Washington State Develops and Tests New Predictor of Patient's Health Needs
In 1997, the Washington Health Care Authority (HCA) the state agency responsible for administering the benefit program for nearly 300,000 Washington public employees, retirees and dependents, and the largest health care purchaser in the state implemented a health-status-based risk-adjusted payment system.
Researchers at the University of Washington School of Public Health developed the system and developed and tested its reimbursement calculations.
Reimbursement calculations take account of the health status (which focuses on the patient's health as a predictor of their yearly health care resource use) of insured individuals under this system.
This project was part of the Robert Wood Johnson Foundation (RWJF) national program Changes in Health Care Financing and Organization (HCFO) (for more information see Grant Results).
- After many discussions, issue papers and simulations, HCA implemented health-status-based risk adjustment in 1997 for enrollment beginning January 1, 1998.
Key to the implementation was finding a balance between the science of developing the payment system and the art of implementing such a system within HCA's existing political environment and policy structure.
- The payment system was phased in with full health-status-based risk adjustment scheduled for the 2000 contract year. The payment system was derived from statistical risk-assessment models based on patients' health conditions that were created during Phase I.
RWJF supported the project with two grants totaling $1,675,845 between November 1993 and June 1998.
Movement toward managed care delivery continued through the mid-1990s. In this environment, the ability of purchasers to pay managed care organizations in a manner that rewarded efficient risk management (how well a plan managed its risks) rather than effective risk selection (how well it selected patients who were good risks) was critical. In an unregulated insurance market, the payment of a single capitation rate for all consumers would be likely to result in effective risk selection, and thus the denial of coverage to sicker, more costly patients. Where the conditions of coverage are constrained by regulation (e.g., guaranteed issue and renewal, mandatory open enrollment periods, elimination of pre-existing conditions limitations), uniform payment rates would be likely to force carriers out of the market. Several strategies had been proposed to develop a system promoting efficient risk management, but they typically used only the demographic characteristics of individuals (such as age and sex) rather than health status (which focuses on the patient's health as a predictor of their yearly health care resource use) or had been applied to individual or small group markets, or special populations such as Medicare.
During Phase I of this project funded by RWJF (grant ID# 023111), the research team developed a statistical risk-payment system for health plans that rewarded efficient risk management rather than effective risk selection in other words, a system in which health plan payments were based on the average health risk of enrollees. The project team developed risk-assessment models that incorporated diagnostic and demographic variables using five years of data from six managed care plans that provided care to 300,000 Washington State public employees, retirees, and their families.
According to the project director, their efforts to develop a health-status-based risk-adjusted payment system focused exclusively on the active employee/non-Medicare retiree population about 240,000 individuals. Since 1988, Washington State had used a demographic risk-adjustment model based on the age and sex of a health plan's enrollees to adjust premiums paid to health plans.
In 1996, the Health Care Authority (HCA) adopted a mathematical process for moving money among the health plans covering active employees to ensure that HCA's aggregate payments to health plans would not increase using a new method of risk adjustment. With input from actuaries from William Mercer, Inc., an international business consulting firm, the project team evaluated the models. Subsequently, the project team recommended, and HCA accepted, a two-part, prospective statistical model that used demographic and diagnostic measures for risk adjustment.
The health-status measures used by the model grouped claims and information on patient visits to physicians and hospitals by diagnosis and cost. The team had evaluated other approaches to grouping the data that looked at ambulatory care information (e.g., visits to physicians and clinics on non-acute conditions) and information for chronic conditions (such as diabetes or asthma).
The overall objective for Phase II of the project (grant ID# 023352) was to implement the new risk-adjusted payment system using the statistical model developed under Phase I. The six specific goals for Phase II were to:
- Examine the stability and precision of the new risk assessment model.
- Develop options to address key policy issues that arise.
- Generate a fair and feasible health-status-based risk-adjusted payment system.
- Simulate health-status-based risk-adjusted plan payments and evaluate the results.
- Communicate with participating health plans.
- Implement health-status-based risk-adjusted payment.
The project team met all six goals. They tested four risk-adjustment approaches through "dry runs" using 1995 claim data from 17 of the 20 plans that contracted with HCA that year:
- HCA's original demographic model.
- An enhanced demographic model.
- The health status model based on demographic and health factors that predicted the yearly resource use of each individual enrolled in the health plan under the state's program and, by extension, for each participating health plan's enrolled population.
- A phase-in of the health status model.
The simulation demonstrated that implementation of the health status-based risk-adjusted model was feasible and that a phase-in approach would work best.
HCA's existing reimbursement infrastructure and the potential impact of the new payment system on various stakeholders (e.g., the health plans and various government entities) was managed through a process involving discussion and proactive identification of concerns. The project team prepared 32 issue papers documenting major policy and technical issues that the team and HCA administration had identified and resolved. Examples of the issues dealt with in the papers included: costs of implementation; dealing with impartial data on enrollees; potential misuse of the new system by health plans to achieve financial gain; and periodic updating of the model.
The project team also created a Technical Advisory Committee comprised of representatives from participating health plans, state government, and national risk adjustment experts. (See the Appendix for the list of members.) The committee met regularly throughout Phase I and II of the project. In addition, all potential contract bidders (health plans) were invited to two general meetings during which the risk-adjustment system was explained. HCA also met with individual health plans that contributed data to the dry run simulation to explain how their results compared with the results of other plans.
This full range of communication was an integral part of the project according to the project team. HCA phased in the new system over two years by calculating payments using both a pure demographic model and the new demographic/diagnostic model for risk adjustment. According to the final report to RWJF, "the desire to phase-in the full health-status model arose because the HCA expected that available health-plan data had sufficient limitations to inappropriately affect the results for some members of the risk pool.
The implementation team designed a method that allowed capitation rates (per-member per-month reimbursement levels) to increase or decrease in the direction that full health-status adjustment would dictate, but not to the full extent. According to the project team, the phase-in method represented a stepping stone toward the full health status-based risk-adjusted payment system, which would "likely pay some plans much more and others much less than they had received under age-sex adjustment alone."
Some 20 health plans bid on contracts for the 1998 contract year and 17 contracts were awarded by December 1997. Full health status-based risk adjustment was scheduled for the 2000 contract year. The payment system was institutionalized and HCA fully expects to continue to use the methodology as part of its operations for Washington State employees.
The project team published two papers in Inquiry and The Journal of Health Economics, both peer-reviewed journals. Several other manuscripts were nearing completion at the close of the grant. A project Web site was developed: http://froya.boat.washington.edu/risk-adjust/html/reports.html. Presentations were made at several conferences, and the project team sponsored a conference attended by representatives from 14 states and the District of Columbia to disseminate information to other states interested in implementing similar systems. The project team also developed 32 issue papers during the course of the project that examined potential obstacles and areas of concern in implementing a new reimbursement system.
AFTER THE GRANT
The HCA will continue to use the health status-based risk-adjustment methodology developed by this project for adjusting capitation payments for Washington State employees. The project team continued to work on refinements of its risk-assessment model, and to test the utility of different health status grouping categories for risk adjustment.
GRANT DETAILS & CONTACT INFORMATION
Implementing a Model for Distributing Risk Among Competing Health Plans
University of Washington School of Public Health and Community Medicine (Seattle, WA)
- Model for Medical Risk Distribution Among Competing Health Plans
Amount: $ 646,229
Dates: November 1993 to July 1995
- Implementing a Model for Distributing Risk Among Competing Health Plans
Amount: $ 1,029,616
Dates: August 1995 to June 1998
Carolyn W . Madden , Ph.D.
Technical Advisory Committee
Arlene Ash, Ph.D.
Associate Research Professor
Department of General Internal Medicine School of Medicine
John Bertko, F.S.A., M.A.A.A.
Principal, Human Resource Advisory Group
Coopers & Lybrand
San Francisco, Calif.
Director of Sales
QualMed Washington Health Plan
Milliman & Robertson
John B. Coombs, M.D.
Associate Vice President, Medical Affairs, and Associate Dean, School of Medicine
University of Washington
Paul Fishman, Ph.D.
Center for Health Studies
Group Health Cooperative
Jinnet Fowles, Ph.D., Vice President
Research and Development
Health Research Center
Park Nicollet Medical Foundation
Paul Ginsburg, Ph.D., President
Center for Studying Health System Change
Chief Financial Officer
Denise L. Honzel
Vice President and Health Plan Manager
Kaiser Foundation Health Plan of the NW
Director, Health Systems Studies
Park Nicollet Medical Foundation
James Lubitz, M.P.H., Deputy Director
Office of Research and Demonstrations
Health Care Financing Administration
Hal Luft, Ph.D.
Professor of Health Economics
Institute for Health Policy Studies
University of California, San Francisco
San Francisco, Calif.
Willard G. Manning, Ph.D., Professor
Institute for Health Services Research
University of Minnesota
Robert P. Power
Senior Director/Senior Medical Economist
Health Partners, Inc.
Douglas W. Roblin, Ph.D.
Kaiser Foundation Health Plan
Department of Medical and Economic Statistics
Dan Rubin, Director of Special Projects
Washington State Department of Health
Chief Financial Officer
Good Health Plan of Washington
DrPH Professor and Deputy Director
Health Services Research and Development Center
Johns Hopkins University
(Current as of date of this report; as provided by grantee organization; not verified by RWJF; items not available from RWJF.)
Books and Reports
Medical Risk Distribution Among Competing Health Plans: Options for Modeling Risk Assessment. Seattle: University of Washington, 1995.
Issue Paper 1.06. Use of Inpatient and/or Ambulatory Data. Seattle: University of Washington.
Issue Paper 1.08. Why Not Surveys? Seattle: University of Washington.
Issue Paper 1.09. Pharmacy (CDS) vs. Diagnosis DCG, ACG) for Assessment? Seattle: University of Washington.
Issue Paper 2.15. ACGs vs. DCGs. Seattle: University of Washington.
Issue Paper 2.16. ACG/DCG Version. Seattle: University of Washington.
Issue Paper 2.03 Confidentiality/Data Security. Seattle: University of Washington.
Decision Memo 3.04 Population. Seattle: University of Washington.
Issue Paper 2.01 Partial Data: Incomplete or Non-available Prior Use Data. Seattle: University of Washington.
Issue Paper 2.02 Partial Data: Missing Data. Seattle: University of Washington.
Issue Paper 2.04 Births. Seattle: University of Washington.
Issue Paper 2.05. Wellness Incentives. Seattle: University of Washington.
Issue Paper 2.06 Costs of Implementation. Seattle: University of Washington.
Issue Paper 2.07 Non Standard Plan Design. Seattle: University of Washington.
Issue Paper 2.08. Audits. Seattle: University of Washington.
Issue Paper 2.09. Gaming. Seattle: University of Washington.
Issue Paper 2.10. Phase In Approaches. Seattle: University of Washington.
Issue Paper 2.11 Subrogation. Seattle: University of Washington.
Issue Paper 2.12 C.O.B. Seattle: University of Washington.
Issue Paper 2.13 Twins. Seattle: University of Washington.
Issue Paper 2.14 Geographic Variation. Seattle: University of Washington.
Decision Memo 3.02. Year of Rate Assignment. Seattle: University of Washington.
Decision Memo 3.03 Outlier Stays: Mental Health & Chemical Dependency. Seattle: University of Washington.
Issue Paper 1.03 Outlier Hospital Payments. Seattle: University of Washington.
Issue Paper 1.04 Hospital Conversion Factor. Seattle: University of Washington.
Issue Paper 1.05 Estimating Annual vs. Monthly Expenses. Seattle: University of Washington.
Issue Paper 1.07 Prices for Dependent Variable. Seattle: University of Washington.
Issue Paper 2.12. C.O.B. Seattle: University of Washington.
Issue Paper 2.17 Data Lag. Seattle: University of Washington.
Issue Paper 1.10. Model Maintenance. Seattle: University of Washington.
Decision Memo 3.01 and Phase I Final Report Data Set Construction. Seattle: University of Washington.
Wilson VM, Smith CA, Hamilton JM, Madden CW, Skillman SM, Mackay B, Matthisen JS and Frazzini DA. "Case Study: The Washington State Health Care Authority." Inquiry, 35(2): 178192, 1998.
Blough DK, Madden CW and Hornbrook MC. "Modeling Risk Using Generalized Linear Models." Journal of Health Economics, 18(2): 153171, 1999. Abstract available online.
"Risk Adjustment for Capitation Payment: Lessons for Purchasers," November 2021, 1997, Seattle. Attended by nearly 100 public and private purchasers from 14 states and the District of Columbia. Four presentations and three panels:
- Gary Christenson, Washington Health Care Authority, "Why We did It: Washington Health Care Authority's Purchasing Vision."
- Ann Robinson, Buyers Health Care Action Group, Minnesota, "Purchasing Technologies for the Evolving Health Care Market."
- James Lubitz, Division of Health Systems Research, Health Care Financing Administration, "Medicare Moves toward Risk Adjustment."
- Richard Dixon, National IPA Coalition, "A longer Term Look at Risk Adjustment Will it Change Managed Care?"
- "Risk Adjusted Capitation for Washington State Employees." Vicki Wilson, Health Care Authority; Cindy Madden, University of Washington; and James Matthiesen, William M. Mercer, Inc.
- "Washington Stakeholders' Perspectives on Risk Adjustment." James Matthiesen, William M. Mercer (facilitator).
- "State Updates: California, Colorado, Oregon, Illinois, Kentucky, New York, Massachusetts, Missouri, and Washington." Tom Bedell, Washington Medical Assistance Administration; Sandra Hunt, Coopers and Lybrand; and Laura Tollen, Colorado Risk Adjustment Project."
Presentations and Testimony
Carolyn Madden, "Findings from Implementing Health Status-Based Risk Adjustment in an Employed Population," at the Washington State Hospital Association, April 1994, Seattle.
Carolyn Madden, "Findings from Implementing Health Status-Based Risk Adjustment in an Employed Population," at the Group Health Cooperative, May 1994, Seattle.
Carolyn Madden, Mark Hornbrook, and Vicki Wilson, "Implementing Risk Adjustment in a Community-Rated Environment: The Case of Washington," at the Workshop on Risk Selection in a Reformed Health Care Marketplace, The Robert Wood Johnson Foundation, June 12, 1994, San Diego.
Mark Hornbrook, Carolyn Madden, and Vicki Wilson, "Medical Risk Distribution Among Competing Health Plans: Payment Reform in Washington State," at Risk Selection in a Reformed Marketplace, Alpha Center, October 6, 1994, Washington.
Carolyn Madden, "Data Issues Related to Implementing Health Status-Based Risk Adjustment in an Employed Population," at the RWJF conference Working with Large Insurance Company Databases: Avoiding and Overcoming Pitfalls," October 1994.
Carolyn Madden, "Panel Discussion on Risk Adjustment," at the Academy of Healthplan Purchasing Cooperatives Third Semi-Annual Conference, July 1995, Seattle.
World Wide Web Sites
http://boat.washington.edu/risk-adjust/html/reports.html (no longer available) provided information on the status and results of the project.
Report prepared by: Karin Gillespie
Reviewed by: Marian Bass
Reviewed by: Molly McKaughan
Program Officer: Nancy L. Barrand