Risk Adjustment under the Affordable Care Act Issues and Expectations. Ross Winkelman, FSA, MAAA

Transcription

1 1 Risk Adjustment under the Affordable Care Act Issues and Expectations Ross Winkelman, FSA, MAAA (720)

2 2 Goals Identify key decisions that states will need to make (and when) Identify key issues that will drive states decisions Provide my thoughts on what decisions states will make Get your thoughts

3 3 Overview The Affordable Care Act (ACA) requires states to implement risk adjustment in the individual and small group marketplace Risk adjustment is the process of adjusting payment to health plans based on differences in expected costs due to differences in measured morbidity Risk adjustment can protect health plans that enroll sicker individuals Risk adjustment can help protect the state s Exchange from anti-selection

4 Risk Adjusted Payment Programs Should Be: Fair Accurate Administratively simple Transparent Budget neutral across health plans

5 Fed role: Model Provide but don t prescribe Data Standardization and Support Process / payment methods Review but allow flexibility

6 2014 is Transformative and Risky! New underwriting and rating rules Covered benefits will change (increase) Previously uninsured without prior experience High risk pool Mitigation Risk adjustment, reinsurance and risk corridors

7 What does this mean? States A lot to do need to get everything done and risk adjustment is just one piece Health Plans A lot to do need to get everything done and risk adjustment is just one piece Pricing is critical a lot of uncertainty Consumers Premium impact!

8 8 Prospective, Retrospective and Concurrent Definitions Data Time Period used can be prospective or retrospective in nature Risk weights can be developed with prospective or concurrent approach Options 1. Use 2013 to predict 2014 (Pros & Pros) 2. Use 2013 to predict 2013 and apply to 2014 (Pros & Conc) 3. Use 2014 to predict 2014 and apply to 2014 (Retro & Conc) 4. Use 2014 to predict 2015 and apply to 2014 (Retro & Pros)

9 9 Data Time Period Prospective Approach Pros of Prospective Method Familiar at Federal level because Medicare is prospective Consistent with medical management principles. Does not pay health plans for acute events, so more incentive to manage. Less uncertainty financially for health plans. Retrospective approach may result in large yearend payable or receivable.

10 10 Data Time Period Retrospective Approach Pros of Retrospective Method Key data may be missing from base period experience required to implement a prospective approach Data on previously uninsured individuals would be missing The retrospective approach is more accurate Budget neutrality more straightforward Would give health plans more time to correct data issues, since 2014 data would be used

11 Setting Risk Weights Choosing appropriate weights for each condition $30,000 Annual Cost by Year Since Diagnosis and/or Acute Event Concurrent Weight $25,000 $20,000 Prospective Weight $15,000 $10,000 $5,000 $

12 What will states do? New benefits and previously uninsured drive states decision In 2014, states likely to use retrospective approach (i.e. will use 2014 data) 2015+, some states may change to prospective Risk weight development to be determined

13 Considerations in Selecting Data Elements to be Used:

14 What will states do? Use age/gender, diagnoses (IP and OP) and National Drug Codes (NDCs) Maybe income level

15 15 Minimum Eligibility & Handling Members Without Sufficient Eligibility A widely accepted minimum eligibility span for risk adjustment models is six months For members without sufficient eligibility: 1. Assign a value equal to average claims based risk adjustment values from those with sufficient eligibility spans 2. Dampen risk scores by the portion of members without sufficient eligibility 3. Assign an age/gender score 4. Assign the age/gender score and adjust the value up or down based on the relationship between this score and claims based scores for those with sufficient eligibility

16 What will states do? Be explicit on what they assume for un-scored members Assume at least some systematic risk profile (i.e. extrapolate scored to un-scored)

17 Selection of the model Accuracy Transparency Data Licensing cost Other costs and resources Susceptibility to gaming Acceptance and familiarity in the marketplace Reliability across settings, over time, etc.

18 Risk Adjustment Tool Selection and Calibration Commercially available risk adjustment models Publicly available models 18

19 19 Risk Adjustment Tool Selection and Calibration Impact of size of population on risk adjustment model sensitivity: MAPE = Mean Absolute Prediction Error (Smaller is better, 0% is perfect)

20 What will states do? Use transparent model Most likely Fed provided model Weights specific to each state My perspective: Weights developed based on information provided by the Feds, but taking into consideration each state s specific market and their desire to balance straight prediction with public policy - does the state want to incentivize and disincentivize? If so, what and how much?

21 21 Integration with Allowable Rating Variables Under ACA, health plans are limited regarding rating variables: Gender rating is not allowed, age rating is limited to 3:1 for the oldest to youngest Adjustment for smoking is allowed, with a limit of 50% higher rate Other health status adjustments are not allowed

22 22 Integration with Allowable Rating Variables Sample Calculation Risk Adjustment Integration with Rating Variables

23 What will states do? Carefully consider interaction and overlap with allowable rating variables Use average rating variables rather than health plan specific variables

24 24 Payments to and from Health Plans Options 1. Make payments based on average overall premium 2. Make payments based on average premium by precious metals tier 3. Make payments based on actual premium being charged Region specific risk pool calculations?

25 25 Payments to and from Health Plans 0.90 Adjustment to (D) represents the portion of premium assumed to be variable (Medical costs + variable portion of admin / profit).

26 26 Integration with Reinsurance and Risk Corridors Academy paper with various stop loss options just refinement at tail of risk adjustment or more like traditional stop loss Dampens risk weights for certain conditions Only costs at risk to the health plans should be included in the risk weights Risk corridors don t impact risk adjustment but risk adjustment does impact risk corridors

27 27 Implementation Steps 1. Data call with health plans 2. Analyze population 3. Develop initial methodology and approach 4. Communicate with key stakeholders (no results!) 5. Refine methodology and approach based on stakeholder feedback 6. Perform test runs 7. Implementation and post-implementation

28 Timeline 28

29 29 What will states do? 1. Retrospective in Age/gender, diagnosis and pharmacy (maybe income) 3. Transparent model, most likely provided by HHS 4. Implementation will include health plan feedback and at least one, hopefully two test runs 5. Level of detail and auditing likely to vary by state health plans should communicate preferences and issues early

30 Questions? 30