When we examine the spending, or patterns of care, or availability of providers of various kinds across cities or other small areas we find what are often big differences. These differences remain a bit of a puzzle about the U.S. health system (or non system). Why do they exist? What patterns do they take across places? Are they a problem worth fixing in their own right?
What do Small Area Variations Look like?
The table below describes some of the U.S. cities and a few indicators for the Medicare population in 2012. All Medicare FFS beneficiaries in these places are included.
The spending data (column 1) is standardized across places for differences in provider prices — so the spending averages per beneficiary are dues to the mix and levels of services provided. The data don’t prove anything, other than there are large differences in the kind of care that are provided to persons in different places, even though these people have exactly the same insurance coverage. McAllen, Tx. and St. Petersburg, FL. have exceptionally high levels of medical utilization, compared to other locations. These are not necessarily the highest or the lowest spending areas in the U.S.
300 or so Hospital Referral Regions are shown below in the scatter plots. They show the very large (vertical) differences across HRRs in attributes of the local health systems.
The earliest work in SAV was done by Wennberg and Gittlesohn (Science 1973). It demonstrated extreme variations in practice in small communities throughout the state of Vermont, as shown here. There were less than 20 small areas in the small state, and the data show the high and the low values for several indicators. The most shocking data is the five-fold differences in tonsillectomy rates across regions of the state.
The researchers who have studied small area variations (SAV) in health care document the fact that places with more spending or higher utilization often have more providers (more beds, more doctors) than do other places. This is sometimes referred to as Supply-induced-demand or Roemer’s Law. In the earliest example of SAV research, Wennberg and Gittlesohn (1972) demonstrated that counties around the state of Vermont had quite different patterns of utilization (hospitalization patterns, surgical utilization patterns, others) and that these patterns were often accompanied (correlated with) patterns of high and low supply of providers. Places with more beds had more hospital utilization than places with fewer beds. And, places with more surgeons did more surgery than places where there were fewer surgeons.
In later work, comparing extremes in Medicare utilization between Miami and Minneapolis the authors also suggests that the differences in utilization is not fully explained by differences in age or population severity of illness. But, these cities just have enormous differences in usage patterns and total spending. During the portion of a lifetime that enrollees are on Medicare, spending in Miami is about $50,000 higher per enrollee than in Minneapolis.
These patterns (Miami high, McAllen high, Minneapolis low, Birmingham low, etc.) are not one-year anomolies. Indeed, the differences remain stable year to year over many years, and are not much ameliorated my market forces, which might be thought to eliminate the differences over time. They seem to remain.
Why do they appear in the first place? The literature suggests that they arise from two kinds of ignorance, and may persist in part because of the way malpractice issues are handled.
Science Ignorance. Evidence shows that about half of the occasions when medical doctors make decisions regarding a test, a procedure, and drug, or whatever, there is no scientific evidence to follow. The first chart shows this evidence. The second chart shows the extent to which evidence is used, when it is available.
On about half the issues in adult medicine science exists about what to do. But, half the time science exists on some issue of care it is actually used to guide practice. So, as Rand’s McGlynn et al report, in practicing with adult patients, the decisions made by physicians are guided by practice only 28.6% of the time—- and not guided by practice 71.4% of the time. Essentially, what this means is that there is a good bit of uncertainty in the practice of medicine— or we might say a good bit of discretion, too.
Consumer Health Ignorance. There is widespread evidence to suggest that patients are not well informed about the benefits, costs and risks of their options when making medical decisions. This ignorance is described by economists as a marketplace where the “demander” is not well informed and the market is characterized by “asymmetric information”; a form of market failure. In such circumstances, customers often seek out advice from experts including suppliers. So, if trust has been established, patients rely on their provider to give advice on what decision to take, given the options available. Well functioning markets do not have such ignorance by the customer, and do not put suppliers in such a conflicted and self-interested role.
The combination of limited science that might guide providers, and the dominant role assumed by providers in deciding what to do for patients creates the basis for SAVs. But, if provider discretion is rampant, why don’t we just observe lots of variation across providers in EVERY market area? Why does each market area assume a somewhat different central tendency about practice: a local standard so to speak. This has been studied, but no “smoking gun” has been identified. Some tried to test whether it comes from where the doctors went to medical school. Or, whether it is age, or some other tangible characteristic of the place. Or maybe level of income of the marketplace? No cigar. The pattern isn’t that obvious to detect.
What seems to be happening is that each place exerts a coercive influence on doctors who practice there, encouraging them to adopt the practice style of the existing physicians. Of course not everyone adopts the exact same criteria for when to recommend surgery, or when to do a MRI for example, but there seems to be some sort of “central tendency” in Miami, that is different than the “central tendency” in Minneapolis. See the illustration of the distribution we might see something liken the following
This simple scatter diagram shows how doctors in these two places (red, black) differ in terms of their practice style in using hospitals. The averages for the two places is shown in the two vertical lines—which are noticeably different.
One of the most novel of the Wennberg studies of SAV was done to compare practice styles in two cities both with dominant cultures of medical education: Boston and New Haven. The study showed huge differences in the use of hospitals, with the Boston practice style being a more intense use of hospitals, by every measure. The key findings were:
Why does this sort of convergence into a more or less distinct practice style happen in each place? Well, it starts from the enormous discretion that is afforded doctors and their influence over patient choices. There may be some peer influences as well. But possibly most critical as a cause for non random “clustering” in each city is the way the malpractice law and precedent has developed over time.
Malpractice
When sued for some act of malpractice, doctors are not held to the standard of “best practice” as might be documented from the science literature, or the most recent textbook from an expert from Harvard Medical School. No, the standard to which doctors are held is a community standard: what would have other doctors in the community have done under similar circumstances? Nobody from Birmingham is held to the standard of care in NYC or Boston, nor to the most recent scientific evidence about what works best. This tells doctors that the risky position is being an outlier in the community.
This probably produces regression to the mean in terms of practice style in every community, though not to the same practice style across communities. So, why are the local standards in Miami and Minneapolis so different? Part of it may result from supply endowments of doctors and hospitals, and also from the historic patterns of care by leading practitioners in those places. That is all that is known.
Why should we care if there are variations in practice styles?
We care primarily because it represents substantial evidence of waste in the health system. The literature suggests that if we limited all communities to the 10th percentile in per capita spending U.S. health care costs would be cut by 15-30%.
professorgaumer 