Irene Papanicolas, Liana R. Woskie, and Ashish K. Jha write,
The United States spent approximately twice as much as other high-income countries on medical care, yet utilization rates in the United States were largely similar to those in other nations. Prices of labor and goods, including pharmaceuticals, and administrative costs appeared to be the major drivers of the difference in overall cost between the United States and other high-income countries.
Pointer from Tyler Cowen.
The paper is very readable, and the tables are very clear. For example, the ratio of specialist physician pay to the mean wage in the U.S. is 5.3 in the U.S. compared with 3.9 in the next-highest country. For general physicians, the ratio is 3.6 in the U.S. compared with 3.3 in the next-highest country.
The study contradicts most of what I believe about comparative health care spending. It also contradicts the findings of Random Critical Analysis. I think that the probability that the study is mostly accurate is less than fifty percent, but greater than zero.
A challenge is that data are very shaky in many areas. The authors write,
Even when data were collected from the same source, issues of comparability remain because of fundamental differences in how systems are organized and, in turn, how care is categorized. Two areas of particular concern are outpatient spending and the primary care workforce. We attempted to address limitations in the workforce data by utilizing a functionality-based approach to identifying who provides primary care services in each country and by cross-referencing resulting numbers with country experts.
Random Critical Analysis used different data on the health care work force and got very different results.
Is the study actually inconsistent with the RCA analysis? Couldn’t they both be correct?
I’ll probably critique this paper on my blog when I have more time, but their data are fairly consistent with the data I’ve reviewed previously and with my arguments. Their analysis and conclusions are largely at odds with my views though.
For instance, it’s quite possible we pay somewhat more for (prescription on-patent) pharmaceuticals and the physician wage premium is somewhat larger, but that our prices are still very much in line with our income levels (as OECD Health PPP series and quasi-price estimates indicate). Physician wages account for maybe 8-10% of health spending in the US, the estimated wage premium relative to average wages is surely even smaller. Returns to skills are also higher in the US and this is likely to explain a large part of this presumed residual. The flip side of this coin is that more modestly skilled workers in the healthcare sector are likely to earn lower wages conditional on the average wage, meaning (amongst other reasons) we shouldn’t assume somewhat higher wages for physicians necessarily implies higher remuneration in the healthcare sector overall (conditional on national average wages).
One point that sticks out: they argue, as many have previously, that utilization doesn’t explain US health spending. This is approximately correct, but what they fail to mention is that these very same indicators explain essentially nothing within and between other OECD countries too, this despite the fact that there is quite a lot of variance in health spending and it is highly correlated with national income levels. If prices explain only a very modest fraction of the difference and rate of health system encounters (“utilization”) explains ~0, then it follows that the source is the difference is largely higher intensity per encounter, i.e., doing more and doing more cutting-edge treatment/diagnostics/etc on a per encounter basis.
In other words, intensity seems to largely explain why health spending increases as countries get richer. High quality, high coverage international data on intensity aren’t available to measure this directly very well, but it is very much consistent with the data that we do have (e.g., MRIs & CT scans, selected surgical procedures, etc) and this story is very much consistent with our domestic data.
I make a living resolving those kinds of data issues for oil companies. I would expect data quality issues in health care to be even worse.
I think I may have lost track of what the issue is here. Can someone remind me?
Are we trying to understand why we spend so much more on health care than other countries? And one possibility is that we have higher costs? So what are the other possibilities? That we consume greater quantities? That a greater proportion of what we consume is of the fancy/expensive sort? Are there others?
I believe that Megan McArdle has argued that, unlike other countries, we can’t afford single payer because during the 70s and 80s we experienced a spike in prices that other countries did not undergo. So is her view consistent with that of the paper quoted here and inconsistent with Arnold’s?
I am working on a response to the RCA post, which relies a great deal on the approach taken by Koechlin, Lorenzoni, and others. Finding time for unpaid effort is a challenge.
But that approach uses techniques that I would think you would usually not be very fond of in other contexts, relying as they do on “quasi-price” estimates and formulas, somewhat arbitrary baskets and indices, GDP/PPP-based purchasing power adjustments, consumption indices as proxy for “wealth” and so forth. Apples to apples comparisons are really hard, so the data is recalcitrant, and will only talk when tortured, but will then confess to anything the particular inquisitor wants. Yes, you can find some special combinations of measures and adjustment which end up looking like they put a bunch of countries on a line, but how do you know that wasn’t the result of cherry picking?
My hunch is the quasi-price/consumption-based wealth-deflator approach is bunk. As an example, it’s hard to beat the transparency and thoroughness of Singapore’s Ministry of Health (Cost and Financing division) published data for hospital bills by condition/procedure, which includes private options, and is a global Mecca for medical tourism. Singapore is up there with the US in terms of NGDP per capita (at market rates of exchange), and 50% richer if you believe the PPP conversions. And yet, consistently, the highest quoted prices on the MoH site (as much as 20 times more than the lowest charged rates) aren’t even close to the typical amount charged in the US. A good example could be to look up appendectomies, where there isn’t much innovation or need for new, expensive devices, and utilization rates are dictated by circumstance, where you can’t put a person on some soft-rationing long waiting list, and which won’t go up or down depending on levels of wealth or quirky fluctuations in demand. The highest quoted price is about $15K in U.S., whereas the US average is over twice that amount.
I think the inputs-based approach (the rival to the outputs-based approach) is more likely to give an accurate picture of what is really going on, since one can more easily analyze sources of discrepencies in the price of inputs, or the use of different amounts of inputs to produce the same outputs (for example, as one would expect to accompany liability-based defensive medicine).
Handle:
Are you using charges or real cost/reimbursement for the US expenses?
Charges are meaningless as a basis for comparison…
Utilization may require larger samples for comparison while most of the cost difference lies in specialization. It is extremely difficult to compare prices even in the US since different arrangements for cost allocation can shift these between doctor, facility, ancillaries, and support. It is also easy to over fit curves when one is such an outlier. It is also misleading to consider this choice when cost and benefit are so widely separated through non transparency and third party remuneration.