Session 18 PD, SOA Health Research Activities. Moderator: Rebecca Owen, FSA, MAAA. Presenters: Ian G. Duncan, FSA, FCA, FCIA, FIA, MAAA

Transcription

1 Session 18 PD, SOA Health Research Activities Moderator: Rebecca Owen, FSA, MAAA Presenters: Ian G. Duncan, FSA, FCA, FCIA, FIA, MAAA Rebecca Owen, FSA, MAAA Juliet Michelle Spector, FSA, MAAA

2 Health Research at the SOA Rebecca Owen, FSA MAAA October 12, 2015

3 Research Department MNG DIR RESEARCH Dale Hall HEALTH RSCH ACTUARY LEAD MODELING RSCHR RESEARCH ACTUARY RESEARCH ACTUARY RET RSCH ACTUARY SR. EXP STUDIES ACTUARY Rebecca Owen Patrick Wiese Steve Siegel Ronora Stryker Lisa Schilling Cynthia MacDonald RESEARCH ADMIN SR RESEARCH ADMIN EXP STUDIES ACTUARY EXP STUDIES TECHNICAL ACTUARY Barbara Scott Jan Schuh Jack Luff Pete Miller EXP STUDIES ACTUARY ACAD RESEARCH ASST Patrick Nolan Erika Schulty RESEARCH ADMIN Korrel Rosenberg 2

4 SOA Research Information Research at the SOA website Glance.aspx About Us Research Opportunities Research Newsletter Completed Studies 3

5 Guidance and direction Health Section Council Health Research Council Regulatory Research Advisory Council Board of Directors Research 4

6 SOA Research Committee Functions Identifying key projects Brainstorming and evaluating ideas for projects Setting priorities, budgets and timetables Finding and selecting the research teams Providing direction to the research teams Monitoring the process of projects Reviewing the results Developing a communication strategy for the results 5

7 Research Areas

8 SOA Practice Area Research 8 volunteer committees focusing on forward looking research for the profession Finance (Investment) Steve Siegel Life Insurance Ronora Stryker & Jan Schuh Joint Risk Management Ronora Stryker & Jan Schuh Knowledge Extension Barb Scott Expanding Boundaries Dale Hall Pension Section Steve Siegel Post Retirement Steve Siegel Health Steve Siegel 7

9 Practice Research Project List In Process Massachusetts Connector Risk Adjustments Provider Payment Risk Adjustment Tools Provider Networks In Concept Stage Medicaid Margin Reserving Long Term Care Retiree Medical Costs Ideas Lots of them, but never enough! 8

10 Centers for Actuarial Excellence Commitment to strengthen the academic branch of the profession. Ronora Stryker and Jan Schuh work with the CAEs on research grants CAE Grants provide mutual benefit expands SOA research work into more theoretical topics, especially in risk management. Large grants for the schools to fund the research. 9

11 Data-Driven In-House Research (DIR) Analytical and Data Driven US Retirement Lisa Schilling/Patrick Wiese CN Retirement Doug Chandler US Healthcare Rebecca Owen 10

12 Current and Upcoming Projects Medical Cost and Utilization-Based Trend Models using commercial data such as HCCI. The impact of weather/climate on costs using Medicare, Medicaid and Commercial data in conjunction with climate scientists. Joint studies with clinical and public health based research. Studies based on the CCIIO publically available data. 11

13 Communicating Research Posted on SOA website Section News Section Update Webcasts Podcasts Blog Posts Meeting sessions (local & national) Media briefings & interviews Trade journal articles 12

14 Opportunities Generating ideas Research proposals Peer review and editing Research Committee Project oversight group Subject matter expert 13

15 Thank you Contact information Rebecca Owen Steve Siegel Dale Hall 14

16 15

17 SOA Risk Adjustment Studies IAN DUNCAN, FSA, FIA, FCIA, FCA, MAAA October 12,

18 Introductions Ian Duncan, FSA, FIA, FCIA, FCA, MAAA Professor, Actuarial Statistics, University of California at Santa Barbara. Founder and former president, Solucia Consulting (now SCIO Health Analytics). Author of several books and a number of peerreviewed studies on healthcare management and predictive modeling Board member, Society of Actuaries. Chair, Risk Adjustment Study POG (substituting for Geof Hileman who is running the NY marathon today). 2 nd Edition

19 SOA Risk Adjustment Studies The SOA s first risk adjuster study, led by Alice Rosenblatt, was published in Two studies followed: in 2002 (Bob Cummings) and 2007 study (Ross Winkelman and Syed Mehmud). 18

20 SOA Risk Adjustment Studies Over time, some differences were introduced including different stop-loss points, prior cost and predictions for members with different conditions. Over time the accuracy of the models has improved (as measured by R 2 ). Whether this is due to model improvements or coding enhancements (particularly the trend to reporting increasing numbers of diagnoses on a claim) is an open issue. 19

21 Commercially-available Risk Groupers Vendors Included in 2007 Developer Risk Grouper Data Source 3M Clinical Risk Groups (CRG) All Claims (inpatient, ambulatory and drug) IHCIS/Ingenix Impact Pro Age/Sex, ICD-9 NDC, Lab UC San Diego Chronic disability payment system (CDPS) Age/Sex, ICD -9 Medicaid Rx NDC Verisk Sightlines DCG Age/Sex, ICD -9 RxGroup Age/Sex, NDC Symmetry/Ingenix Episode Risk Groups (ERG) Pharmacy Risk Groups (PRG) Symmetry/Ingenix Episode Treatment Groups (ETG) ICD 9, NDC NDC ICD 9, NDC Johns Hopkins Adjusted Clinical Groups (ACG) Age/Sex, ICD 9 MedAI was also included in this comparison with a custom model 20

22 Expanded Field of Models and Vendors The number of vendors evaluated over the years did not change much although the number of models did increase. The 2007 study evaluated 12 models from 6 distinct vendors. The updated study will include up to 23 models from 10 distinct vendors. New entrants to the field include Milliman s MARA model, Truven Health s Cost of Care Model, Wakeley s Risk Adjustment Model, and SCIO Health Analytics Chronic Conditions Model. We will also include the commercial HHS-HCC model. Returning models include those developed by UCSD (CDPS), Verisk (DxCG), 3M, Johns Hopkins (ACG), and Optum. 21

23 Data Source: MarketScan As for the 2007 study, the data source for the evaluation will be Truven s MarketScan commercial claims database (calendar years 2012 and 2013). MarketScan offers the advantage of high-quality and high-volume data representing commercially insured lives. MarketScan represents a convenience sample of data available to Truven. 22

24 Data Source: MarketScan While the MarketScan data was considered the best available data source, its use did raise concerns for the comparison. This year, Truven has submitted its own model for comparison. A number of models are currently specified using MarketScan data. The POG has discussed these issues and concluded that the evaluation will be performed in such a way that Truven would not have an advantage. This is (in part) because of the proposed re-calibration. The study will provide raw comparisons with offered weights and comparisons using recalibrated model weights to minimize the effect of potential bias in favor of models calibrated on Truven data. 23

25 Re-Calibration Prior SOA studies included re-calibration; We originally intended to perform comparisons without re-calibration (smaller health plans use offthe-shelf models); Inclusion of Truven and models developed on Truven data could give an advantage to some vendors; CONCLUSION: This year we will do both. 24

26 Re-Calibration: Methods We explored 3 re-calibration methods: Full re-calibration (re-run the regression models); p-value Re-calibration (used in 2007) (This method adjusts the original risk score; does not create a new model); Ridge Regression 1 : minimizes a function of residuals that includes the absolute value of the coefficients. Results Full re-calibration requires model detail that is not available for all models; results of tests of p-value and ridge regression re-calibration not materially different Continue to use p-value re-calibration, as is in

27 Study Outcomes The study will compare key predictive power metrics common to earlier studies, including R 2 and MAPE. Will also compute predictive ratios by various attributes, including cost tier, age and gender group, and specific disease cohorts. 26

28 Innovations in this Study Some new analyses will be introduced, in part due to the increase in prevalence of risk adjustment in healthcare finance (and specific needs of certain arrangements). New areas of focus: Predictive ratios by additional variables, including benefit richness and geographic area. Comparisons to the HHS-HCC model and the degree to which the various models are able to prospectively identify individuals that represent an expected gain or loss in the risk-adjusted marketplace. Comparisons of the degradation in predictive power of each model in the face of imperfect input data. 27

29

30 Annual Meeting & Exhibit Juliet Spector, FSA, MAAA Provider Payment Arrangements, Provider Risk, and Their Relationship with the Cost of Health Care October 12, 2015

31 Provider Payment Arrangements, Provider Risk, and Their Relationship with the Cost of Health Care 30

32 Core Message: Assessing all of the risks of a value-based contract is critical 1. Define risks and various payment models 2. Understand the general pricing process and how to vary it and extrapolate from it. 3. Evaluate the results from key experiments in the past. What works? 31

33 What makes risk contracting risky? All payment arrangements have the potential for adverse risk as well as opportunity, depending on the circumstances. No one payment structure is the best in all circumstances. Risk is loosely defined as exposure to harm or loss. For today s discussion we will focus on four risks to help us evaluate various contracting models: Utilization Risk Performance Risk Insurance Risk Technical Risk 32

34 Payment Models Payment Models Fee-for-service Global capitation Shared savings DRG/case rates Bundled payments Reference pricing Provider excess loss reinsurance Pay-for-performance Service Delivery Models Walk in clinics Accountable care organizations Patient centered medical homes Provider excess loss reinsurance and pay-for-performance are tandem models that must be paired with one of the six other payment models 33

35 Utilization Risk Defined The impact of changes in utilization (volumes) of services on provider profitability depends on the relationship of payment changes to operating cost changes (variable costs). Volume-related payment changes differ among models. Utilization Risk Performance Risk Insurance Risk Technical Risk 34

36 Performance Risk Defined Drivers of performance risk include: inefficiency suboptimal quality high cost of care Fraud, waste, and abuse Elements of reducing performance risk can include care efficiency gains performance metrics reduction of operating costs resulting from efficient work. Will vary across payment models and within payment models, depending on payments that tie to performance metrics and payer s ability and willingness to monitor clinical inefficiencies and abuse. Utilization Risk Performance Risk Insurance Risk Technical Risk 35

37 Insurance Risk Defined This type of risk is related to the normal variation in demand for medical services over time and differences in utilization within segments of insured populations. Some examples that may cause insurance risk include: Age / gender / acuity differences Number of high-cost cases vs. average Year-to-year variation in patient demand for services Proportion that has zero or near zero claims in a year Utilization Risk Performance Risk Insurance Risk Technical Risk 36

38 Technical Risk Defined The risk of appropriately structuring technical elements of a contract to match population and circumstances. Models with low technical risk are easy to design, implement, and monitor. To mitigate insurance risk, you have to add complicated elements to your contract, which introduces technical risk. Utilization Risk Performance Risk Insurance Risk Technical Risk 37

39 Pricing process, the actuary s role Lead the pricing process Help quantify some of the risk Calculate the price for the appropriate model Project and model cash flows Help the payment reform team decide which payment model is right for its target population 38

40 Pricing process 1. Obtain data most difficult step some times 2. Identify and understand total cost of care of target population. 3. Understand how reimbursement changed over historical period 4. Collaborate with clinicians to help fill in picture. 5. Estimate care management savings ITERATIVE PROCESS 39

41 Pricing process, no feasible payment model, what next? They can opt for one of the proposed payment models that best suits their needs based on the existing target population and service delivery model. They can start with a fresh target population and/or service delivery model that can be utilized to develop a new payment model. In other cases, they may continue with the status quo and hope that this choice will continue to be appropriate for the market. 40

42 Pricing process, Quality (really process measures!) Access to Care Outcome of Care ARHQ Quality Domains Structure of Care Experience of Care Process of Care 41

43 Conclusions and Best Practices Results are decidedly mixed. Success in provider payment arrangements ultimately boils down to good risk management by the payment reform team. The organization must understand its exposure, volatility, probability, severity, time horizon, and correlation to the risk 42

44 Payment Reform Team Actuary quantify risks and financially model them. An actuary can also calculate needed capital CFO and the actuary can set a budget to maintain the return on investment (ROI) of the payment reform model. CFO can allocate resources to keep the health system within the predefined budget of the payment model. Clinician provide high-quality care to the patient in order to achieve customer satisfaction and good outcomes. However, clinicians must also choose costeffective treatments in order to keep within their service and administrative cost budgets. Coding specialists, data analysts and information technology specialists need to work together with all of the other members of the team to make sure that the clinicians, CFOs and actuaries are receiving timely and accurate information. Policymaker fix systematic issues such as shortages of primary care entrants into the workforce or the adequacy of care to the most vulnerable and remote populations. 43

45 Paper also addresses in the appendices foundational knowledge The Role of the Affordable Care Act in Payment Reform Hospital and Physician Organizations Types of Other Provider Organizations Data Tools 44

46 Core Message: Assessing all of the risks of a value-based contract is critical 1. Define risks and various payment models 2. Understand the general pricing process and how to vary it and extrapolate from it. 3. Evaluate the results from key experiments in the past. What works? 45

47 Follow up papers? 46

48 New! Provider Payment Arrangements Projects/Health/2015-provider-paymentsarrangements-risk.aspx 47

49 Thank You! 48