As mentioned earlier, aside from developing national estimates, we also modeled the President’s proposal and the Congressional tax proposal for each state. Again, we used the HBSM to estimate the state impacts. The policy options were modeled using similar assumptions and methods as described above for the national estimates. This facilitates comparisons between the nation and states, as well as across states. However, it was necessary to incorporate data from some different data sources in order to ensure that the model results are as relevant to each state as possible.
Baseline Data. As with the national estimates, we begin by developing the current-law estimates of health spending in each State. This process is similar to the national process. We used state-specific population and expenditure data to develop baseline estimates for each state. We used the state-specific sub-samples of the CPS data for 2004 through 2006 in order to estimate the population figures needed, which were described above.
Unfortunately, no single entity maintains a detailed accounting of all health expenditures by state. A major reason is that our current multi-payer system does not require the kind of centralized systems for the payment of health care services that would be conducive to collecting and evaluating overall health expenditures. For example, private employer health plans generally maintain separate health data systems that are not conducive to tracking health expenditures for individual geographic areas such as states (e.g. some workers are employed in firms where the corporation and its health plan are headquartered in a different state).
Our approach to developing state health expenditure accounts is to piece together estimates of health spending by source of payment and type of service from the limited data that are available. One source is the Centers for Medicare & Medicaid Services, which has developed estimates of total health spending, as well as Medicare and Medicaid spending by type of service for each state between 1980 and 2004.10
We also used a report prepared for the National Association of State Budget Officers to get expenditures for State and Local programs, such as state-run AIDS treatment programs, chronic disease programs, and funding for services provided at safety net clinics.11 This data was used to estimate the proposals’ impacts on spending for the existing other State and Local government health programs that focus on funding services for the uninsured. It may be possible for savings in these other public programs to be incurred as the uninsured become covered because of the proposals.
While data on spending for government programs in the state are available, comparable information on health spending under specific types of private insurance and household out-of-pocket spending generally is not available for individual states. We estimated these spending amounts using data from MEPS household and employer surveys. The employer survey contains information representative at the state level such as average health insurance premiums, including splits for employer and employee shares, by firm-size. We use this data as the basis for our private health insurance spending estimates by state.
As mentioned above, the household survey provides nationally representative information on the sources and uses of funds. We use the CPS-based, state-specific population estimates to re-weight the MEPS household data in order to develop state estimates for out-of-pocket spending and uncompensated care. We also use the household data to distribute source of funds by type of service.
Information from all of these sources were incorporated into our analysis to develop a detailed accounting of health spending by state. This process required converting some of the health spending data from a government fiscal year basis to a calendar year basis. We also needed to project all health spending estimates to 2009. Our projections were based on the trends in national health care expenditure projections estimated by CMS, with some adjustments for differential spending across states based on historical spending trends.
After estimating baseline spending for all fifty states, we calibrated the state estimates to ensure that the sum of the state estimates is equivalent to our national estimates. This is necessary as different data sources were used to model the national and state baselines.