Objective: To determine whether a given doctor treating Medicaid patients is likely to practice in a predominantly minority area, and whether a minority patient is likely to be treated by a physician who is heavily influenced by Medicaid policy decisions.
Study Design: Retrospective pharmacy claims database analysis combined with zip-5-level demographic analysis.
Methods: Data extracted from a large prescription claims database were used to categorize all active prescribers in the United States by the proportion of prescription claims paid by state Medicaid programs from May 31, 2003, to April 30, 2004. US Census data from 2000 were used to assess the ethnic composition of each physician's zip code. Descriptive analyses were conducted to explore any associations between zip code racial composition and proportion of prescriber prescription claims adjudicated by state Medicaid programs.
Results: Physicians with more than 75% of their prescriptions adjudicated through Medicaid versus those with fewer than 1% of their prescriptions adjudicated through Medicaid practiced in zip codes that were 47% versus 24% nonwhite, respectively. Residents in Medicaid-intense zip codes were 59% nonwhite versus 31% nonwhite in the nation as a whole.
Conclusion: Nonwhite residents are much more likely than white residents to live in a zip code where Medicaid prescribing rules will affect their physician. Any legislation-induced changes in prescribing patterns seem likely to disproportionately impact both Medicaid and non-Medicaid minority residents in these areas.
(Am J Manag Care. 2005;11:SP21-SP26)
The Medicaid program is the largest source of funding for medical and health-related services for America's poorest people.1 By paying providers for services, Medicaid has reduced financial barriers to medical and health-related services for many with low income and low wealth. This has almost certainly improved the health and well being of those who lack resources and are at risk of having relatively poor health. Because African Americans and Hispanic Americans constitute disproportionately large shares of this vulnerable population, there can be little doubt that Medicaid has been a force to reduce racial and ethnic disparities in access to insurance coverage. Racine et al found that expansions in eligibility for Medicaid between 1989 and 1995 produced greater reductions in uninsured rates among poor minority children than poor white children,2 and an Urban Institute study found similar gains between 1997 and 2002 for Medicaid and a related program, the State Children's Health Insurance Program.3 It seems clear that without Medicaid, any disparities in healthcare access or health outcomes between minorities and others would probably be much worse.
Despite much progress in closing the health insurance coverage gap, minorities continue to experience significant disadvantages in health outcomes.4 Coverage is an important ingredient in solving the problem, but closing the outcomes disparity is complicated by structural details of that coverage as well as a host of economic, genetic, and cultural factors associated with patients and providers. A body of literature has found that minorities sometimes receive different treatment than others even when they have the same insurance coverage and see the same physicians. For example, minority utilization of prescription drugs tends to be lower even when disease and insurance coverage are controlled for.5
Some insight into this dilemma may be obtained by examining how Medicaid policy may affect minority patients. Each state Medicaid agency creates contracts with physicians, pharmacies, hospitals, and other healthcare service providers. These contracts sometimes set rates and requirements for provider payment, as well as limits on the type, amount, duration, and scope of services. The ability to limit services provides a tool to constrain costs, but at the same time also may have implications for health outcomes. One area where policymakers have spent significant energy recently is prescription drug benefit design. As of the end of 2003, at least 29 states had obtained legislative approval to implement a preferred drug list (PDL) with expanded prior-authorization requirements for nonpreferred drugs.6 Such programs attempt to save money on prescription drugs by establishing lists of "preferred"medicines based on a set of criteria, including price, and creating obstacles when nonpreferred medicines are prescribed. According to a 2003 Kaiser Family Foundation Report, there is substantial variation in who sets the policy and criteria for inclusion on the list.7
Recent evidence indicates that the task of crafting Medicaid policies that provide an optimal set of incentives for providers is complicated by the racial and ethnic composition of a state's residents. Burroughs et al present a medical rationale and clinical evidence that suggest racial differences in benefits from specific pharmaceutical treatments for specific diagnoses.8 They conclude that "[t]here is good evidence to show that therapeutic substitution of drugs within the same class places minority patients at greater risk?"8,p2 Soumerai observes that "[t]argeting essential drug classes with heterogeneous patient responses and side effects could reduce appropriate care, adversely affect health status, and cause shifts to more costly types of care."9,p135 These studies suggest that the optimal health production function (as defined by Grossman) differs by race and ethnicity.10
Given the sheer numbers, even small changes in policy could have a disproportionately large impact on minority patients. As calculated from data downloaded from www.statehealthfacts.kff.org and illustrated in Table 1, 25% of all nonelderly Medicaid recipients are African American, and 23% are Hispanic (vs 12% and 11% for the general population).11 Table 1 also includes information about whether a state has put a PDL into place.
Based on the enrollment and race data for the states that have implemented PDLs, we calculate that among the nonelderly population, 61% of white Medicaid enrollees, 65% of black Medicaid enrollees, and 68% of Hispanic Medicaid enrollees are now in programs with PDLs in place.
The underlying racial composition of Medicaid is more skewed among children than among the general Medicaid population. According to the Urban Institute, 44% of African American children and 37% of Hispanic children are covered by Medicaid. In contrast, fewer than 8% of all non-Hispanic whites (and about 15% of white children) are covered by Medicaid.3
The high degree of racial and ethnic heterogeneity across states within the Medicaid program, and especially within states that have adopted a PDL, begs the question of whether such states have incorporated relevant racial and ethnic information into the development of and adjustments to their PDLs. Most of the changes implemented in state Medicaid programs have been adopted within just the past few years, and the evidence is only now beginning to become available to evaluate the impact such changes have had on patient and physician behavior and health. One fundamental question that must be answered is whether and to what extent Medicaid-induced changes in physician prescription-writing behavior impact racial and ethnic health disparities. Designing an optimal policy would be a less complicated task for a state with a Medicaid population that is relatively homogeneous with respect to race and ethnicity than it would be for a state with a Medicaid population that is heterogeneous with respect to race and ethnicity. In the latter case, a set of incentives that did not recognize those differences may exacerbate rather than reduce health disparities.
Although changes to Medicaid prescription drug coverage rules may impact patients, physicians, and pharmacists through different mechanisms, for this analysis we focus on physicians. Our assumption is that attempts by policymakers to influence physician prescribing behavior are successful in changing behavior, at least to some degree. To the extent that racial and ethnic minorities live disproportionately in areas where physician practice patterns are affected by Medicaid, and to the extent that those changes impact the practice generally (not just Medicaid patients), changes in policy will have a disproportionate impact on minorities. There is good reason to expect the effects to permeate throughout the community; a recent analysis by researchers at Dartmouth College suggests that a large part of the overall problem of health disparities in the United States may be the result of regional differences in treatment and outcomes.12 Additionally, researchers at the Center for Studying Health System Change and Memorial Sloan-Kettering Cancer Center have found that black and white patients are largely treated by different physicians; 80% of the primary care visits for blacks receiving Medicare are handled by just 22% of all physicians.13 That study also found that the physicians treating black patients are more likely to face obstacles in getting their patients access to high-quality services.
The present study was designed to test the following hypotheses:
Physicians who are most affected by Medicaid prescribing regulations practice in neighborhoods that are disproportionately comprised of racial and ethnic minority residents.
We obtained data from NDCHealth (Atlanta, Ga) for prescriptions filled from May 2003 through April 2004. The data contain codes that include the zip code of the location of the prescribing physician's practice as well as insurance coverage/source of payment information for the prescriptions filled. These data are collected from pharmacies throughout the country and include detailed information on more than 70% of the prescriptions filled in the United States. We included the top 90% of prescribing physicians in the sample, to exclude those who stopped practicing midyear or who do not generally prescribe medicines to their patients. We used these data to analyze the aggregate demographics and source of payment (Medicaid, cash, or third party) data within the 5-digit zip code where each physician practices, and then segmented doctors into categories based on the proportion of prescriptions they wrote that were covered by Medicaid. In cases where 5-digit zip codes were very lightly populated (fewer than 5000 residents), we analyzed the data for the 3-digit zip code of the physician's practice instead of the 5-digit zip code. All areas of the United States were included in the analysis, yielding 299 164 physicians practicing in 12 177 distinct geographic areas.
We combined the data with 2000 US Census data at the zip-code level, which provide racial and ethnic data. This allowed us to examine the demographic features of zip codes where many or most physicians treat a significant portion of Medicaid patients. We categorized zip codes by the percentage of residents who identified themselves as African American, white, Hispanic, and other. The Census Bureau allows residents to choose both "Hispanic" and "Caucasian" or "Hispanic" and "African American."For our analysis we classified people into 4 mutually exclusive categories: non-Hispanic white, Hispanic, non-Hispanic African American, and others. For convenience, we abbreviated these categories as white, Hispanic, African American, and other. Only 4.5% of Hispanics in the sample were identified as African American. In our analysis we used the term "nonwhite" to include Hispanics of any race, African Americans, and others who did not identify themselves as Caucasian.
To determine which zip codes were most likely to be impacted by Medicaid policy changes, we segmented physicians based on the proportion of their prescriptions that are paid for via Medicaid. We then organized those physicians by 5-digit zip codes and determined the proportion of physicians within each zip code whose practices would be considered Medicaid intense, which we defined as having more than 50% of their prescriptions paid for by Medicaid. We also examined the same data using a looser definition of "Medicaid intense" for physicians who have at least 25% of their prescriptions paid for by Medicaid. All of the analyses are simple means, calculated and aggregated at the zip-code level based on a combination of the available data sources.
Table 2 presents the basic results of this analysis for primary care physicians; the data for all physicians show similar trends. We found that physicians who prescribed to a high proportion of Medicaid patients (the bottom few rows of the table) tended to practice in areas with a higher nonwhite population than other physicians.
Practices where 75% to 100% of prescriptions were paid for by Medicaid were located in neighborhoods where 47% of the population was nonwhite (18% African American, 22% Hispanic, and 7% other). Practices where 50% to 75% of prescriptions were paid for by Medicaid were located in neighborhoods where 42% of the population was nonwhite (18% African American, 18% Hispanic, 7% other). In contrast, the average area had a population that was 29% nonwhite (12% African American, 11% Hispanic, 6% other).
The Figure presents the results of our zip-code-level physician practice analysis. Within each zip code, we estimated the share of physician practices that we considered Medicaid intense, which we defined as having more than 50% of their prescriptions paid for by Medicaid. We then examined the demographic characteristics of Medicaid-intense zip codes.
In most zip codes, fewer than 10% of physician practices were Medicaid intense. The demographics in Medicaid-light (fewer than 10% of physicians) zip codes matched up fairly well with the national average: about 25% of residents of such areas were nonwhite.
However, the patients who lived in areas served by Medicaid-intense physicians were disproportionately nonwhite: 59% of all people living in the most Medicaid-intensive zip codes were nonwhite (26% African Americans).
We also examined how likely an individual nonwhite person was to live in a Medicaid-intense zip code. Using the 5-digit zip code again as a proxy for a neighborhood, we characterized neighborhoods by the percentage of physicians with a high share of Medicaid patients. For this analysis, we used 2 definitions of Medicaid-intense practice: one where physician practices were Medicaid intense if at least 50% of their prescriptions were paid for by Medicaid, and a less strict version where the practices were Medicaid intense if at least 25% of their prescriptions were paid for by Medicaid. A Medicaid-heavy zip code was defined as one where at least 20% of the physician practices were Medicaid intense under 1 of the 2 definitions. We found that nonwhites were much more likely than whites to live in a Medicaid-intense zip code under both the stricter definition (24% vs 9%) and the looser one (59% vs 35%).
DISCUSSION AND LIMITATIONS
Two key findings resulted from the analysis. First, a physician who writes a significant share of Medicaid prescriptions is likely to be located in a zip code with relatively high concentrations of racial and or ethnic minority residents. Second, market areas that are characterized by higher concentrations of racial and or minority residents also have proportionately more physician practices that are Medicaid intense. The size of the latter effect is substantial–under the looser definition of "Medicaid intense," nearly 60% of all nonwhites live in zip codes where a significant number of physicians are likely to be affected by Medicaid policy decisions. Thus, to the extent that Medicaid affects the poor as it was designed to do (with respect to medications, by affecting physician prescribing practices), it is likely to have a disproportionately large effect on racial and ethnic minority enrollees. Other work has demonstrated that formulary changes tend to spill over from Medicaid patients to other patients when physicians treat significant numbers of Medicaid patients,14 suggesting that because minorities are highly likely to be treated by physicians with a significant Medicaid business, any impact that Medicaid policies have will reverberate through the minority community.
Changes in rules for accessing prescription drugs in Medicaid are not intentionally designed to discriminate against minorities. However, it is plausible that such rules have disproportionately adverse consequences for minorities because they are a relatively large segment of the enrollee population. There are at least 2 other reasons why rule changes might affect minorities differently than other groups. First, minorities may respond differently as a group than nonminorities to specific medicines, and if those medicines are subject to access restrictions, minorities may be impacted disproportionately. Second, minority patients may respond differently to different rule changes based on cultural factors. For example, Burroughs states that "[p]atients' beliefs regarding the properties and effects of medications are of central importance in determining compliance"with prescribed treatment regimens.15,p3
Finding the optimal treatment regimen for any patient is complicated, and treatments must be tailored to individual needs. Several recent studies have highlighted some of the genetic differences that play a role in how well medicines work in different kinds of people. One study found that genetic differences among minorities "may influence a drug's action by altering its absorption, distribution, metabolism, excretion,"or overall effect on the body, which can help explain why some medicines are more effective in some patients than others. These differences may increase or decrease the intensity and duration of the expected drug effect and "substantial dosage adjustments may be necessary."8
The National Medical Association study by Burroughs et al found that genetic polymorphisms affect the metabolism of different classes of medicines (including antiarrhythmics, antidepressants, beta blockers, and isoniazid).8 According to the article, polymorphisms can be major determinants of drug efficacy. This means that they may influence how well a drug works for individual patients by altering the drug's pharmacokinetics (ie, metabolism, absorption) or pharmacodynamics (effect on the body). Clinically, these variations mean that doctors would need substantial latitude in choosing the right drugs and dosages for individual patients from various populations to achieve optimal outcomes. Also of note is that different drugs of the same class often are cleared through different metabolic pathways and drugs of a certain class may vary in their susceptibility to genetic differences in metabolism. Additionally, the pathophysiology of different disease states (ie, hypertension) differs among racial groups, and some drugs will be more effective in one racial/ethnic group than in another.
To our knowledge, such considerations are rarely factored into calculations to determine whether a medicine will make a state's PDL. The current study is merely descriptive in nature and therefore is limited in its ability to assess and understand the underlying mechanisms by which different kinds of Medicaid rules actually impact prescribing behavior. Future research should consider whether differences in program designs have different impacts on minority groups and should examine in greater detail the propensity for physicians to be influenced by Medicaid prescribing rules.
To be sure, this analysis simply points to the need for a more thorough, multivariate framework that would capture details about the individual patients and their characteristics alongside the practice and payment-mix characteristics of the physicians we studied. Such a framework could include a sophisticated model of drug choice by physicians that would allow for much more refined estimates of the interplay between payer mix, legislative decision making, racial and ethnic factors, and physician choice of treatment. Such a model is beyond the scope of the current project but would be highly valuable as policymakers consider the potential consequences of their decisions on potentially vulnerable groups.
Over the past few years, many states have moved toward an aggressive cost-control approach to prescription drug access in Medicaid. The new approach is fueled by a combination of budget pressures and a sense that the pharmaceutical spending is controllable relative to other parts of the budget without hampering quality. Such policies save money in the pharmaceutical budget in the short run by impacting physician behavior, specifically by penalizing physicians for prescribing medicines not on a preferred list or by creating obstacles (prior-authorization programs and other paperwork) to discourage use of certain medicines. However, such policies may come with several types of important hidden costs in terms of both dollars and patient care.
There is ample research to suggest that underuse of medications–whatever the cause–can raise costs elsewhere in the system by causing unneeded hospitalizations, physician visits, and other services.16 Studies also have found that newer medicines–precisely the type that are usually targeted for savings–tend to bring offsetting reductions in other medical costs compared with older medicines.17 These types of hidden costs eventually will appear in Medicaid budgets, and it seems likely that analysts will be able to uncover the costs of restricted access as they become visible.
The negative impact of Medicaid access restrictions on health disparities, however, is an area with high potential to go unnoticed because it may be eclipsed by the overall good that Medicaid does to reduce disparities. Policymakers may not realize the extent to which their decisions on Medicaid affect both the disproportionate numbers of minorities in Medicaid and, indirectly, the large portion of minorities who live in neighborhoods served by physicians with a large proportion of Medicaid patients. To the extent that changes in prescription drug policy have an impact on enrollees generally, our findings suggest that the impact on minorities will be amplified considerable by demographic, physical, and cultural factors.