Aligning payer-provider incentives to improve coding and ACA risk transfer payments

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Jason Siegel, Kimberley K. Hiemenz, Simon Moody | 22 December 2015

With the passage of the Patient Protection and Affordable Care Act (ACA), many commercial health plans were subject to risk adjustment for the first time in 2014. Experience for 2014 showed the impact of risk adjustment can be substantial, and health plans operating under the ACA need a thoughtful, innovative plan for managing risk scores. While providers may not realize it yet, the increased focus by health plans on coding will also have implications for providers and the incentive structures they establish with health plans.

Based on our work with health plans and providers in this space, we arrived at the following key takeaways:

  • Health plans should engage providers in their networks to code diagnoses accurately and completely.
  • The most effective initiatives occur when the incentives of health plans and providers are aligned, and they work together to coordinate coding efforts.
  • A successful risk adjustment strategy includes continuous monitoring and strong execution by both the health plan and the providers.

Coding improvement for health plans

Health plans operating in the Medicare Advantage- and Medicaid-managed care markets have faced a risk-adjusted environment for years. As a result, they tend to be well versed in the activities necessary to properly manage risk adjustment revenue:

  • Assessing new members
  • Making sure those in need of care visit a provider
  • Ensuring diagnoses are coded at the correct level of severity and specificity
  • Identifying and validating potentially missed diagnoses
  • Verifying all appropriate data is carried through the submission process
  • Properly documenting all steps along the way
  • Managing the subsequent risk adjustment and data validation audits

Commercial health plans can also achieve a positive return on investment (ROI) through disciplined coding efforts (see “Coding improvement for commercial exchange plans: Is it worth the cost?” for more information). In fact, identifying a missing diagnosis code in a commercial health plan can have a larger percentage impact on risk score than in Medicare Advantage given the lower average risk score of a commercial population.

However, in our experience, there is a common misconception regarding how Medicare Advantage and Medicaid coding efforts can properly be leveraged to inform commercial risk adjustment. For example, 80% of all Medicare beneficiaries suffer from at least one chronic condition and the marginal impact of finding a missed diagnosis under Medicare is high given the high revenue per member associated with an aged or disabled population. As a result, even relatively expensive solutions to identify missed diagnoses can still result in a positive ROI.

Heavyweight tactics like home visits, for example, which are popular in Medicare Advantage, will not necessarily yield a positive ROI for a commercial population. Commercial health plans will likely have more luck with less labor-intensive strategies that fit within existing activities, such as designing appropriately aligned provider incentive structures, integrating coding efforts into existing case management work flows, and building proper reconciliations and other analytics on top of EDGE server output files.

Similarly, models built to identify missed codes under Medicare Advantage can be fairly broad-brush, erring on the side of capturing too many suspects. Thus, models intended for a Medicare Advantage market will engage far too many members and will likely have limited impact on the health plan’s risk adjustment results.

What we found is that models designed to identify members suspected of missing diagnosis codes achieve the greatest level of specificity by doing the following:

  • Layering together knowledge from a variety of markers to identify missed codes (e.g., drug utilization, procedures, comorbidities, specialist office visit patterns, etc.).
  • Using markers with negative coefficients to rule out otherwise false positives or distinguish between cases where a drug or procedure can be indicative of any of several possible conditions.
  • Using machine learning techniques to ensure all interactions are properly accounted for without over-fitting the calibration data.
  • Leveraging data incurred prior to enrollment, such as that available through arrangements with pharmacies and pharmacy benefit managers (PBMs).
  • Customizing ROI analytics based on the state, metallic tier of the member plan, market share of the issuer, etc.

Undisciplined coding efforts tend to be expensive. They are resource intensive and, if not properly structured, can result in a disproportionate number of diagnostic deletions relative to additions. The resulting ROIs may be difficult to track or potentially negative, even when everything appears to be working properly. For example, a health plan may conclude that a diagnosis found through chart review could have been identified through simpler and less expensive efforts or automatically in the normal course of business. Or many missing diagnoses may be identified, but that does the health plan no good if the health plan cannot properly engage members and providers.

To help avoid some of these pitfalls, start by considering how the financial incentives of partners (providers, risk adjustment vendors, etc.) are aligned with the health plan’s strategy. For example, if the health plan reimburses a risk adjustment vendor for each chart review, the health plan will want to ensure the suspect list is targeted based on parameters likely to lead to a positive ROI. Tracking the resulting ROIs of risk adjustment coding improvement efforts in a transparent manner can be insightful and can help a health plan fine-tune its risk adjustment strategy.

Coding improvement for providers

It is not uncommon to incentivize and reimburse providers for improved and focused medical coding efforts in the Medicare Advantage market. However, analogous coding improvement incentives for providers in the commercial market are still rare. As commercial health plans now understand the potential impact of risk adjustment from the 2014 settlements, we are starting to see a similar emphasis and emerging financial incentives in the commercial market.

Typical shared savings/risk arrangements generally include a risk adjustment mechanism to reflect changes in the relative morbidity of the provider’s attributed population between the baseline period and the measurement period. Therefore, providers in these arrangements are inherently incentivized to ensure risk scores fully and accurately reflect the morbidity of the attributed population. Further, improved coding in the measurement period relative to the baseline period is likely to result in a higher risk score and the likelihood of greater savings/lower deficits, all other things being equal. That said, the primary purpose of the risk adjustment mechanism in these arrangements has not been to improve coding, and the impact is often buried deep within a complicated savings formula.

We are beginning to see commercial reimbursement contracts (fee-for-service and shared savings/risk contracts) with a risk adjustment component to incentivize coding improvement. The intent of these new arrangements is to reward the provider for more accurately and completely coding diagnoses for the health plan’s population. In other words, the provider has the opportunity to share in the health plan’s potentially enhanced ACA risk adjustment payment from the improved coding. These arrangements provide a direct incentive for the provider to focus on risk score coding and improvement. Thus, risk adjustment analyses and tools commonplace in the health plan arena may also be applicable and critical to providers. Specifically, risk stratification tools, suspecting models, and a disciplined coding strategy may quickly become part of a provider’s commercial ACA contracting strategy.

The use of risk adjustment as a reimbursement lever in the commercial space is still in its infancy. In addition to risk score coding improvement operational analyses, there are many strategic contracting issues that need to be considered. Some examples include:

  • Reimbursement incentives
    • Structure
    • Risk adjustment
    • Audits
    • Proper compensation for increased work flow
    • Avoiding adverse incentives
  • Coding improvement
    • Measurement method
    • Level of support from health plan to provider
    • Re-basing cost/quality targets
  • Estimating final bonus payments
    • Claims run-out/incurred but not paid claims
    • Diagnoses run-out

Further, the nature of the ACA risk adjustment program (i.e., revenue-neutral, with some health plans receiving risk adjustment transfers and other health plans paying into the program) introduces a host of complications that require thoughtful consideration:

  • The current ACA risk adjustment formula includes complexities that can create “winners and losers” beyond the health plan’s risk score relative to the market (e.g., metallic tier or age distribution).
  • The provider may improve coding for all health plans in its market (not necessarily just the health plan that is incentivizing the provider to improve coding) as part of the provider’s overall process to code patients with all pertinent diagnoses.
  • Quantifying the financial impact of coding improvement to the health plan in light of the other dynamics at play in the ACA risk adjustment transfer calculation.
  • How to revise a multiyear agreement if and when the ACA risk adjustment program is revised in the middle of an agreement.

As the ACA risk adjustment program continues to take shape and health plans and providers better understand the potential upside (and downside) of ACA risk scores in the transfer calculation, we anticipate seeing greater focus on ACA coding improvement strategies, with perhaps more alignment of incentives between health plans and providers through innovative reimbursement models. As a result, many of the considerations discussed in this paper will be pertinent to both health plans and providers.