Funding the VA of the Future: Testimony of Jessica Banthin
Statement of Jessica Banthin, Ph.D., Director of Modeling and Simulation, CFACT, AHRQ, DHHS, Before the Committee on Veterans' Affairs, U.S. House of Representatives
April 29, 2009
Good morning, Mr. Chairman and Members of the Committee. Thank you for the opportunity to testify before the Committee on the issue of modeling long term projections. Before beginning the substance of my remarks, I want to state that the Agency for Healthcare Research and Quality (AHRQ), an agency of the Department of Health and Human Services (HHS), has benefited from extensive collaboration with the Department of Veterans Affairs (VA) in the areas of health services research, patient safety, and clinical quality of care. We consider the VA an important partner in improving health care.
I serve as the Director of Modeling and Simulation in the Center for Financing, Access, and Cost Trends at AHRQ. At AHRQ, we have extensive experience with working on sophisticated health care models. For example, we developed a simulation model that estimates the number of eligible uninsured children in the U.S. and can be used to project enrollment in Medicaid and the Children's Health Insurance Program (CHIP), and informs outreach efforts to increase enrollment of eligible children ages 1-4. We worked closely with actuaries at HHS's Centers for Medicare and Medicaid Services (CMS) to benchmark national health expenditure estimates.5 In addition, researchers at AHRQ designed an economic microsimulation model that predicted consumer choice of health insurance in response to changes in health insurance offerings.6 The model also projected changes in total health care spending resulting from the change in insurance offers.
I have had the opportunity to review the RAND report on the VA Enrollee Health Care Projection Model (EHCPM).7 The EHCPM includes three major components: an enrollment projection model, a utilization projection model, and a unit cost projection model.
The RAND report draws distinction between actuarial models that are based on historical trends and economic models that incorporate behavioral parameters. I have worked with both actuarial and economic models. I have also worked with models that combine elements of both approaches. There are caveats to all long-term projection models.
In my testimony, I will briefly describe an enrollment model that we have constructed at AHRQ that can be used to project children's enrollment in Medicaid and CHIP. I will also discuss the benefits, caveats, and limitations that affect long-term cost and utilization projection models.
An Example of Modeling Medicaid and CHIP Eligibility and Enrollment
In AHRQ's modeling efforts, we model Medicaid and CHIP enrollment using survey data from our Medical Expenditure Panel Survey (MEPS) as well as State-specific eligibility rules. We make use of information on family structure and family income and then apply State-specific eligibility rules to all sampled children in the MEPS data. We simulate the eligibility of each child for public coverage through Medicaid or CHIP. We then compare the simulated eligibility status to the child's reported insurance status. Many eligible children are enrolled in public coverage, and our model supports the calculation of take-up rates.
Next, we use output from our eligibility simulation model to develop economic models that explain why some children are more likely than others to enroll. These models, as with all actuarial and economic models, are limited by the available data. We cannot easily measure the effects of factors that are not observed or measured. Nonetheless, the enrollment (or take-up) model identifies the factors that have the largest marginal effects on enrollment. We find, for example, that among children who are eligible for public coverage age, children's health and disability status and parents' employment status are strong predictors of enrollment.4 These models can easily support longer term enrollment projections and are flexible enough to account for changes that may affect enrollment decisions.
In the aforementioned studies, MEPS data were used. Data from the American Community Survey (sponsored by the Bureau of the Census) also measure veteran status. As of 2008, the American Community Survey is also measuring health insurance status.
Cost and Utilization Projections
The long-term projection of costs and utilization is very difficult because of the number of factors that affect use of health care services. Factors include unpredictable changes in both the demand for and the supply of various services. Technological change can yield new treatments for medical conditions and improved diagnosis of ailments. Changes in the prevalence of disease can affect the demand for care. When AHRQ projects health care expenditures, we refrain from applying complex models and assumptions and instead apply publicly available projections from census data (regarding demographic changes) and from CMS (regarding expenditure growth), so we project expenditures using a more conservative approach that is more aligned to actuarial methods. AHRQ-projected expenditure data are publicly available, so modelers can then use these data to develop more complex microsimulation models that predict the cost changes resulting from various behavioral parameters and assumptions. These more complex microsimulation models with behavioral parameters are critical for policy analysis, but their long-term accuracy in projecting expenditures is very hard to gauge. The advantage of having extremely detailed information from private claims data on the use of health care services is that the data project use and costs associated with an array of specific health care services. Breaking down long-term projections in this way avoids the need for relying solely on these behavioral parameters.
Issues in Projecting Enrollment, Utilization, and Costs
Programs such as the VA face several challenges in projecting utilization and costs for its patient population when there is limited information on the other non-program sources of care patients may use. This issue is more pronounced for patients under age 65 without Medicare claims data to examine. To the extent that the VA patient population is unique and differs in many ways from the commercially insured population, such data limitations present additional challenges in projecting future utilization and costs.
It is important to account for illness severity or morbidity when projecting costs. Morbidity is a strong predictor of both enrollment and use of services. Morbidity can be measured with clinical measures but can also be accounted for with some survey-based measures of patient-reported physical and mental health status, functional status, and work disability. These patient-reported measures have strong predictive power in many economic models of demand for services.
In conclusion, I want to emphasize that there are caveats associated with all long-term projection models, whether they use actuarial or economic methods. In addition, the accuracy of all projection models depends critically on the available data. Without sufficient data there may be areas in the models that rely on best guesses rather than solid data. As most modelers know, long-term projection models can constantly be improved and enhanced. This is usually an ongoing process. Nevertheless, the VA Enrollee Health Care Projection Model is a very sophisticated model that benefits each year from better information on the current veteran population.
Mr. Chairman, this concludes my prepared testimony. Thank you, and I would be happy to answer any questions you may have.
1. Hudson J, Selden T. Children's eligibility and coverage: recent trends and a look ahead. Health Affairs 2007;26(5).
2. Hudson J, Selden T, Banthin J. The impact of SCHIP on insurance coverage of children. Inquiry 2005;42(3):232-54.
3. Selden TM, Hudson JL, Banthin JS. Tracking changes in eligibility and coverage among children, 1996-2002. Health Affairs 2004;23(5):39-50.
4. Selden TM, Banthin JS, Cohen JW. Projecting eligibility and enrollment for the State Children's Health Insurance Program. 1999; AHCPR Pub. No. 99-025.
5. Sing M, Banthin JS, Selden TM, et al. Reconciling medical expenditure estimates from the MEPS and NHEA, 2002. Health Care Financing Review 2006;28(1):25-40.
6. Zabinski D, Selden TM, Moeller JF, Banthin JS. Medical savings accounts: microsimulation results from a model with adverse selection. Journal of Health Economics 1999;8(2):195-218.
7. Harris KM, Galasso JP, Eibner C. Review and evaluation of the VA Enrollee Health Care Projection Model. RAND 2008.