Editor’s Note: This post by Tom Miller of the American Enterprise Institute appears in tandem with last week’s publication on the Health Affairs Web site of an article by Jack Hadley of George Mason University, John Holahan of the Urban Institute, and coauthors, which estimates the cost of covering uninsured Americans. Miller’s post follows posts on the Hadley-Holahan article by Henry Aaron of the Brookings Institution and Len Nichols of the New America Foundation.

Jack Hadley, John Holahan, and their colleagues go well beyond simplistic back-of-the-envelope estimates of the added cost of covering the uninsured. They look inside an imperfect box of data to reveal a number of important policy insights. What’s most notable right off the bat is that this latest study continues to rain on the political parade that claims that the uncompensated care costs of the uninsured are largely recycled into higher private insurance premiums.

The key numbers reported by Hadley et al are that the uninsured received $56 billion in uncompensated care for 2008 (2.2 percent of all health spending), of which approximately 75 percent ($43 billion) was paid for through public funds directed to the costs of the uninsured. No matter how one tries to slice and dice this, there’s just not much of a residual of implicitly subsidized health costs to be potentially picked up by shifting those costs to private payers of higher premiums. The purported cost shifting hydraulic is running short of uncompensated care fluid. Its so-called “hidden tax” component may be as hard to find as someone living in a witness protection program. Hadley and Holahan’s older numbers from 2004, using somewhat less refined methods, were in similar proportional ranges, amounting to $40.7 billion in total uncompensated care costs for the insured and an offset of $34.6 billion in public funding for them.

The authors complete this analysis by separating fact from fiction in the hospital sector. They demonstrate that, to the extent that some hospitals are able to raise their prices, it is not a response to increased demand for care by the uninsured but primarily due to changes in the ability of insurers to limit price increases and fluctuations in reimbursement rates in Medicare and Medicaid. However, this remains a murky area of creative accounting, given the fuzzy boundary lines between charity care and bad debt and the lack of more comprehensive data that would include for-profit hospitals.

The authors’ takeaway conclusion on this health policy front is that cost shifting finances a relatively small amount of uncompensated care and it has only a very small impact on the level of private insurance premiums. This finding is not popular with those who spin the political theory that the privately insured should support subsidized universal insurance coverage, because “they are already paying” for most of the costs of uncompensated care anyway. Several recent studies in the “hidden tax” genre of cost shifting fall short of Hadley and Holahan’s more accurate estimates. Nichols fails to distinguish between the uncompensated care costs of the full-year uninsured and the part-year uninsured. Thorpe and Families USA overstate the number and costs of the full-year uninsured, while undercounting the amount of public funding for uncompensated care.

Hadley et al excel in adjusting for the clustering of health care spending by the part-year uninsured during periods of the year when they have coverage (85 percent of their annual spending), which thereby accounts for their much smaller amounts of uncompensated care. The use of the statistical fiction of the “full-year equivalent” number of uninsured by Nichols overlooks this distinction, as well as the inherent limits of relying on Current Population Survey estimates of the full-year uninsured. Leveraging the below-cost reimbursement rates of Medicaid to boost the aggregate amount of uncompensated care to be “cost shifted” not only mixes public program apples with private market oranges; it runs head on into the contradictory policy recommendation of expanding insurance coverage through public program expansion that only compounds the below-cost payment problem (or as bankrupt retailers once hoped to achieve: “Making up our losses on volume”).

Cost Shifting To The Privately Uninsured
Is The Weakest Argument For Expanding Coverage

The bottom line is that the cost shifting case for covering the uninsured is the weakest of several arguments for insurance coverage expansion, even though it allows non-profit hospitals, less-efficient private insurers, and public health program appropriators to try to shift blame to someone else while hoping to increase their market share and incoming revenue streams. The unanswered question remains: If the uninsured eventually receive so much uncompensated care later due to the deferral of more timely early care that would have averted costly conditions, why do they still, on average, persistently receive less than half the amount of care received by the insured over time? The most likely answer is that they continue to receive less care, and pay for it through poorer lifetime health.

Hadley et al had to maneuver through difficult statistical terrain by making a number of mostly defensible judgment calls, including updated adjustments to the fraction of Medicaid and Medicare disproportionate share (DSH) payments, and other supplemental provider payments, that are allocated to financing of uncompensated care (an area of creative financing abuse that is long overdue for necessary health policy reform). They also indirectly rebut recent contentions that the uninsured are plagued by costly medical conditions, in pointing out that many of the uninsured are younger and healthier than the insured, and should be expected to have lower medical spending independent of their lack of insurance. Their allocation of some Veterans Health Administration spending ($5.4 billion) to the financing of uncompensated care spending is relatively inconsequential, because, under their methodology, reallocating it to the “insured” spending category would also reduce their overall uncompensated care totals by the same amount, and leave the residual potentially prone to cost shifting to private premiums unchanged as well.

Some Assumptions And Methods Used
By Hadley And Coauthors Can Be Questioned

However, some of the authors’ assumptions and methods can be questioned. First, these latest, updated estimates of the amount of health care spending received by the part-year and full-year uninsured differ noticeably from their previous ones. In 2004, Hadley and Holahan estimated that the full-year uninsured received 55 percent of the dollar amount of health care that the full-year insured (including both those with private coverage and with Medicaid) spent. In this latest study, the full-year uninsured received in 2008 only 36 percent of the per capita health spending by the full-year insured, and 43 percent of per capita spending by the full-year privately insured. The part-year uninsured received 83 percent of the dollar amount spent by the full-year insured in 2004, but only 67 percent in 2008 (and 76 percent of per capita spending in 2008 for the full-year privately insured).

Brad Herring used similar MEPS data from 1996-2000 (Hadley and Holahan’s 2004 work relied on the 1996-1998 MEPS data) in a Journal of Health Economics study of the full-year uninsured which estimated that they received about 65 percent of the care spent by the privately insured. Interestingly, Herring also found in related work that the amount of “free care” provided to low-income individuals (those below 300 percent of the federal poverty level, paying 31.7 percent of their care out of pocket ) is higher than the amount provided to high-income individuals (those above three times FPL, paying 47 percent OOP), and that people with the highest amount of health care utilization paid the lowest share out of pocket (12.5 percent). In other words, the safety net of uncompensated care expands the more that one needs it, on the basis of higher costs and lower income – like a means-tested catastrophic insurance policy.

Hadley et al do not delve into the distributional allocations of uncompensated care to that degree, but it would be helpful if they explained in more detail why and how their 2008 findings on how much the part-year and full-year uninsured receive in care differ from these earlier estimates, and why they chose to change their baseline of comparison from all of the full-year insured population to just the lower-spending full-year privately insured.

Second, several other questionable judgment calls that could change estimates of total uncompensated care involve the authors’ decisions not to adjust their estimates of physicians’ uncompensated care to account for the recent estimates by Gruber and Rodriguez of more than full offsets from higher cash payments received from the uninsured, and to use the higher MEPS rather than lower CPS estimates of the number of uninsured (which increased estimates of uncompensated care costs by $12 billion in 2008).

Third, Hadley et al also make new estimates for the total increase in health spending – approximately $122 billion more — under universal coverage. They assume that an uninsured person’s new benefits under universal coverage would be similar to the distribution of health benefits currently received by lower- and lower-middle-income people with either private or public coverage. However, that overlooks the different dynamics of demand for health spending by the higher-income uninsured (very conservatively, numbering at least several million persons).

Moving Towards Universal Coverage May Increase Costs
In Ways That Hadley And Coauthors Do Not Consider

More significantly, the authors too conveniently assume that provider payment rates, administrative costs, and the generosity of coverage would stay largely the same under universal coverage. Perhaps they should have provided some alternative estimates for likely costs under even the less-than-universal-coverage health plan of Senator Obama. The range of first-year costs stretch from the optimistic, low-end in-house estimates by his advisers of at least $65 billion, to a more recent, fully-loaded one by health economist Roger Feldman at a staggering annual cost of $452 billion.

A more interesting omitted variable involves the dynamic effects of a major coverage expansion on the price, availability, and quality of health care and how cost effective the initiative might prove to be. Amy Finkelstein’s 2006 analysis of the impact of the introduction of Medicare in 1965 on real per capita health spending concluded that market-wide changes in health insurance may have fundamentally different effects on the health care sector than what partial equilibrium analyses such as the Rand Health Insurance Experiment would suggest. Finkelstein estimated a 37 percent increase in spending associated with the first five years of Medicare and also suggested that a dynamic feedback loop could have produced long-run effects of Medicare on technological change and health spending beyond the initial ten-year post-Medicare window she analyzed.

Hadley et al’s more static modeling of the cost effects of universal coverage are not on as large a scale as the early years of Medicare expansion. However, they did not even appear to consider how market-wide changes in health insurance might fundamentally alter the character of medical care both for individuals who experience a change in insurance coverage, and for those who do not as well. A related paper by Finkelstein and McKnight further found that the introduction of Medicare (universal health insurance for the elderly) appears to have had no impact on elderly mortality in its first ten years.

All of the above aggregate quantitative analyses still omit other key policy variables. Moving bundles of private and public dollars around does not ensure that the value of our investments in health — measured by long-term health outcomes relative to their all-in resource costs — will be improved if not optimized through health-spending centric means alone, as compared to a more diverse portolio of inputs beyond the health services sector, including improvements in education, information transparency, navigational assistance, decision support, and time preferences. Nevertheless, we know more than we did before, if not enough about what matters most, thanks to the contributions of the authors of this thorough and generally well-crafted study.