Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 215
7
Data Collection and Monitoring
The preceding chapters illustrate the complexity and variety of fac-
tors—including healthcare financing arrangements, institutional and or-
ganizational characteristics of healthcare settings, aspects of the clinical
encounter, and the attitudes, perceptions, and beliefs of healthcare pro-
viders and their patients—that influence healthcare disparities. The com-
plexity of these factors, coupled with the fact that disparities in care are
not always apparent to patients or providers in clinical encounters, in-
creases the need for data to better understand the extent of disparities and
the circumstances under which disparities are likely to occur. Unfortu-
nately, standardized data on racial and ethnic differences in care are gen-
erally unavailable. Federal, private, and state-supported data collection
efforts are scattered and unsystematic, and many health plans, with a few
notable exceptions, do not collect data on enrollees’ race, ethnicity, or pri-
mary language, pointing to significant obstacles to the collection and
analysis of such data (Perot and Youdelman, 2001).
Standardized data collection, however, is critically important in the
effort to understand and eliminate racial and ethnic disparities in health-
care. Having data on patient and provider race and ethnicity would al-
low researchers to better disentangle factors that are associated with
healthcare disparities. In addition, collecting appropriate data related to
racial or ethnic differences in the process, structure and outcomes of care
can help to identify discriminatory practices, whether they are the result
of intentional behaviors and attitudes, or unintended—but no less harm-
ful—biases or policies that result in racial or ethnic differences in care that
cannot be justified by patient preferences or clinical need. Data collection
215
OCR for page 216
216 UNEQUAL TREATMENT
and monitoring therefore provides critically needed information for civil
rights enforcement. Further, collecting and analyzing patterns of care by
patient race, ethnicity, and other demographic data can help health plans
to monitor plan performance. Such monitoring can help to ensure ac-
countability to enrolled members and payors, improve patient choice, and
allow for evaluation of intervention programs. Such evaluations are likely
to improve service delivery for racial and ethnic minority populations,
and therefore may result in cost savings that would offset the costs of data
collection.
The collection of racial and ethnic data in health systems poses special
challenges, however. Traditionally, the practice of healthcare has been
dominated by individual practitioners who delivered care in settings rela-
tively unaffected by regulation, oversight, or government intervention.
Hospitals enjoyed little external monitoring, and their professionally
dominated and autonomous organizational structure was rarely chal-
lenged prior to the emergence of the federal government as the largest
healthcare payor. Today’s cost-conscious healthcare systems present an
opportunity for greater healthcare practice accountability, but medicine’s
traditional autonomy and self-government presents little history of over-
sight, particularly with regard to civil rights, that can be expanded upon
(Smith, 1998).
Specific recommendations regarding the types of healthcare data that
should be collected, and how this information should be analyzed and
reported has been the subject of intensive study and debate by govern-
mental (U.S. DHHS, 1999) and private groups (National Quality Forum,
2001; Perot and Youdelman, 2001), and is beyond the scope of this report.
Selecting indicators of healthcare disparities that can be readily measured,
analyzed and reported, and developing methods to ensure reliable data
collection will require careful consideration of costs, benefits, and other
potential problems inherent in collecting and reporting patient care data
(see discussion of obstacles to racial/ethnic data collection, below). These
issues will be weighed by a forthcoming National Academies study com-
mittee that has been asked by Congress to assess the adequacy of racial
and ethnic data within U.S. Department of Health and Human Services
(DHHS) systems. Ideally, however, all patient encounters should be as-
sessed for the quality of care and patient outcomes. This would enable
the data to be aggregated to many different levels of the healthcare deliv-
ery system, including health plans, medical groups, and hospitals. Most
of the information collected should be recorded as part of the patient’s
medical record, a task that in the future will be assisted greatly by the
development of electronic patient records. These data should be stratified
by race, ethnicity, as well as socioeconomic status and, where possible,
primary language.
OCR for page 217
217
DATA COLLECTION AND MONITORING
OBSTACLES TO RACIAL/ETHNIC DATA COLLECTION
The need for data on patients’ race and ethnicity and quality of care
must be balanced against other significant considerations. Foremost, pa-
tient privacy must be protected. The confidentiality and security of pa-
tient information and data transactions must, at minimum, conform with
standards set forth in the Health Insurance Portability and Accountability
Act of 1996 (HIPAA). Secondly, the costs of data collection must be
weighed relative to its benefits. When and how such data are collected
will have broad cost implications; collection of patient race and ethnicity
data at the point of plan enrollment, for example, will likely be less expen-
sive than data collection among members already enrolled in plans whose
race or ethnicity is unknown. Similarly, administrative and paperwork
burdens are likely to increase as the numbers of patient data elements are
increased. Formal Congressional checks on such administrative burdens
(e.g., the Paperwork Reduction Act) require that administrators of pub-
licly-funded programs assess such costs and demonstrate the utility of
additional data collection relative to costs.
Other legal constraints must be assessed, as well. While the vast ma-
jority of states do not prohibit collection of patients’ race and ethnicity
data, some may impose restraints on when and how such data may be
collected (Perez and Satcher, 2001). The extent of these restraints must be
assessed and this information provided to managed care organizations
(MCOs) and payors to avoid confusion over what kinds of data collection
are allowed, and under what circumstances.
Political concerns must be also addressed to ensure cooperation from
all parties in data collection efforts. Resistance to data collection efforts
may come from healthcare providers, institutions, plans, and patients,
unless the purposes and benefits of data collection are clearly explicated.
Providers, as noted earlier, may resent perceived intrusions on autonomy.
Patients, particularly minority patients, may worry that racial or ethnic
data collection will result in “redlining” of services, selection of enrollees,
or rationing of services on the basis of race or ethnicity.
Efforts to enforce data collection from the federal level may also meet
resistance from state authorities, who retain primary responsibility for
determining data requirements of health plans with whom states contract
for Medicaid MCO services. Federal efforts to require the collection of
patients’ racial and ethnic data would raise challenges from those who
find federal reporting requirements already burdensome and the federal
role in dictating the terms of managed care contracts too extensive. Fi-
nally, it should be noted that some individuals are broadly opposed to
government involvement in monitoring race and ethnic trends among the
U.S. population, and are mounting challenges to the notion that the gov-
OCR for page 218
218 UNEQUAL TREATMENT
ernment should collect any information about race or ethnicity. Ward
Connerly, for example, the California businessman who led efforts to re-
peal affirmative action in that state, is spearheading a ballot initiative to
prevent the state from collecting any information about race or ethnicity,
except for a few limited circumstances (Jordan, 2001). This initiative
would likely undercut efforts to assess racial and ethnic inequities in
healthcare, as well as in other potentially discriminatory practices.
In addition, health plans have raised significant concerns regarding
the collection of patient race and ethnicity information. Many plans, led
by American Association of Health Plans (AAHP), increasingly see the
collection of information on patient race and ethnicity as an important
means to evaluate their own efforts to reduce disparities in care and de-
velop better strategies to serve growing minority patient populations
(Ignani and Bocchino, communication with Alan Nelson, M.D., March 19,
2001). However, some plans have operated under the erroneous assump-
tion that federal and/or state law prohibits the collection of patient race
and ethnicity information. Efforts by the U.S. Department of Health and
Human Services Office for Civil Rights (OCR) and Office of Minority
Health (OMH) to clarify federal law (Perez, 2000; Perez and Satcher, 2001)
have helped to dismiss this assumption.
Many health plans, however, remain concerned that their ability to
serve minority patients could be hampered should data collection efforts
be seen by these populations as an effort to ration care. In addition, plans
that serve disproportionately minority and lower-income populations
could be hurt by the release of “report card” information that reveals their
enrolled members to be less healthy or to require more services than the
majority population. In such instances, information about the health sta-
tus of plans’ enrolled populations and case-mix may largely reflect con-
ditions of poverty and the generally higher incidence of morbidity and
mortality among lower-income and minority populations, and may not
necessarily reflect poor service on the part of health plans. This kind of
information might unfairly hurt health plans’ efforts to expand their mar-
ket share among minority populations, and should be taken into account
(Fiscella et al., 2000).
Other challenges include the accuracy of racial and ethnic data. As
noted earlier, “race” and “ethnicity” are fluid, socially defined concepts
that are not consistently understood or applied in data collection efforts.
Racial or ethnic identity is determined by multiple factors and may vary
depending on the contexts in which these constructs are defined and the
manner in which data are collected. Observers recording race and
ethnicity data are notoriously inaccurate, particularly with regard to His-
panic or American-Indian populations (e.g., death certificates commonly
misreport the race of American Indians). Further, a small but increasing
OCR for page 219
219
DATA COLLECTION AND MONITORING
proportion of individuals define themselves using two or more racial and
ethnic categories, making simple classification difficult. Finally, efforts to
address disparities in care must acknowledge the significant heterogene-
ity within each of the federally defined racial and ethnic groups (whites,
African Americans, Native Americans, Asian Americans, Pacific Island-
ers, and Hispanics). Wide variations within each of these groups can be
found in health status, health practices and behaviors, and healthcare re-
sources. It is therefore important that data be collected on subgroups
within these categories (e.g., Cuban American, Puerto Rican, Mexican
American, Central American among the “Hispanic” ethnic group). Where
possible and appropriate, data collected over several years can be com-
bined to achieve sufficient analytic sample sizes (U.S. DHHS National
Committee on Vital and Health Statistics, 1999).
These challenges underscore the need for consensus among health
plans, providers, and consumers regarding data collection policies, and
best practices regarding how data will be analyzed and to whom it will be
presented. To this end, the committee believes that efforts by public and
private groups, such as the National Quality Forum (NQF), the National
Committee on Vital and Health Statistics (NCVHS), and the Agency for
Healthcare Research and Quality (AHRQ), to convene experts and pro-
vide specific recommendations regarding the collection and analysis of
data on patients’ race and ethnicity will prove fruitful to help achieve
broad consensus on best policies and practices. Development of a full,
national database of healthcare quality that can be analyzed by race and
ethnicity will take time, however, and it is clear that a sequence of steps
must be undertaken to reach this goal. An important first step would
involve an assessment of existing data sets within public and private plans
that allows for an analysis of patient care by race and ethnicity.
THE FEDERAL ROLE IN RACIAL, ETHNIC, AND
PRIMARY LANGUAGE HEALTH DATA
Several agencies of the DHHS, recognizing the importance of racial,
ethnic, and primary language healthcare data, have attempted to promote
data collection and monitoring efforts, particularly to address the chal-
lenges noted above. Despite these efforts, federal data collection remains
unsystematic and lacks an overall guiding structure to ensure account-
ability and cooperation by HHS agencies, states, and private sector part-
ners involved with federal health programs (Perot and Youdelman, 2001).
The Summit Health Institute for Research and Education, Inc. (SHIRE)
and the National Health Law Program (NHeLP), with support from The
Commonwealth Fund, analyzed an array of statutes, regulations, federal
OCR for page 220
220 UNEQUAL TREATMENT
agency policies, practices, and data collection vehicles related to race,
ethnicity, and primary language in healthcare settings. This analysis in-
cluded an assessment of the extent to which federal policies mandate or
encourage collection and reporting of race, ethnicity, and primary lan-
guage data and an assessment of how current law is understood, inter-
preted, and implemented by federal officials. SHIRE and NHeLP ana-
lyzed 80 program-specific statutes and over 100 data collection vehicles,
and developed 25 findings and 10 recommendations regarding federal
data policies (Perot and Youdelman, 2001). These recommendations are
listed in Box 7-1.
BOX 7-1
Recommendations, Racial, Ethnic, and Primary Language Data
Collection in the Healthcare System: An Assessment of Federal
Policies and Practices (Perot and Youdelman, 2001)
1. Ensure that Medicare data, as well as other data regarding individuals
who are served by HHS programs or who participate in HHS research
activities, are readily available and accurate by race, ethnicity, and
primary language. Independent analysts have estimated that the Medi-
care beneficiary eligibility file compiled by the Social Security Admin-
istration is less than 60 percent accurate for all racial/ethnic classifica-
tions other than black or white.
2. Enforce state collection and reporting of data by race, ethnicity, and
primary language for enrollees in Medicaid and the State Children’s
Health Insurance Program (SCHIP). Currently, data collection and
reporting by states are often inconsistent and incomplete.
3. Revise the standards for implementation of the Health Insurance Port-
ability and Accountability Act (HIPAA) to designate the code set for
race and ethnicity data as mandatory for both claims and enrollment
standards. Racial and ethnic categories used under HIPAA must be
compliant with OMB standards.
4. Recommend that quality measurement and reporting tools such as the
Health Plan Employer Data and Information Set (HEDIS) should col-
lect and report health data by race, ethnicity, and primary language.
5. Ensure access to quality healthcare for people with limited English
proficiency by effective monitoring of adherence to guidelines and
collection of requisite data.
6. Include statutory conditions in new program initiatives, including
block grants, stating that data must be collected and reported by race,
ethnicity, and primary language, and that programs should allocate
adequate resources to promote compliance, address technological dif-
OCR for page 221
221
DATA COLLECTION AND MONITORING
SHIRE and NHeLP draw four principle conclusions regarding the fed-
eral role in racial, ethnic, and primary language data collection. First, the
collection of such data is legal and authorized under Title VI of the Civil
Rights Act of 1964. Second, a growing number of federal policies empha-
sizes the need for the collection of race, ethnicity, and primary language
data. Third, such data is an indispensable tool for the assessment of
progress toward federal goals of eliminating health disparities (U.S. De-
partment of Health and Human Services, 1999). SHIRE and NHeLP found
broad consensus within U.S. DHHS on this point, but a fourth conclusion
of the investigators is that DHHS policies and practices fail to reflect this
ficulties, ensure privacy and confidentiality of data collected, and
implement effective educational strategies to maximize beneficiary
and provider cooperation with data gathering efforts.
7. Encourage public and private agencies to participate in the develop-
ment and implementation of approaches to improve data availability
and promote data collection and reporting. In support of agencies,
HHS should:
• create a “tool kit” containing information on effective data-related
techniques, technologies, and privacy safeguards currently in use;
• bolster the HHS Data Council’s efforts to identify and document
the benefits of collecting and reporting; and
• support national policies to facility data-sharing among all federal
and state agencies.
8. Expand or create public and private educational efforts to:
• inform insurers, health plans, providers, private/public agencies,
and the general public that data collection and reporting by race,
ethnicity, and primary language are legal and in many instances
required by federal law and regulations;
• raise public awareness that the collection and reporting of these
data are prerequisites for the achievement of Healthy People 2010
goals and essential to demonstrate compliance with the nondis-
crimination requirements of Title VI; and
• inform decision-makers that effective strategies exist for achieving
compliance with data collection and reporting policies, including
risk-adjustment, and make such compliance a condition for receiv-
ing government resources.
9. Provide states and healthcare providers with greater access to ag-
gregated and disaggregated racial, ethnic, and primary language data
acquired at the federal level, subject to privacy and confidentiality
regulations.
10. Support research on existing best practices for data collection.
OCR for page 222
222 UNEQUAL TREATMENT
consensus, as data requirements and methods for collection and report-
ing vary across federal agencies, and no single HHS blueprint exists to
provide a framework and rationale for the department’s activities. Fur-
ther, no department-wide mandate exists for racial, ethnic and primary
language data collection and reporting, leaving only a patchwork of ef-
forts across agencies to promote data collection and reporting (Perot and
Youdelman, 2001).
The SHIRE-NHeLP report notes that two significant developments in
early 2001 illustrate the “disconnect” between federal consensus and prac-
tice. In one instance, HHS finalized regulations regarding standard data
elements for the electronic transmission of health information authorized
under HIPAA, yet these rules failed to identify race or ethnicity as a re-
quired code, an omission that many HHS officials saw as a “lost opportu-
nity.” In another instance, HHS published regulations for Medicaid Man-
aged Care and the State Child Health Insurance Program (SCHIP) that
would require states to report the race, ethnicity, and primary language of
enrollees on a quarterly basis, yet these regulations were suspended for
further review following the change of presidential administrations in
2001 (Perot and Youdelman, 2001). Notably, the National Committee on
Vital and Health Statistics (NCVHS), which serves to advise the federal
government on health information and data policy, warned in a 1999 re-
port that the limited data-collection practices of MCOs who serve Medic-
aid beneficiaries threatened to inhibit HHS’s ability to monitor the quality
of care provided by Medicaid MCOs. NCVHS urged that HHS develop
more specific guidance about the manner and format in which Medicaid
MCO data should be collected and reported by states (Mays, 2001).
Despite the lack of a framework or mandate for systematic data col-
lection at the federal level, data on enrollee race and ethnicity is available
to a limited degree for the two largest federal healthcare programs, Med-
icaid and Medicare. The Centers for Medicare and Medicaid Services
(CMS—formerly the Healthcare Financing Administration [HCFA]) has
generally required states to report patient encounter data for Medicaid
enrollees, but has not required that states report data by race and ethnicity.
Most states have voluntarily supplied CMS with data on Medicaid benefi-
ciaries’ race and ethnicity, and cumulative totals of beneficiaries’ race and
ethnicity are available from all states. As noted above, however, the pro-
posed rule requiring all states to report the race and ethnicity of Medicaid
and SCHIP recipients has yet to be implemented. Further, states would
be expected, via CMS’s proposed rule issued in August, 2001, to provide
Medicaid MCOs with information regarding enrollees’ race or ethnicity,
but these data are often incomplete or inconsistent, and the rule did not
require that this data be reported back to the agency (Perot and You-
delman, 2001). Medicare enrollees’ race or ethnicity has been typically
OCR for page 223
223
DATA COLLECTION AND MONITORING
extracted from the Medicare Enrollee Database, which is based on Social
Security Administration (SSA) information. Enrollment data is available
for all Medicare beneficiaries, but SSA data are limited, particularly for
data obtained prior to 1994, as SSA only identified beneficiaries’ race or
ethnicity as “white,” “black,” “other,” and “unknown.” Efforts by HCFA
to reconstruct this data by surveying the 2.1 million beneficiaries whose
race was listed as “other” or “unknown” reduced the number of unidenti-
fied race codes significantly, but accuracy of these data for beneficiaries
identified as other than “black” or “white” is estimated to be less than
60% (Perot and Youdelman, 2001).
OTHER DATA SOURCES TO ASSESS
HEALTHCARE DISPARITIES
Several other federal, state, and private data sources currently exist or
are planned that can be tapped to assess racial and ethnic disparities in
care. As will be noted later in this chapter, data from these sources can be
used to help identify sources of disparities in care and/or monitor changes
in racial and ethnic disparities in care over time. The following summary
of data collection systems is not intended as an exhaustive listing of fed-
eral, state, or privately funded data sets that may be used to assess racial
and ethnic healthcare disparities. For a more exhaustive listing of federal
data collection systems, see the HHS Directory of Health and Human Ser-
vices Data Resources (U.S. DHHS, 2001).
Several relevant national-level data sources that can be used to assess
aspects of racial and ethnic healthcare disparities include:
Consumer Assessment of Health Plans Survey (CAHPS)
The Consumer Assessment of Health Plans Survey (CAHPS), sup-
ported by the AHRQ, provides information to healthcare consumers, pur-
chasers, health plans, and others regarding the quality of healthcare plans
and services. CAHPS surveys ask consumers about their experiences with
health plans, such as the quality of communication with providers, the
provision of translation services for patients with limited English profi-
ciency, and the timeliness and quality of care provided for a variety of
medical conditions and procedures. CAHPS survey data can be analyzed
by respondents’ race or ethnicity to assess group differences in patient
experiences.
Medical Expenditure Panel Survey (MEPS)
The Medical Expenditure Panel Survey (MEPS), the most recent of a
OCR for page 224
224 UNEQUAL TREATMENT
series of federal surveys of medical care costs, was initiated by the AHRQ
in 1996 for the purpose of assessing the types, frequency of use, and costs
of healthcare services used in the United States. MEPS data yield infor-
mation on health services expenditures and how they are paid for, as well
as the extent of health insurance coverage among the U.S. population.
MEPS consists of four components: the Household Component, which
samples families and individuals to assess health status, insurance cover-
age, healthcare use and expenditures, and sources of payment for health
services; the Nursing Home Component, which samples nursing homes
and residents to assess characteristics of facilities and services offered,
costs, and sources of payment of these services; the Medical Provider
Component, which supplements information from the Household Com-
ponent by surveying hospitals, physicians, and home healthcare provid-
ers; and the Insurance Component, which assesses the amount, types, and
costs of health insurance available to employees. The Household Compo-
nent collects data on respondents’ race/ethnicity, and while the Nursing
Home Component has racial and ethnic data available, only the African-
American and white samples are large enough to permit analysis (U.S.
DHHS, 2001). These data can be assessed by race and ethnicity, as well as
other socio-demographic indicators, such as level of education, income
and assets, and employment. Several of the studies summarized in Chap-
ter 1 utilize MEPS data to assess patterns of disparities in care.
Medicare Beneficiary Enrollment Database
Medicare’s Enrollment Database (EDB), supported by the CMS, is the
principal database for Medicare beneficiary services, including access to
and use of services covered under Medicare. The primary source for EDB
beneficiary information, however, is the Social Security Administration’s
Master Beneficiary Record database. As noted above, these data are unre-
liable with respect to racial and ethnic populations other than black and
white beneficiaries.
Medicare Current Beneficiary Survey
The Medicare Current Beneficiary Survey (MCBS), supported by CMS,
is a continuing sample of Medicare beneficiaries to assess healthcare use,
costs, and who pays for it. A variety of demographic data are collected
from respondents during an initial interview, including race/ethnicity,
health and insurance status, and education level. Data can be used to assess
racial and ethnic differences in costs and utilization of care, and costs paid
by Medicare as well as other public and private insurance sources.
OCR for page 225
225
DATA COLLECTION AND MONITORING
Public and privately funded healthcare plans can take advantage of
survey instruments developed as part of broader quality improvement
initiatives, such as the Health Plan Employer Data and Information Set
(HEDIS).
Health Plan Employer Data and Information Set (HEDIS)
The Health Plan Employer Data and Information Set (HEDIS), devel-
oped by the National Committee for Quality Assurance (NCQA) in con-
junction with public and private purchasers, health plans, researchers, and
consumer advocates, is a set of standardized performance measures that
assesses the quality of healthcare and services provided by managed care
plans. HEDIS was developed to ensure that purchasers and consumers
have access to information to compare the performance of managed
healthcare plans. HEDIS measures the effectiveness and availability of
care in areas such as childhood immunization, breast cancer screening,
cholesterol management, and treatment of heart attack. In addition,
HEDIS offers information on structural attributes of health plans, such as
practitioner turnover and rates of board certification and residency com-
pletion. HEDIS also includes a standardized survey of consumers’ expe-
riences that evaluates plan performance in areas such as customer service,
access to care and claims processing.
At the state level, new data sets being developed, such as the Califor-
nia Health Interview Survey (CHIS), may allow researchers to explore
regional and subpopulation variation in healthcare access and use.
California Health Interview Survey (CHIS)
The California Health Interview Survey (CHIS) is a collaboration of
the UCLA Center for Health Policy Research, the California Department
of Health Services, and the Public Health Institute to assess the health
status, health behavior and risks, and healthcare access and utilization of
the state’s diverse population. Data from its survey of 55,000 California
households will be available in early 2002 and will be made available
through published reports, public-use files, and an Internet-based system
that will allow requestors to gather information tailored to particular
health topics, population groups, and geographic areas. In particular,
CHIS asks respondents to provide information about their usual source of
care, access to and use of specific services, experiences of discrimination
in healthcare settings, and recall of provider advice, among other items.
Results will be analyzed by respondents’ race and ethnicity, with particu-
lar attention to racial and ethnic subgroups. Funding for CHIS has been
OCR for page 226
226 UNEQUAL TREATMENT
provided by the California Department of Health Services, The California
Endowment, the National Cancer Institute (NCI), California Children and
Families Commission, the U.S. Centers for Disease Control and Preven-
tion (CDC), and the Indian Health Service (IHS).
MODELS OF MEASURING DISPARITIES IN HEALTHCARE
Many models of healthcare “report cards” have been developed over
the past few years, as healthcare consumers and purchasers of plans have
expressed great interest in timely and accurate information about the qual-
ity of care delivered by plans, hospitals, and individual providers. Few
such “report cards,” however, have focused exclusively or in part on ra-
cial and ethnic disparities in care. This paucity of information on dispari-
ties in care is likely to change in the near future, as federal and private
initiatives are increasing visibility and attention to the problem. In one
instance, the Office of the Assistant Secretary for Planning and Evaluation
of the U.S. Department of Health and Human Services (U.S. DHHS) has
recently commissioned a review of measures of discrimination in health-
care settings. In another federal initiative, AHRQ has initiated plans to
develop a national report on racial and ethnic disparities in healthcare,
and plans to incorporate measures of racial and ethnic disparities in care
in a national report of quality of care. Within the private sector, the Na-
tional Quality Forum (NQF), with support from The Commonwealth
Fund, has produced a report on measuring and reporting the quality of
care for minority populations. These activities are likely to spur efforts to
increase information available to consumers and purchasers of plans and
promote greater choice when selecting plans, to promote accountability
to consumers and purchasers, and to spark action on the part of plans,
providers, and legal and regulatory bodies to reduce disparities in care.
Two models of “report cards” that specifically address racial and eth-
nic disparities in healthcare are reviewed below.
“Health Accountability 36”
Smith (1998) proposes a report card to assess racial and ethnic dis-
parities consisting of 36 consensus indicators that have been developed
and utilized in other settings by a range of public and private entities.
The indicators were selected based on the availability of data, sensitivity
of the indicators to key health conditions for vulnerable populations, and
their amenability to public health and healthcare intervention. The first
12 indicators include measures adapted from the U.S. DHHS initiative
Healthy People 2000, and are routinely collected and reported by the Na-
OCR for page 227
227
DATA COLLECTION AND MONITORING
tional Vital Statistics report system to evaluate the health of geographi-
cally defined populations. The second 12 indicators include measures of
managed care plans to provide consumers and purchasers with informa-
tion about plan performance. Of these, the first six were developed by the
National Committee for Quality Assurance for HEDIS, while the subse-
quent six indicators were selected by the former Agency for Healthcare
Policy and Research (now AHRQ). The third set of 12 indicators was
developed by the Joint Committee on Accreditation of Health Care Orga-
nizations (JCAHO) as part of its accreditation process to measure hospital
performance, and reflect measures of obstetrical, oncologic, and cardio-
vascular outcomes. Smith (1998) notes that data for these indicators are
currently available and can be analyzed using the standard categories for
race and ethnicity adapted by the Office of Management and Budget (see
Chapter 1, Table 1). A goal for public health agencies and health systems,
Smith suggests, would be to bring racial and ethnic disparities to within
80%. These measures are listed in Table 7-1.
Several of the measures proposed by Smith can be criticized on the
grounds that as indicators of population health, they are influenced to a
far greater extent by social and economic forces such as income inequal-
ity, residential segregation (and subsequent substandard living condi-
tions, especially for lower-income minority groups), environmental risks,
and other social problems. As such, they are less amenable to health sys-
tem intervention. Further, health systems that disproportionately enroll
lower-income and minority patients will have a greater challenge in im-
proving the health of a generally sicker population with higher rates of
co-morbidities, and thus, may not demonstrate improvement on many of
the measures. Smith (1998) notes, however, that the impact of plans’ case-
mix can be adjusted statistically. In addition, he notes, some health plans,
such as not-for-profit integrated delivery systems, recognize the impact of
social and economic forces on the health of their enrolled populations and
attempt to address these forces by improving screening and primary and
preventive healthcare services, and by addressing housing and other so-
cial service needs of their patients.
Integrated Approaches
LaVeist and Gibbons (2001), in their report to U.S. DHHS1 on poten-
1U.S. DHHS commissioned LaVeist to “summarize the literature on racial/ethnic discrimi-
nation within healthcare settings, with the primary goal of describing how discrimination
has been measured” (LaVeist and Gibbons, 2001, p. 1). In this review, the authors note that
the existence of racial and ethnic disparities in healthcare does not necessarily reflect dis-
crimination, but focus their analysis on indicators that may detect patterns of discrimination
apart from disparities that are not inherently discriminatory.
OCR for page 228
228 UNEQUAL TREATMENT
TABLE 7-1 “Health Accountability 36” Report Card Indicators
Unit of Analysis Source Indicators
Geographically Healthy 1. Total age-adjusted death rate
Defined People 2. Automobile death rate
Population 2000 3. Suicide death rate
4. Lung cancer death rate
5. Breast cancer death rate
6. Cardiovascular death rate
7. Homicide death rate
8. Teen births
9. Inadequate prenatal care
10. % Low birthweight births
11. Infant death rate
12. Children in poverty
Health Plan HEDIS 1. % Women for whom prenatal care began in the first
Covered Lives AHCPR trimester
2. % Children receiving all childhood immunizations by 24
months
3. Cholesterol screening age 40-64 once in 5-year period
4. % Women 51-64 continuously enrolled for 2 years who
received mammogram breast cancer screening
5. % Women 21-64 continuously enrolled for 3 years who
received a Pap test
6. % Members 2-19 with one or more asthma admissions
7. % Diabetics 31-64 who had retinal exam during the
preceding calendar year
8. % Members 23-39 who visited a health practitioner in the
past year
9. % Rating how well the doctor listened as excellent
10. % For whom last visit to doctor fully met their needs
11. % Choice of doctors not a problem
12. % Satisfied with overall plan
Hospital Patient JCAHO Obstetrical Indicators:
Clinical 1. % Low birthweight infants
Population 2. % Term infants admitted to NICU within one day of delivery
3. % Neonates with an Apgar of 3 or less at 5 minutes and a
birthweight > 1,500 grams
4. % Neonates with a discharge diagnosis of significant birth
trauma
Oncology Indicators:
5. Survival of patient with primary cancer of the lung, colon/
rectum, by state and histologic type
6. Use of test critical to diagnosis, prognosis, and treatment
7. Use of treatment approaches that have an impact on
quality of life
8. Interdisciplinary treatment and follow-up
Cardiovascular Indicators:
9. Intrahospital mortality as a means of assessing multiple
aspects of CABG care
10. Extended postoperative stay as a means of assessing
multiple aspects of CABG care
11. Intrahospital mortality as a means of assessing multiple
aspects of PTCA care
12. Intrahospital mortality as a means of assessing multiple
aspects of acute MI care
SOURCE: Smith (1998).
OCR for page 229
229
DATA COLLECTION AND MONITORING
tial measures of discrimination in healthcare settings, note that such mea-
sures must not only address structural differences in receipt of care (e.g.,
the proportion of women receiving prenatal care in the first trimester, as
suggested by Smith [1998]), but should also assess the quality of interper-
sonal interactions in healthcare settings. Structural differences shape the
parameters of care provided to different populations, they note, but indi-
vidual, subjective factors affect the quality of care in clinical interactions.
They argue for an integrated approach that includes multiple measures,
and meets the following criteria:
1. Applicable to multiple racial/ethnic groups—the indicators must be ap-
plicable to all racial and ethnic groups that make up the U.S. population.
2. Produce unique scores for individual healthcare facilities—the report
card must be producible for individual healthcare facilities and not merely
produce scores for the nation or a particular region.
3. Data sources must be accessible—the report card must be easily un-
derstandable to a broad audience of healthcare consumers and the indica-
tors must have high “face validity.”
4. No confounding—indicators must not be confounded with other
variables such as health insurance, patient preferences or larger societal
factors. If there is confounding, there must be a way to adjust for it.
5. Longitudinality—the indicators must have the ability to be repli-
cated over time (LaVeist and Gibbons, 2001, p. 7).
LaVeist and Gibbons weigh the merits of four potential approaches to
measuring discrimination in care, including Smith’s (1998) “Health Ac-
countability 36,” patient assessments, administrative claims audits, and
assessments of substandard care. The “Health Accountability 36” mea-
sures draw largely upon existing data, and can be applied to geographi-
cally defined populations, individuals in health plans, and hospital and
clinic patients. LaVeist and Gibbons note, however, that many of the mea-
sures, particularly those assessing racial differences in health status, are
confounded with larger social and economic factors.
Several measures of patient satisfaction have been extensively evalu-
ated, according to LaVeist and Gibbons, and several studies have assessed
racial and ethnic differences in patients’ perceptions of the quality of care
they receive (reviewed earlier). Few of these measures, however, have
explicitly assessed patients’ perceptions of racial discrimination in care
settings (the Seattle-King County survey of patient perceptions of dis-
crimination in care, reviewed earlier, is a notable exception). Such mea-
sures have the potential of providing unique scores for individual health-
care facilities and can be used to assess changes over time. Patient
perceptions of care, however, can be influenced by a wide range of fac-
OCR for page 230
230 UNEQUAL TREATMENT
tors, and may not reflect whether patients are receiving care appropriate
to their needs. Nonetheless, such perceptions form an important compo-
nent of a multi-pronged assessment profile, particularly if measures can
assess the degree of patient participation in treatment decisions and un-
derstanding of their diagnosis and course of treatment.
Administrative claims data have been used extensively in prior re-
search to audit care and demonstrate racial disparities in access to diag-
nostic and therapeutic procedures (much of this research is reviewed in
Chapter 1). Well-controlled studies using claims data have adjusted for
many potentially confounding factors, such as co-morbid conditions and
insurance status, to isolate the influence of patient race on receipt of care.
LaVeist and Gibbons (2001) suggest that administrative audits can pro-
duce unique scores for individual hospitals and healthcare facilities. Such
data often fail, however, to illuminate process-of-care variables, such as
referral patterns or participation in treatment decisions (e.g., whether pro-
viders present all treatment options and whether patients accept or refuse
them). Prospective studies are therefore needed to supplement typically
retrospective analyses of administrative claims data (see Chapter on “Re-
search Needs”).
Measures of adverse events due to practitioner or healthcare setting
error are also an important component of assessing disparities in care,
according to LaVeist and Gibbons (2001). Increasingly, healthcare pro-
viders and consumers have focused on the problem of medical errors and
patient safety, and at least two methodologies have been developed to
evaluate adverse events. Both involve an initial screening of potentially
problematic cases, typically by two trained healthcare professionals, but
screening methods differ in that one approach utilizes actual medical
records, while the other uses administrative claims data. Such analysis
could indicate whether minority patients are differentially more or less
likely to face substandard care. This method has the advantage of yield-
ing objective data on the quality of care provided, relative to standard
criteria. Data are free of confounding, and the accuracy and validity of
these methods has been demonstrated, the authors note.
LaVeist and Gibbons (2001) conclude that a two-tiered, multi-assess-
ment approach may be useful to assess discrimination in healthcare set-
tings. In the first tier, routine monitoring of healthcare facilities can be
accomplished by audits of administrative data and analyses of data on
substandard care. This initial “screen” could identify facilities that should
be investigated more closely. In the second assessment tier, facilities are
informed of the disparities and are given a period of time to address them.
If progress has not been made, LaVeist and Gibbons suggest, a method
used more commonly to assess housing and employment discrimina-
tion—paired testing—may be used to further assess the possibility of ra-
OCR for page 231
231
DATA COLLECTION AND MONITORING
cial or ethnic discrimination. In this strategy, individuals are trained to
present the same needs and background information to targeted health-
care facilities, but vary only in race or ethnicity (see Chapter on “Racial
Attitudes and Discrimination”). The purpose of such testing, according
to the authors, is to enhance awareness and to facilitate voluntary efforts
to address racial disparities in care. Should these efforts fail, judicial rem-
edies could be explored if clear violations of civil rights laws are found
(LaViest and Gibbons, 2001). Unlike paired testing in housing and em-
ployment, however, the use of such strategies in healthcare settings poses
unique legal and ethical challenges that should be addressed before such
strategies are adopted.
Reporting of Racial and Ethnic Disparities Using Existing Data Sets
As noted earlier, the HEDIS data sets developed by NCQA offer a
ready set of measures of plan performance that are widely used and ac-
cepted by health plans, purchasers, and consumers. Health plans volun-
tarily report this information to NCQA, which then disseminates data as
part of its Quality Compass database in regular publications such as the
NCQA State of Managed Care Quality report. Quality Compass 2000 con-
tains measures of plan performance in several clinical areas, such as can-
cer screening, childhood and adult immunization, timely outpatient care,
and evidence-based treatments for hypertension, cardiovascular disease,
asthma, diabetes, and depression. Approximately half of the nation’s
HMOs participate in Quality Compass, with another 90% participating in
NCQA’s Accreditation and HEDIS programs.
Some researchers and plan administrators have raised concerns that
health plan performance on these or other quality measures is affected by
the sociodemographic mix of plan enrollees. According to this view, plans
that enroll a high percentage of low-income or racial and ethnic minorities
(who tend to be sicker, face a greater number of barriers to accessing care,
and are less likely to utilize preventive and primary care services) may tend
to face poorer health plan performance scores as a result of factors exog-
enous to the health system (Zaslavsky et al., 2000). Zaslavsky et al. (2000)
tested this hypothesis by studying the relationship between plan perfor-
mance on HEDIS measures and sociodemographic mix, including enrollee
age, gender, and area of residence as an indicator of race/ethnicity and
household income. The authors found that plan performance was nega-
tively associated with the percentage of individuals receiving public assis-
tance and the percentage of African Americans and Hispanics in enrollees’
area of residence, and positively associated with the percentage of college-
educated and Asian-American residents. Adjusting for these demographic
OCR for page 232
232 UNEQUAL TREATMENT
variables, however, had a limited effect on plan performance, as most plans
changed by less than 5% in performance measures.
Romano (2000) argues that even if case-mix differences could be ad-
equately adjusted statistically, such adjustment does not necessarily im-
prove analysis of the quality of care that plans deliver. To the contrary, he
argues that statistical adjustment may hamper accurate assessment of plan
performance by failing to identify the direction of the relationship between
case-mix and plan performance—in other words, does the plan’s case-mix
result in poor performance, or does poor performance lead to the observed
case-mix? In addition, statistical adjustment may “excuse” health plans
for failing to address socioeconomic and racial/ethnic health disparities.
Adjustment for case-mix may inadvertently remove plans’ incentive to
reduce disparities, according to Romano, by masking differences in the
level of care provided to racial and ethnic minorities and low-income en-
rollees. He argues for reporting of data stratified by race, ethnicity, and
socioeconomic status, which would offer the advantage of highlighting,
rather than masking, sociodemographic disparities, and would allow con-
sumers to make better informed choices about plans based on their own
sociodemographic profile. In addition, by presenting performance data
stratified by race, ethnicity, and socioeconomic status, plans could be re-
warded for efforts to reduce disparities (Romano, 2000; Fiscella et al.,
2000).
DATA NEEDS AND RECOMMENDATIONS
The preceding discussion illustrates that despite the many challenges
inherent in efforts to collect data on patients’ race and ethnicity and moni-
tor the quality of their care, data collection and monitoring are a feasible,
critically important step in understanding and eliminating disparities in
care. As Tom Perez (this volume) notes, “Effective data collection is the
linchpin of any comprehensive strategy to eliminate racial and ethnic dis-
parities in health.”
Currently, data collection efforts are unsystematic and inadequate to
monitor the quality of care provided to racial and ethnic minorities. These
efforts must be improved to ensure accountability of plans and providers
to healthcare payors and consumers, to track disparities and assess the
impact of quality improvement efforts, and to identify best practices that
may be replicated by other plans and health systems. Federal leadership
is needed to spearhead data collection efforts; for this reason, the commit-
tee advocates that the Secretary of the U.S. Department of Health and
Human Services produce periodic studies to assess progress in eliminat-
ing racial and ethnic disparities in healthcare. The private sector, how-
ever, also shares a role in encouraging data collection and reporting of
OCR for page 233
233
DATA COLLECTION AND MONITORING
patient care data by race, ethnicity, and where possible, primary language.
Accreditation bodies, such as JACHO and NCQA, should require the in-
clusion of data on patient race, ethnicity, and highest level of education
attained (in case of children, highest level of education attained by mother)
in performance reports of public and private providers as part of health-
care performance measurement. Such an emphasis would help to ensure
that addressing healthcare disparities is seen by plans, providers, and
purchasers as central to broader healthcare quality improvement efforts.
Data collection should be accomplished using a standard racial/eth-
nic classification scheme. Current OMB standards can be used, but data
categories must go beyond the existing minimum standards to reflect the
diversity within racial and ethnic populations, particularly at the local
level (e.g., subgroups of Hispanics, African Americans, Asian Americans,
etc.). In addition, information is needed on patients’ socioeconomic sta-
tus and primary language. These data should be stratified, where pos-
sible, to better understand the relative contributions of race/ethnicity, so-
cioeconomic status, and other demographic variables to variations in care.
In the future, a standardized, central database is needed, with safe-
guards for privacy and confidentiality, which can be merged with other
data systems. This database should be consistent with efforts to develop
electronic patient medical records, and should be compatible to merge
with other data systems. Such a long-term goal will require federal lead-
ership and financial support.
Recommendation 7-1: Collect and report data on healthcare access
and utilization by patients’ race, ethnicity, socioeconomic status,
and where possible, primary language.
Standardized data should be collected on the race, ethnicity, and
highest level of education (in case of children, highest level of educa-
tion attained by mother) of all patients enrolled in publicly funded
health programs and reported to Congress. Collection of data on
patients’ primary language should be encouraged, where feasible, as
part of this effort. Data on healthcare access, use, and outcomes
should be reported by race, ethnicity (including subgroups, and pri-
mary language where possible), and adjusted for highest level of
education.
Recommendation 7-2: Include measures of racial and ethnic dis-
parities in performance measurement.
JCAHO and NCQA should require the inclusion of data on patient
race, ethnicity, and highest level of education attained (in case of chil-
dren, highest level of education attained by mother) in performance
reports of public and private providers as part of health care perfor-
OCR for page 234
234 UNEQUAL TREATMENT
mance measurement, such as NCQA’s HEDIS indicators. The collec-
tion of data on patients’ primary language should be encouraged.
These performance reports should make elimination of healthcare dis-
parities a focus of quality improvement efforts.
Recommendation 7-3: Monitor progress toward the elimination of
healthcare disparities.
The secretary of HHS should conduct periodic studies to monitor
the nation’s progress toward eliminating racial and ethnic health-
care disparities, to provide insight into the root causes of these dis-
parities, and to assess opportunities for intervention and improve-
ment.
Recommendation 7-4: Report racial and ethnic data by OMB cat-
egories, but use subpopulation groups where possible.
Current OMB categories for race and ethnicity should be used in all
reporting and monitoring efforts, but data categories must go be-
yond the existing minimum standards to reflect the diversity within
racial and ethnic populations (e.g., subpopulations), particularly at
the local level.
Representative terms from entire chapter:
primary language
