The Affordable Care Act was based on homogeneity. Not only would its provisions be disseminated equally, but smoothing the peaks and valleys of health care utilization would liberate the funds necessary to finance it. Under reform, Newark would come to resemble Grand Junction CO, and Mayo would be the model for Manhattan. No longer would Los Angeles, home to the nation’s largest concentration of poverty, consume more resources than Green BayWI, where poverty is infrequent. Regional variation in income and poverty could be ignored all together. The problem is “practice variation,” and health care reform will fix that.

Of course, the US  is not homogeneous, and poverty cannot be ignored. In fact, the principal cause of geographic variation in health care utilization is geographic variation in poverty. And now the hilly terrain of health care is coming more sharply into focus.  Over the course of a few short days, from December 15^th to 18th, that message sprung from three separate sources.

First was an observation in the New England Journal by David Blumenthal, the most qualified member of the Obama health care team and the first director of its health information effort. Dr. Blumenthal noted that “good legislation does not guarantee successful implementation. Never before had a country as large, complex, politically decentralized and diverse as theUnited Statesattempted to create a nationwide electronic health information system.” The nation is too varied.

On the same day, the Wyden-Ryan plan for Medicare was announced. While defined contribution plans, including this one, have the potential to skew benefits away from the poor, a Wall Street Journal editorial endorsing it correctly noted that one of its benefits is that “it relies on local information and adjusts with the behavioral and organizational responses that will vary from region to region.”

And the very next day, HHS Secretary Sebelius announced that the Obama administration would no longer insist on defining a single uniform set of essential health benefits for the nation because “coverage that works inFloridamay not work inNebraska.” The New York Times observed that “the move would allow significant variations in benefits from state to state.

The consistent message is that there is a great deal of geographic variation in America. Its hilly terrain includes peaks and valleys of costs, outcomes, preferences, poverty and needs. Reforming health care in an imagined nation of sameness must give way to reforms for an economically and demographically varied nation, and real solutions must be sought not only within the system but within the structure of the terrain in which our patients live.

It is true. The ACA does blame “ill-incentivized health care” for excess readmissions, ignoring the reality that readmissions are most strongly associated with poverty (an ill-advised social condition). Misidentifying causal factors leads to faulty conclusions, and faulty conclusions breed faulty regulations and payment incentives. And so it is in the ACA. Thanks to Kocher and Adashi for illuminating this outrage.

This notion may seem shocking, since it is generally believed that low-income patients receive less health care. After all, many have little or no health insurance, and most have poor access to primary care. Isn’t it the wealthy whose access is best and who use the most? The answer is yes to the first, but no to the second. Access is better for the wealthy, but they use less because they need less. They have better underlying health, and they have social environments that are more conducive to attaining and sustaining health. It is the poor whose health status is poorest and whose needs are greatest.

That doesn’t mean that the care they receive is always efficient, timely or convenient. Nor that it is as effective as the care received by more affluent patients. Inequality exists throughout. But the principal inequality is in their underlying health status. It is worse among the poor, and only some of the difference can be narrowed, even with the best care.

The reasons are well known. They begin in early childhood with poor nutrition, inadequate education and unhealthy behaviors, often accompanied by physical and emotional abuse. The impact of these early experiences often persists into adult life, where the web of causation expands to include inadequate housing and transportation, poor access to proper foods and weak family and social support systems. Their effects become magnified in dense urban environments, with their complexities, segregation, discrimination and threats to personal safety, and further compounded by chronic unemployment.

It is not surprising, therefore, that the poor have more disease and disability, both physical and mental. Nor should it be surprising that, as a result, they have more and longer hospital admissions, more readmissions, more out-patient visits and higher health care spending. Yet this added care is still not commensurate with their greater burden of illness, and despite it, the gap in life expectancy between poor and rich continues to widen.

One reason many believe that low-income patients receive less rather than more care is that they did receive less forty years ago, both in the US and Britain. It was not until the late 1980s that parity was reached, and health care spending among low-income patients has grown disproportionately ever since. Among Medicare enrollees, the poorest one-fifth now use about 30% more than the richest. This difference is even greater among working-age adults. At the extremes, health care utilization in urban poverty ghettos is more than double the rate of affluent suburbs. Indeed, if utilization throughout such regions were at the levels of their wealthiest enclaves, overall spending would be 25-30% less. This decrement is similar to Sir Michael Marmot’s estimate that about one-third of health care spending in England can be attributed to income-inequality.

Unfortunately, some policy-makers have misinterpreted the increaseutilization by the poor as wasteful care in regions that use more, and this has led to policies that disproportionately penalize providers who disproportionately care for the poor. The reality is that income inequality leads to high health care spending and that this draw on resources is unsustainable, all the more so as poverty rates continue to rise.

Not all who are demonstrating on Wall Street may have specific recommendations, but the message that I get from them is that a more equitable America would work better. That certainly is true for health care. Costs are higher and outcomes poorer in countries where income inequality is greater, and the US is #1. This is not to say that our system doesn’t have other problems, from over-regulation to over-utilization, nor that better ways to care for the poor cannot be found. What it does say is that, no matter what else is done, the US will not be able to afford the high health care costs of income inequality and that the only durable solutions are to deal with its root causes.

First, using the conventional approach to measuring physician supply (AMA Masterfile), they  counted the number of primary care physicians and found no significant correlation with either Medicare spending or mortality. This directly contradicts earlier claims by Shi & Starfield that areas with more primary care physicians have lower mortality and by Baicker & Chandra that areas with more primary care physicians have lower Medicare spending, both of which are widely quoted. But we now know that neither is true. In fact, they never were.

Like the current study, Shi & Starfield had also failed to find a correlation between mortality and the total supply of primary care physicians, both family practice (FP) and general internal medicine (GIM). The correlation they found was only with FPs, although they called it “primary care.” However, as I’ve explained, this correlation exists because FP training programs (and therefore FPs) are more prevalent in the upper-Midwest, where minorities are sparse, poverty ghettos are rare and mortality rates are low. FP supply has nothing to do with it, and we now know that primary care supply doesn’t correlate with mortality. It never did. Ouch!

The same for Baicker & Chandra’s claim that areas with more primary care physicians have lower adjusted Medicare spending – about 30-40% lower. The current study shows that this isn’t true. In fact, it never really was. Baicker & Chandra’s claim was based on a statistical shell game that I exposed in Health Affairs. They responded that I was wrong, and Susan-the-Editor (and Dartmouth board member) responded with feminine furry, but it turns out that I was right. As the current Dartmouth study shows, areas with more primary care physicians do not have lower adjusted Medicare spending. Ouch again!

Now for the big story. In addition to using the AMA Masterfile, the Dartmouth team counted primary care physicians a new way. They measured how many primary care services were billed and then converted services into the number of FTE physicians that would, if working full-time, be able to supply them. But forget about the conversion. Stick with the measure. It’s a good measure. It measures services per beneficiary. The Dartmouth team found that areas with more primary care services also had more total clinical services (primary care plus specialty care), and these areas also had more Medicare spending. It is areas like these that the Dartmouth group previously called “high spending areas,” where patients were no sicker and didn’t get any better despite all of that added spending. This gave rise to the mantra about the unwarranted use of supply sensitive service needlessly consuming 30% of the health care budget. But the current study found that patients in the high-spending areas were sicker. And despite being sicker, their mortality was lower. Mortality was lower in areas with more physician services and more Medicare spending. More was more. OUCH!

That’s the exact opposite of Elliott Fisher’s conclusion that more care is associated with higher mortality. According to Fisher and colleagues, this added care is not only wasteful. It’s dangerous. Only a few years ago, David Goodman (senior author of the current study) said, ”more physicians will make health care worse.” And Susan-the-Editor said ”the greater the amount of health care you provide, the more likely it is to kill you,” somewhat awkward syntax but quite damning. And now the Dartmouth team has shown that more health care is associated with lower mortality. Ouch, ouch and ouch again. Three strikes and you’re out.

So how did the new paper spin these observations? It concluded that “a higher level of primary care physician workforce was generally associated with favorable patient outcomes.”  Therefore, medical homes, train more, pay more, bla, bla, bla. But the workforce wasn’t measured. All that was measured was primary care services. And primary  care services where greater where specialty services were greater, and these areas had more Medicare spending. The real conclusion is not about primary care. It’s about medical care. Medicare beneficiaries who received more medical care had better outcomes, even when they are sicker. MORE was MORE.

In publishingcthe same conclusion two years ago, I said ”let the simple truth that health care quality is better in states with more physicians, both primary care and specialists, sweep away the myths and permit greater clarity as planners work to solve the crisis in physician supply that now confronts the nation.” Possibly this latest study from Dartmouth will allow us to do just that.

The full report is available online. The brief summary below, which quotes from it, attempts to capture its major themes.

Geographic Variation in Health Care: Changing Policy Directions     by Jill Bernstein, James D. Reschovsky and Chapin White

As the debate in Congress about health care reform began in late 2008, the Congressional Budget Office (CBO) highlighted work from the Dartmouth Atlas project, which estimated Medicare spending would fall by close to 30 percent if spending in higher-cost areas of the country were somehow reduced to the level of low-cost areas. In policy circles, the idea took hold that modifying payment policies in high-cost areas could reduce unwarranted, inefficient spending. 

Some sources of geographic varia­tion in health spending are warranted, such as input price and the burden of illness. Moreover, if higher spending produced higher quality, it might be warrant­ed, while ineffective or inappropriate treatments are not warranted. Over time, as research methods have improved, less geo­graphic variation in health care appears to be unexplained, and there is no sound way to attribute the remaining, unexplained variation to any particular cause.

Dartmouth researchers have associated the unexplained portion of geographic variation with the supply of specialist physicians or hospitals, potentially leading to incorrect inferences about the causes of geographic variation—so-called supplier-induced demand. However, recent studies indicate that health sta­tus is a more important factor driving variation in spending than previously believed and that demographic and economic factors, as well as the structure of local health care markets, shape patient preferences and provider practice styles in far more complex ways than early analyses suggested. High spending might reflect inadequately measured health status or some other factor, such as poverty.

Growing evidence suggests that failing to adequately address these complexities may overstate both the extent and implications of geographic variation in health care spending and use.

Some policymakers have used the Dartmouth work to assert that “more is not better.” However, a growing body of research supports the opposite conclu­sion. For example, several recent studies found that patients admitted to higher-intensity, costlier hospitals had better inpatient and post-discharge survival rates. Still other stud­ies indicate greater total spending results in better health sta­tus and survival rates.

These findings raise questions about whether narrowly targeted geographic policies can drive critically important system-wide improvements in efficiency and quality of care.

The paper was authored by Michael Sargen, a Penn medical student, Rod Hooker, an authority on physician assistants and other nonphysician clinicians and a member of The Lewin Group, and Buz Cooper, who has published authoritative projections of the physician workforce over the past 15 years. A pre-publication copy of the paper is available online and a summary follows:

Gaps in the Supply of Physicians, Advance Practice Nurses and Physician Assistants

Based on the goals of health care reform, the demand for physicians will continue to increase. As physician shortages deepen, advanced practice nurses (APNs) and physician assistants (PAs) will play larger roles. Together with physicians they constitute a workforce of “advanced clinicians.” The objective of this study was to assess the capacity of this combined workforce to meet the future demand for clinical services.

Projections were constructed to the year 2025 for the supply of physicians, APNs and PAs, and these were compared with projections of the demand for advanced clinical services, based on federal estimates of future health care spending and historic relationships between spending and the size of the health care labor force.

If training programs for APNs and PAs grow as currently projected but physician residency programs are not further expanded, the aggregate per capita supply of advanced clinicians will remain close to its current level, which will be 20% less than the demand in 2025. Increasing the numbers of entry-level (PGY1) residents by 500 annually will narrow the gap, but it will remain >15%.

The nation faces a substantial shortfall in its combined supply of physicians, APNs, and PAs, even under aggressive training scenarios, and deeper shortages if these scenarios are not achieved. Efforts must be made to expand the output of clinicians in all three disciplines, while also strengthening the infrastructure of clinical practice and facilitating the delegation of tasks to a broadened spectrum of caregivers in new models of care.

In a series of articles in the Contra Costa Times last year, Susanne Bohan and Sandy Kleffman described the striking differences in life expectancy in poor vs. wealthy ZIP codes in East Bay. Life-expectancy in Walnut Creek (94597) was 87.4 years, but it was only 71.2 years in Sobrante Park (94603), where household incomes are about half and poverty >20%. That’s a gap of 16.2 years. We find that, in addition to a shorter life-expectancy in Sobrante, the inpatient hospital utilization rate is double the rate in Walnut Creek. Poverty is not only tragic. It’s expensive.

Bayview/Hunter’s Point is poor area that’s across the Bay in San Francisco. In an article in last month’s New Yorker about Nadine Burke’s clinic for the poor in Bayview/Hunter’s Point, Paul Tough described the community as “a bleak collage of warehouses and one-story public housing projects.” Like Sobrante Park, its poverty rate is >20%, double San Francisco’s average, and hospital utilization in Bayview/Hunter’s Point is double the rate of San Francisco’s wealthy areas, such as Marina and Twin Peaks.

Now let’s look at education. In Sobrante Park and Bayside/Hunter’s Point, where life is short, health care spending high and poverty prevalent, only 40% of adults completed high school and only 5% achieved bachelor’s degrees. In contrast, in Walnut Creek, Marina and Twin Peaks, where lives are long, poverty rare and spending low, >95% completed high school and 40% have bachelor’s degrees. These high-education areas resemble Japan, where high school completion rates are also >95%, as they have been for decades, and where life expectancy is best and health care spending is least.

So I agree with Professor Dow. While other factors contribute to high health care spending, poverty contributes the most, and if the goal is to control health care spending, we must educate children. Of course, they will need more than good schools. They’ll need safe neighborhoods, adequate nutrition, a nurturing environment and more. But without vast improvements in how poor children grow into adults, constraining health care spending will remain a distant dream.

You be the judge. Are areas of the nation with the most chronic disease diagnoses really areas of lowest mortality?  But watch out. Dartmouth statistics are quicker than the eye.

In their new report, MedPAC went a step beyond measuring expenditures, which they adjusted for prices and other factors in their last report, to measuring the actual units of service, a far better way to assess the health care system. MedPAC’s new findings on the distribution of service use in MSAs are graphed below:

Based on this new approach, MedPAC concluded, “although service use varies less than spending, the amount of service provided to beneficiaries still varies substantially. Specifically, service use in higher use areas (90th percentile) is 30 percent greater than in lower use areas (10th percentile); the analogous figure for spending is about 55 percent. What policies should be pursued in light of these findings is beyond the scope of this paper, which is meant only to inform policymakers on the nature and extent of regional variation in Medicare service use. However, we do note that at the extremes, there is nearly a two-fold difference between the MSA with the greatest service use and the MSA with the least.”

As with their previous analysis, the variation seen in their new graph is largely explained poverty and related social circumstances. If the goal is to reduce variation, then patients’ poverty, not physicians’ practices, must be the object of efforts, not that improvements in the way we practice are not warranted. They are, and more must be done, but the metric is not geographic. Poverty is the principle determinant of geographic variation.

The profound contributions of poverty are exemplified by examining two distinct areas of the nation with diametrically opposite characteristics. The first, shown below, is the rural Midwest and west, which extends from Washington and Oregon in the west to Lake Michigan in the east and from the Canadian border to the southern edges of Utah and Nebraska.

This region covers 30% of the land mass of the US but encompasses only 6% of the total population. Its demographics are unusual. Only 0.5% of the population is African-American, compared with 14% elsewhere, and only 10% are Latino, compared with 18% elsewhere.  And while poverty exists in the rural Midwest and west, it exists at a much lower rate and rarely exists in anything resembling a poverty ghetto, as it does in major metropolitan areas. Although it covers a vast area, the rural Midwest-west has only 76 MSAs, and given their favorable demographics, service use in these MSAs was low.

It should be noted that this region became famous for being home to Grand Junction Colorado, designated in the figure above by the small red sun near the bottom of the illustration. President Obama held Grand Junction out as a model of health care because of its low spending, but on closer inspection this much-hyped conclusion proved to be nothing but a grand deception, as I recently discussed on this blog. Grand Junction is like the rest of the rural Midwest-west: white, middle class or more and low disease burden.

A second region of the country presents a very different picture. It is the “southern seven,” shown in red below. I described the characteristics of these states in my previous analysis of MedPAC’s 2009 data. Poverty is the highest in the nation, minorities comprise a larger portion of the population than elsewhere in the nation and the utilization of health care services is the greatest in the nation. While there are questions about fraud and abuse, as well, particularly in Florida, their most distinctive characteristic is poverty, and that is the major cause of high health care utilization.

Striping away the rural Midwest-west and the southern seven from the entire set of MSAs leaves 239, 60% of the total, which display very little geographic variation, as shown in the figure below. In fact, service use is greater than 10% above the mean in only two MSAs.

The picture that emerges is clear and unambiguous. More poverty in the south and less in the rural Midwest and west is the principal factor governing the observed geographic differences in service use. If that is true, there also should be a statistical relationship between the prevalence of poverty in various MSAs and the amount of care utilized by Medicare beneficiaries living there, and, as shown in the figure below, there is.  

Across the full set of MSAs, there is a strong correlation between the amount of health care utilized, as measured by MedPAC, and the percent of seniors who are below the poverty line, as reported by the Census Bureau. Given the large populations of MSAs (averaging almost 1.0 million people) and their diverse physical configurations and social circumstances, it is remarkable that poverty proved to be such a strong correlate of the use of medical services, but that simply reflects the strength of the association between poverty and health care.  

An interesting byproduct of this analysis was the observation that the three MSAs with the very highest levels of poverty and  also very high levels of health care utilization were Laredo, Brownsville and McAllen Texas, which Gawande made famous by claiming that there was no relationship between their health care high utilization and poverty or other patient factors. But of course there is. Just as Dan Gilden said - the reason for high use in McAllen is its high burden of disease, which is a product of its high burden of poverty.

MedPAC, the IOM and countless other organizations are on a quest to explain geographic variation in health care.  Yet the puzzle has been solved, and it is solved again here. Geographic variation in health care is a manifestation of geographic variation in poverty. The logic is obvious to all of us in our every day experiences.  Poverty is associated with more disease, and poor people cope with disease more poorly. And poverty is geographic. So it should not be surprising that health care utilization and spending are geographic.

I believe it is time to stop looking for imaginary causes of geographic variation, or even worse, finding imaginary causes and constructing destructive policy based on them. Rather, it is time to recognize that the major underlying factor is poverty and its myriad social extensions. Effort would be better directed toward trying to understand how to better help poor people to cope better. And more than that, they should be directed to doing what is necessary to decrease the poverty in our society. More than the deficit, more than the war, poverty is our greatest challenge.