The legal and policy questions following the US Supreme Court’s ruling in Gobeille v. Liberty Mutual Insurance Co, 577 U.S. ____ (2016) have been widely discussed, including in Health Affairs Blog posts by Sage, Brown and King, and Newman. The Court ruled that states could not require health insurance plans under The Employee Retirement Income Security Act (ERISA) to submit their health care claims data for use in the state’s all-payer claims database (APCD).
What are the implications to APCDs if ERISA data is not available? While we understand that the potential loss of ERISA data is viewed with concern (that the lack of comprehensive data will undermine the utility of APCDs), the Court’s decision may be not fatal to policy-relevant research. Some or many ERISA plans may continue to voluntarily share their data with states, which is the best case scenario from a research perspective — more data is preferable to less.
The Court’s decision should only be fatal to APCDs if the ERISA data are fundamentally different from the non-ERISA data and the data available to APCDs cannot be used to approximate the ERISA data. If the ERISA populations are identical (or nearly identical) to the non-ERISA populations, the loss of the ERISA data is inconsequential since the non-ERISA data can substitute for the ERISA data.
If the ERISA data are somewhat different but those differences are understood, then one should be able to use the non-ERISA and any available ERISA data to adjust for the missing data. In general, there are likely to be methodological, research, and agenda-related implications, but the data available to APCDs should still be useful for a variety of important activities.
There never was an ‘All’
Researchers deal with incomplete and missing data in many contexts. In fact, to date, there has never actually been an APCD that has collected “all” data. No state has literally collected payment and utilization data for all health care services from all payers.
Perhaps the most glaring omission, and potentially most important, are those payments made by individuals outside of an insurance scheme. States have generally shied away from attempting to collect data on the uninsured because it was difficult and expensive.
Maine makes an effort to estimate payments using pseudo-claims for services provided through Maine Health, the largest health system in the state. However, this approach simulates an insurance claim for tracking purposes without actual paid amounts for an uninsured individual. From a public policy perspective, it is important to know both what services are consumed by the uninsured, and what they are paying. In some states the uninsured still represent a significant proportion of the population. In Texas, for example, the figure is as much as 20 percent.
In addition to the uninsured, free care offered by providers, sample pharmaceuticals provided by physicians, and services individuals elect to pay for outside of their traditional coverage are not captured. Federal schemes such as TRICARE, the Indian Health Service, the Veterans Administration, and the Federal Employees Health Benefits program are generally excluded from state APCDs. Moreover, some state APCDs do not get their state’s Medicaid data and others do not have access to the Medicare data.
Finally, to the extent that APCDs can be used to evaluate providers or suppliers on prices or quality, there will be a number of providers whose “markets” or patient catchment areas cross state borders. The APCDs in those states where the providers are located will be missing any out-of-state patients’ claims from for those providers. For instance, the Mayo Clinic in Rochester, MN likely draws patients from neighboring states (e.g., ND, SD, NE, MN, IA, and WI), at a minimum, as well as nationally and internationally. However, the Minnesota state APCD is only able to collect data from those insured in Minnesota.
ERISA/Non-ERISA—should there be a difference?
The great utility of claims data for research purposes comes from the industry-wide use of standard coding schemes and the fact that everyone is paid in dollars. These two conventions allow researchers to use claims data to examine the services consumed (utilization) and the prices paid. The amount spent on health care (expenditure) in any population is simply a function of price times utilization.
From a state APCD’s perspective, fundamental differences between ERISA and non-ERISA plans in price or utilization are the most relevant concerns. If the prices paid for health care services or if the rates of utilization of health care services varies, then APCDs would need data from both populations or would need to know how the populations’ prices and/or utilization differ.
Generally, when an employer decides to create an ERISA plan, the employer engages a third-party administrator (TPA) to manage the benefits. These services can include arranging the network and negotiating with providers, managing enrollment, and processing claims. Many national and local insurers offer their network, and presumably their network pricing, to TPA’s and ERISA plans. To the extent that this is the case, then pricing for the ERISA plans should not look fundamentally different from pricing for non-ERISA plans.
Similarly, there may be some consumption of health care services that is induced or reduced by differences (e.g., benefit designs) in ERISA versus non-ERISA plans. However, there is no theoretical reason that a significant proportion of health services utilization should be driven by an employer’s decision to offer a health insurance benefit under ERISA or not.
We recognize that there could be differences in the distribution of age, gender, or benefit design across plan types (i.e., health maintenance organization, preferred provider organization) between ERISA and non-ERISA populations, which may affect utilization or possibly prices. Hypothetically, it could be the case that workers age into larger more established employers that have a higher probability of having an ERISA plan. If it is the case that providers negotiate higher rates for certain services that are more common among specific age-gender groups, there will appear to be difference in utilization, price, and/or expenditures between ERISA and non-ERISA populations. However, if the differences in member characteristics are known, methodological or statistical approaches to adjust for these differences can be applied to the data.
ERISA/Non-ERISA—is there a difference?
While we have suggested above why price and utilization should not substantially differ across ERISA and non-ERISA data, we are in a position to test these propositions empirically using a national claims data set with both types of claims. In general, the results are consistent with the intuition — there are only small differences in price and utilization between the ERISA and non-ERISA populations.
We focus here on inpatient hospital prices and individuals with point-of-service (POS) plan types. In our data set, POS plans accounted for 85 percent of the ERISA membership in 2014. Among the non-ERISA population, POS plan types accounted for 51 percent of membership. A full set of results and a complete discussion of the methodology can be found in our issue brief and we encourage others with access to similar data to replicate our analyses to determine how generalizable our findings are.
First, with 2014 data, we compared average price levels between the two populations to assess whether price levels differ. Using the average price of each diagnosis-related group (DRG) in the ERISA and non-ERISA populations, we calculated the ratio of weighted average inpatient prices assuming the distribution of services was the same in both populations. This assumption eliminated any differences in utilization patterns between the two populations.
The POS inpatient price ratio ranged from 0.96 to 0.99 over the five-year study period (2010-2014). This ratio can be interpreted as the percent difference between non-ERISA and ERISA data. In other words, the non-ERISA weighted average inpatient price was 1 to 4 percent less than the ERISA weighted average inpatient price, suggesting only modest differences between the populations.
We also examined DRG-level prices for a subset of inpatient services, the 10 most expensive DRGs in the non-ERISA population in the 2014 data. On average, the non-ERISA DRG-level prices were 2 percent less. However, there was variation in prices at the DRG-level within each plan type. The smallest variation in price at the DRG-level was among the POS plan type.
Finally, we assessed utilization by comparing the top ten most frequent DRGs in both the ERISA and non-ERISA populations in 2014. The 10 most common DRGs were the same across populations and ranked in the same order. Moreover, those 10 DRGs accounted for approximately 43 percent of all admissions and about 22 percent of inpatient spending in both populations. This suggests that prices and utilization, and thus expenditures, are similar for the two populations for the most common services.
It is highly unlikely that that all ERISA plans will opt out of participating in state APCDs. This is particularly the case if states and stakeholders can standardize the process of reporting and resist periodic efforts to change the requirements.
If the rate that employers opt-out of participating is relatively small and somewhat random, the available ERISA and non-ERISA data is likely sufficient for policy research purposes. If, however, only minimal ERISA data is available, the non-ERISA data may be sufficient if one concludes, as we are inclined to, that the ERISA and non-ERISA data are not substantially different. With information about the distribution of ERISA member characteristics, statistical adjustments can be made to the non-ERISA data to study trends in health care costs and utilization in a state.
While Gobeille wasn’t helpful to health policy research, it doesn’t need to be fatal.
from Health Affairs BlogHealth Affairs Blog http://ift.tt/29PTeZR
No comments:
Post a Comment