looking Forward
Plans will need to be prepared for 2022, as there are several significant changes to the modeling and data used for risk adjustment. These changes make it even more important to have systematic processes in place for validation purposes. As with many risk adjustment changes, the impacts will vary across MA plans depending on the patient mix of their enrollment and available infrastructure for gathering and validating the data.
There are also some open questions that may be addressed over time. For example, given that the filtering logic centers around procedure codes, CMS may decide to validate procedure codes at some point in the future. Additionally, at some point encounter data could be used to calibrate the CMS-HCC risk adjustment model replacing traditional Medicare FFS claims as the source data for calibration. Such a decision would likely take years to implement, as CMS would need to resolve several policy and technical issues. In the shorter term though, improvements in the use of encounters could make it possible to conduct more robust analyses on the value of care provided under the MA program. Such analyses would leverage the additional information available in encounters relative to RAPS. Plus, the more frequent updating of the encounter data could provide more real-time insights to plans and CMS on broader cost and utilization trends in the MA population.