Group Health’s Medical Home: Leaving the Poor Out in the Cold

Group Health recently published two papers, one in Health Affairs and the other in JAMA, both extolling the virtues of its Medical Home. These follow a brief report last fall in the NEJM and a lengthy description in the American Journal of Managed Care. In addition, Group Health’s Medical Home has been promoted by the Commonwealth Fund and others, and it is praised in an editorial in the current issue of Lancet. The big news is that costs for patients in their Medical Home were a full 2% lower than in conventional practices, hardly a great success – it wasn’t even statistically significant.  But was even this small difference due to the Medical Home, or was it  because Medical Home patients were less likely to consume care?

Group Health’s answer is that the 7,000 patients in their Medical Home were the same as the 200,000 controls, because the burden of disease, as measured by Diagnostic Cost Groups (DvCGs), was similar in the two groups. But while the DxCG system adjusts for diagnoses, age and sex, it does not adjust for sociodemographic factors, the strongest determinant of utilization. Nor does it appear to have accounted for health status. The chart below, from data in the AJ Managed Care publication, shows just how different these two groups are. Sadly, these data were not included in papers in the NEJM, JAMA or Health Affairs, which are read more widely.

Anyone should be able use fewer resources caring for patients who are more highly educated (and presumably higher income), who are more often white and whose baseline health status is better. Indeed, it’s remarkable that the DxCGs could have been so similar, since health status was so much better among Medical Home patients. What’s most amazing is that this more favorable group consumed only 2% less resources. I would have expected at least 20% less.

But even if Group Health’s model of care were valid, it’s important to recognize the practical limitations in generalizing from it. It took eight physicians to constitute the six FTE physicians who provided Medical Home care, and these physicians had patient panels that were almost 25% smaller per FTE than in Group Health’s usual practices. Nonetheless, Medical Home patients were more frequently referred to specialists (and that was statistically significant). With eight Medical Home physicians providing the same care as four full-time physicians working in the usual practices, it’s not surprisingly that those in the Medical Home had less stress. Patients were more satisfied, too. But there are not 25% more primary care physicians available to allow all of the primary care physicians in America to reduce their panels, particularly with many working part-time. And when there are too few, the poor come in last (see:  No One is Home in the Medical Home).

Beyond these basic concerns, I’m left with two nagging questions. Why, if the Medical Home is patient-centered, did it start with 9,200 patients in 2006, decline to 8,094 by the end of 2007 and fall further to 7,018 by the end of 2009, a loss of 24% of the patients in less than three years? Not too “continuous, comprehensive and coordinated ” for them. But more important in terms of study design, where did these 2,000 patients go? And why? And how much does their care cost? And why aren’t those costs in the final calculations?

And one last question. How can Group Health be a model for the nation when, according to its Service Area Maps, it accepts commercially-insured patients from eighteen counties (top panel) but Medicaid patients from only three (bottom panel)?

If we want high-performance primary care, it will have to be delivered in high-performance systems that use scarce physician resources more efficiently. Panel size will have to be increased, not decreased, as physicians defer more care to others. And physician satisfaction will have to increase not because of less stress but because physicians are rewarded for exercising the complex knowledge that they worked so hard to attain.

Most of all, if we want to decrease health care spending, we will have to recognize that the major remedial costs are associated with the added care that is provided to low-income patients. It’s time to stop talking about wasteful Medical Homes for college grads and start talking about safe neighborhoods, high-quality schools and workable systems of care  for a diverse and needy nation.

The Vagaries of Risk Adjusting for Income – Both Geographically and Individually

Health care utilization is over-estimated in high-income regions that include low-income households. And individual income is over-estimated for low-income individuals living in high-income regions. Both lead to spurious conclusions, the former in studies of geographic variation and the latter in risk adjustment. This is a bit arcane, but it’s critically important, so read on……………

The most powerful characteristic associated with wellness, mortality and health care utilization is income. Low-income patients are sicker, die earlier and consume 2-4 times as much health care as affluent patients. And, because income inequality is greater in the US than in any other developed country, the extremes are greatest in the US.

Aside from its devastating social consequences, these circumstances present a challenge for risk adjustment. While some factors associated with poverty are captured by measuring severity of illness and co-morbidities, others are not. Examples include the effects of a nurturing childhood, adequate education and language skills; the impact of a caring environment with adequate physical resources and available care-givers; access to grocery stores, medications and transportation; safe neighborhoods; racial equality; and myriad other elements of life that impact on wellness, disease and healing. Many are a direct consequence of low income. Others are manifestations of poor neighborhoods – “poverty ghettos.” All correlate generally with income, even though income is not a measure of wealth, particularly in retirement, nor does current income necessarily reflect childhood income, which impacts on lifelong health. Despite these limitations, income is the index of importance.

How, then, to correct for income? Survey studies, such as the Medical Expenditure Panel Survey, include specific information about personal income, but most do not. Some studies of Medicare patients include Social Security income, a poor surrogate for personal income. Most studies lack specific information and rely, instead, on ZIP code income; i.e., the average income of individuals or households within the ZIP code of residence.

There are approximately 43,000 ZIP codes in the US. If the population were dispersed equally, each would have 7,500 people, but ZIP code populations range from fewer than 100 to more than 50,000. Moreover, because ZIP codes are formulated around mail routes rather than sociodemographic characteristics, many include a individuals with range of incomes. The errors that result from the use of ZIP code income are of two types: 1) high-income ZIP codes that include low-income residents are seen as utilizing more health care resources than their average incomes would predict (the geographic error); and 2) low-income residents of ZIP codes that include high income residents are assumed to use fewer resources than their actual incomes would predict (the risk-adjustment error). These two errors confound individual risk adjustment (as in hospital comparisons) and geographic risk adjustment (as in studies of regional variation).

Los Angeles is a case in point. While it possesses a dense poverty core and several affluent zones, it is a patch-quilt of wealth and poverty, often juxtaposed. The Pasadena-Altadena area is one patch in this quilt. A small concentration of poverty overlaps four adjacent ZIP codes, each with more than 20,000 people. In all, this area has 46,000 households with a median income of $60,000, but median incomes among the four ZIP codes range from $36,000 to $110,000, and individual households within them have incomes ranging from $10,000 to more than $1.0M. The utilization of hospital care also ranges widely among these ZIP codes, from 205 days per 1,000 in the wealthiest to 1,045 per 1,000 in the poorest.

How does the mix of affluence and poverty within ZIP codes affect the determinations of health care utilization?  The figure below displays the income distribution in these four ZIP codes. From studies such as the graph above, we know that low-income households (shaded) contribute disproportionately to utilization and that utilization increases steeply at the lowest incomes. In contrast, high-income households contribute much less to utilization but contribute strongly to determining income.

The figure below illustrates the distribution of income in the four ZIP codes and indicates their average levels of hospital utilization. Utilization is influenced most strongly by the low-income households. Conversely, median incomes are influenced most strongly by high-income households. ZIP codes containing such mixtures of affluence and poverty form the “elbow” in the graph at the top. In ZIP 91011, which is predominantly high-income, the smaller proportion of low-income households contributes little to median income but disproportionately to average utilization. Only in ZIP 91003, with its predominance of low-income households, can a valid comparison be made between income and utilization. The consistent error is that utilization appears to be greater relative to income when higher-income ZIP codes include appreciable numbers of poor households. Aggregating and averaging such ZIP codes into hospital referral regions, as occurs in the Dartmouth Atlas, magnifies the error. In this way, affluent regions that include pockets of poverty, as most urban areas do, appear to use excessive resources, and because outcomes among the poor are poor, these resources are seen as “wasted.”

Risk adjustment is affected in the opposite direction. Because of where they live, low-income patients are assumed to have higher incomes than they actually have, sometimes very high, and adjustments that are  based on their assumed incomes fail to fully account for the added risk among low-income patients. This phenomenon is clear in the figure above. Depending on their ZIP code of residence, patients with low incomes will be viewed as wealthy, middle income or poor, and their “adjusted” utilization will be determined accordingly. The result is that, by under-correcting for risk associated with low income, hospitals that care for the poor are seen as “wasteful.”

These vagaries of household incomes within ZIPs cast a shadow over geographic comparisons of health care utilization and outcomes, and they raise serious concerns about risk adjustment applied to individual patients. It is important, therefore, to examine ZIP codes  to determine whether what they report about geographic variation is valid and, if not, to modify the analyses accordingly. No such remedy other than actual knowledge of patients’ incomes exists for risk adjusting individual patient experiences. ZIP code income is simply not a valid estimate. Hospitals that employ such an metric, as most must, will over-estimate  income for poor patients and under-adjust for risk.

Poverty and Health Care: Getting Attention on the Web

Readers interested in the relationships between health care and poverty will want to read several new postings on the Web.

One is an article about my Rhoades Lecture (view slides here) given to  the Medical Society in Detroit. It’s entitled “Poverty and Health Care in America.”

Second is by James Marks, MD, MPH, Vice President of the Robert Wood Johnson Foundation, entitled “The Poor Feel Poorly.” It is posted on the Huffington Post site.

Third is “Health and Health Care in America’s Poorest City,” a tragic and dramatic portrayal of America’s failures to its own in Detroit.

Finally, here is a link to a collection of papers on social inequalities in health by the McArthur Network on SES and Health, published by the New York Academy of Sciences under the title, “Biology of Disadvantage.”

More Than Just Microsoft

A thoughtful physician from Everett WA asked a perceptive question about the redistribution of Medicare payments from “low efficiency” to “high efficiency” areas, for example from South Bronx to Mayo, or, in the example that he gave, from Miami-Dade County to King County WA, the home of Microsoft. “Maybe I am wrong, but I think this was intended to reward high-quality, low-cost care.”  

You’re not wrong. That was the stated purpose. But it simply rewards wealth and penalizes poverty. Costs are lowest where poverty is least, and quality follows. It’s best where poverty is least. High poverty areas (like Miami) have poorer and sicker seniors (see the chart below). The opposite is true for wealthy places with little poverty, like Everett and the surrounding King County.

This is not to say that there’s not waste in Dade and efficiency in King. Both may be true. It’s just that the income effect is so large, it swamps the others.

March Madness – Mayo $400M, the Poor $0

The final House “Manager’s Amendment to Reconcilliation“  provides $400M for hospitals located in counties in the lowest quartile of Medicare spending, adjusted for age, sex and race but not income. Coupled with annual cuts of $10B in DSH and $1.5B for readmissions, this is bad news for the poor and the hospitals that care for them. Mayo wins!   

Note that adjustments cannot be based on counties. Urban counties are too big and economically varied. When the extremes of wealth and poverty are averaged, mean household income is 128% of average in Washington DC, 113% in LA and 108%  in Chicago (Cook County), all with dense and costly poverty ghettos. Without any poverty, mean household income in Olmsted County (home to Mayo) is the same as in LA. Very few truly poor counties will qualify for such payments. This is another example of the truism that “Poverty is the Problem; Wealth is the Solution.”

WSWS: Poverty, Health Care Reform and the Dartmouth Atlas

 Joanne Laurier from the World Socialist Web Site (WSWS) interviewed me recently about Poverty, Health Care Reform and the Dartmouth Atlas. Here’s what I had to say:

“There are basically two problems with the Dartmouth group’s approach. One is methodological and the other is ideological. Although they are quick to point out that they have published 100 papers, these are based on only a few methodologies—and each is flawed. I’ll get into what’s wrong with their methodology later.

But even if they were right, they’re burdened with another problem—ideology. It’s not unusual for policy research to be burdened in this way. In the case of Dartmouth, it’s to an extreme. And, worse, through Peter Orszag, director of the Office of Management and Budget, their ideology has become the cornerstone of health care reform.

It was John Wennberg and his associate, Elliott Fisher, who led Orszag and others to believe that studies of geographic variation prove that doctors and hospitals over-treat and over-charge, to no benefit. And it was they who proposed the 30 percent solution, claiming that the money needed for health care reform was easily available—no new taxes would be required (as President Obama had promised).

If only health care were “more efficient,” the nation could save 30 percent of health care expenditures, $700 billion annually. And to create that “efficiency,” all that was needed was to force all providers to function like the Mayo Clinic (which cares predominantly for white, middle-class patients) and to utilize more primary care physicians (which Mayo doesn’t).

That’s what I call the sin of commission—the tragedy of misleading the process of health care reform. There’s a second sin—the sin of omission, or obfuscation. It’s not simply that the Dartmouth work on geographic differences is methodologically wrong and its conclusions incorrect, nor simply that its policy implications misdirected health care reform. It’s that there is another explanation for the geographic differences, which has to do with differences in the distribution of poverty.

So all the while that they talked about saving money by reducing wasteful geographic variation (by providing less care where it’s actually needed), the fundamental needs of the poor and the large added costs of caring for them were ignored.

It’s actually worse. Poverty was denied, because it couldn’t be both ways. Either the Dartmouth group was right and the high costs in some areas were because of too many specialists and hospitals doing too many unneeded things, or this higher spending was due to the added costs of caring for the poor. The truth is that it is the latter.

Therefore, the only way to really save money is to make a long-term commitment to ameliorating the high health care costs that are a result of poverty and other social determinants of disease. Not that there aren’t inefficiencies. But physicians have been dealing with inefficiencies as long as I’ve been a doctor—which is 50 years—and certainly before that.

As medicine evolves, there are always more inefficiencies to deal with, but as fast as we deal with them, new ones emerge. So constant diligence is necessary. But is medicine more efficient than in 1960? You bet it is. And is poverty a bigger problem for health care spending now than it was then? You bet. We seem to know how to make things more efficient. But as a nation, we aren’t very good at reigning in poverty. It just grows.”

 Read more on the WSWS Web page.

The Demise of Dartmouth, but Who’s To Blame?

An important article appeared in today’s NYT, describing a new paper by Peter Bach, which is in today’s NEJM. Peter’s paper (“A Map to Bad Policy“) debunks the Dartmouth Atlas and cautions against its use. As I said in the Wash Post in September, the Dartmouth Atlas is the ”Wrong Map for Health Care Reform.”

More damning even than Peter’s analysis was Elliott Fisher’s reply: “Dr. Fisher agreed that the current Atlas measures should not be used to set hospital payment rates, and that looking at the care of patients at the end of life provides only limited insight into the quality of care provided to those patients. He said he and his colleagues should not be held responsible for the misinterpretation of their data.” Really? It was someone else’s interpretation? OK, Elliott, you’re not responsible. Just stand in the corner.

Peter is not the only leading epidemiologist to debunk Dartmouth in recent days. There’s also the report this week from the U of Wisconsin and RWJ by Pat Remington (another leader), showing that people who have the poorest health (and, therefore, the highest health care costs) live in the poorest counties (see my blog report and an earlier discussion of poverty and health care). And there’s the recent paper by Ong and Rosenthal (co-authored by Jose Escarce, editor of HSR, the leading health services research journal), showing that, when all care is measured (not simply end-of-life care, as measured by Dartmouth), hospitals that provide more have lower mortality, which was confirmed in the current issue of Medical Care by Barnato and associates at the U of Pittsburgh. When it rains, it pours.

What’s doubly important about the death of the Dartmouth Atlas is that it was the cornerstone of health care reform. Right from the start, Peter Orszag, director of OMB and the administration’s architect of health care reform, accepted Dartmouth’s ideological principles that health care spending was driven by doctors and hospitals who over-treated and over-charged, to no benefit. The funds for health care reform were readily available by simply getting rid of geographic differences. That alone would save 30% of health care spending ($700B). And that could be accomplished by making everything look like Mayo (white, middle class and efficient) and by having more primary care physicians (which Mayo doesn’t). And best of all, it could assure that no new taxes would be needed, just as President Obama had promised. 

The problem is that it didn’t make sense. Voters knew it, even if they didn’t know the methodological details. And the CBO figured out. And congressmen had to scramble to find ways to pay for health care reform without actually paying for it, because it was supposed to be for free – the 30% solution said so, and folks all over Capitol Hill cited it.  And now the spiral of hypocrisy has finally unraveled. Like Madoff’s investments, the Dartmouth Atlas was shadows and mirrors. But this time, the price tag is more than the $50B that Madoff cost. It’s the likely loss of health care reform. But don’t worry, Elliott. We won’t blame you.

County Health and Poverty

A new study from the University of Wisconsin and RWJ identifies the five counties in each state with the poorest health (also the five healthiest). According to Marketplace, the research concludes that economic status may be even more important than access to care. The map from this study is reproduced below. Beneath it is a map showing counties with the most (and the least) poverty. Circles identify counties that appear on both maps. There is a remarkable degree of overlap between the very poorest counties and the sickest counties. Note a similar relationship between the wealthiest and the healthiest.

The lesson is that poverty is geographic, poor health tracks poverty and expenditures track poor health. If we’re going to do health care reform again, let’s get it right. No more ”30% solution” this time. But a lot of attention to the links between poverty, health status and health care spending.

House Members Urge Pelosi to Limit DSH Payment Cuts

In a letter that was organized by Reps. Reyes (D-TX), Lewis (D-GA) and Schakowsky (D-IL), 104 members of the House urged Speaker Pelosi to ensure that the final health reform bill does not lower Medicaid and Medicare Disproportionate Share Hospital (DSH) payments more than in the House-passed bill, which already is too much. “To retain our health care safety net’s stability, we believe that future DSH payments must continue to recognize financial losses sustained by these providers due to Medicaid reimbursement shortfalls and uncompensated care,” they wrote. The House bill would cut DSH payments by $20 billion over 10 years, while the Senate bill would cut $43 billion, approximately 25% of what otherwise would be paid. The rationale for lowering DSH payments is that more patients will be insured. But half or more of them will have incomes low enough to qualify for Medicaid, even at the Senate’s eligibility level of 133% of poverty (the House bill calls for 150%). And it’s because the costs of caring for low-income patients are so high that DSH now subsidizes hospitals that provide a lot of it. So the last thing anyone should want to do is cut DSH as more poor patients are brought into the system.

Who would be most hurt? The map below shows the amount of DSH paid by Medicare on behalf of poor patients in the various counties, expressed as DOLLARS per POOR POPULATION (individuals below the poverty level). Note: orange is the most, dark blue is next, pale blue is least. Areas with the densest concentrations of poverty, such as in the south, northeast and California, receive the most, and they will lose the most. The northwest and upper-Midwest win again. Of course, the Dartmouth Atlas tells us that these latter regions are “efficient” and deserve to be rewarded. That’s malarkey! Paying for health care reform on the backs of the poor is poor social policy.

Map by Matthew Cooper

Deficits, Jobs and Health Care

Commenting on the President’s budget, an editorial in the Times on Feb 2nd juxtaposed three of our nation’s dilemmas: the deficit, jobs and health care.

“President Obama got his priorities mostly right. The deficit, compared with what it could have been, is $120B. That’s a lot of money. But it’s not too much at a time of economic weakness, when deficit spending is needed to put Americans back to work.”

“Medicare and Medicaid will cost $788B; that should be another reminder of why the country needs health care reform.”

The fundamental question about health care spending is, therefore, what does it mean for jobs?  Approximately 15 million people work in health care, and that doesn’t count jobs at the 140 companies that specialize in constructing health care facilities or 56,000 pharmacies or the dreaded health insurance companies, nor does it include all of the jobs of people supplying goods and services to the 18 million folks engaged in health care in these various ways.  But a more important question is, where is the job growth? The answer is health care. Over the past decade, the growth in health care jobs has equaled the total growth of jobs. Many are high-skilled, but many are entry-level jobs that help to move people out of poverty. So we had better be careful in measuring the impact of health care. Quite apart from its beneficial effects on well being, it just may be the engine of the economy.