State-Based Collection of Data on Race, Ethnicity, Socioeconomic Position, and Acculturation and Language Use
State health agencies collect a large amount of data on health and health care services to aid in their missions of providing health programs and services to the populations of their states. States also collect these data as part of their missions to license and regulate health care providers and insurers, train and distribute a health workforce, and to generate measures for market and policy decisions. In addition, they collect such data to meet federal data collection requirements for programs that are run by individual states with some funding from federal sources.
The federal government relies on states for much data collection because its regulatory powers and service provision activities are not nearly so broad. The states, in fact, are the collectors of much of the data used to study health and health services at both the state and federal levels. This is why there are many cooperative data sharing efforts, such as the Vital Statistics Cooperative Program (VSCP) for vital statistics, the Healthcare Cost and Utilization Program (HCUP) for hospital discharge data, the Surveillance, Epidemiology, and End Results (SEER) program for cancer, the Behavioral Risk Factor Surveillance System (BRFSS), and Medicaid. In this chapter, we briefly discuss the major state-based data collection systems and the racial and ethnic, socioeconomic position (SEP), and acculturation and language data that are collected in them. The systems reviewed are vital statistics birth and death records, hospital discharge abstracts, cancer registries, health interview surveys, and Medicaid and the State Children’s Health Insurance Program (SCHIP), with examples of how these data have been used to understand disparities in health and health care. We also discuss
gaps in these state-based data collection systems. The chapter concludes with recommendations calling for states to push for the collection of data on race, ethnicity, SEP, and acculturation and language use as much as possible, and for the Department of Health and Human Services (DHHS) to provide technical assistance, resources, and incentives to states to improve the collection and use of these data.
The panel commissioned a background paper on state-based data collection for its Workshop on Racial and Ethnic Data in Health (see the paper by Geppert et al., in Appendix E). This paper, which was presented at the workshop, drew on interviews with officials from four states about their racial and ethnic data collection as case studies on how states use such data and the problems they encounter in collecting and analyzing them. This chapter draws on the results presented in the paper.
In considering state-based data collection, it is important to note that states’ needs for social and demographic data on their populations are different from those of the federal government. State governments directly implement and evaluate health intervention programs in order to ensure the health of their populations. While the federal government also implements health programs and monitors public health, it does so, for the most part, in an indirect way through contracts that is more removed from actual implementation of the programs (the Indian Health Service is an exception).
The federal government must monitor the nation as a whole and cannot, because of resource constraints, regularly collect data on ethnic groups that do not represent a sizable portion of the national population. It therefore aggregates data on racial and ethnic groups into generally broad categories. States, on the other hand, must provide services for their own populations, which may include concentrated populations of particular ethnic groups; for example, Dominicans and Puerto Ricans in New York, Cubans in Florida, Hmong in Minnesota, and Mexicans in Texas.
These subgroups may have differing health and health care needs. For example, Puerto Ricans are more likely than other Hispanics to have low-birthweight babies (9.3 percent of live births compared with the 6.5 percent national average for all Hispanics) (National Center for Health Statistics, 2003); and Hawaiians have a lower rate of early prenatal care than other Asian and Pacific Islanders (79.1 percent, compared with 90.1 percent for mothers of Japanese descent, 87.0 percent for mothers of Chinese descent, 85.0 percent for mothers of Filipino descent, and 82.7 percent for mothers of other Asian or Pacific Islander descent) (National Center for Health Statistics, 2003).
For these reasons, the broad OMB categories may not be as appropriate for all states as they are for most federal-level data collection. States need data for specific ethnic subpopulations in order to target public health interventions and measure health outcomes and disparities, and these sub-
populations may not correspond to the broad racial and ethnic categories stipulated by the federal government. It should be noted, however, that the OMB standards establish minimal racial and ethnic categories and therefore do not prohibit the use of more detailed categories.
VITAL STATISTICS BIRTH AND DEATH RECORDS
Each state issues birth and death certificates as part of the country’s vital statistics system. These data provide states with information that can be used to assess and improve the health of the population. For example, states use vital statistics for prenatal care interventions and infant mortality reduction. Information for birth certificates is recorded at the birth of an infant by a health care professional, and information for death certificates is usually collected by a funeral home director. It is believed that these two record systems are essentially complete in their coverage of births and deaths in the United States (see U.S. DHHS, 1997).
All states and territories provide these core data to the National Center for Health Statistics (NCHS) under the Vital Statistics Cooperative Program (VSCP). The program (and its predecessors) was implemented to record national vital statistics and to encourage comparable reporting of these events across states and U.S. territories. Standards for the reporting of minimum basic data items were developed (and continue to be reassessed) by NCHS working with state vital statistics organizations. States are funded to provide the standardized data to NCHS, with each state’s federal funding level based on its reporting of these minimum basic data.
Standards for reporting racial and ethnic data are included as are standards for reporting the education levels of the parents (on birth certificates) or of the decedent (on death certificates). For birth certificates, the race and ethnicity of the infant are not reported; rather, the race and ethnicity of the infant’s mother and father (if known) are reported. The education levels of both parents are recorded in the same manner, as are their countries of origin. The form is filled out by a medical records clerk, who is supposed to ask the parents for this information. In the case of death registrations, the race and ethnicity and education level of the deceased are usually recorded by the funeral home director or a health care worker who requests the information from either the decedent’s next of kin or a family representative while filling out the forms. No data are collected on the decedent’s country of origin or language.
The racial and ethnic categories currently used in the vital statistics system were reviewed as part of the regular vital statistics standard certificate review process (NCHS, 2000). This review resulted in a recommendation for expanded racial and ethnic categories that would include separate Asian categories (Asian Indian, Chinese, Filipino, Japanese, Korean, Viet-
namese, and other Asian) and separate Pacific Islander categories (Native Hawaiian, Guamanian or Chamorro, Samoan, and other Pacific Islander)—all of which can be aggregated to the five minimum OMB categories—and “specify” lines where individuals can indicate their tribe or “other” status. In addition, the review panel recommended that question about Hispanic ethnicity be listed before the race question. The recommended categories are shown in Box 5-1. The recommendations of the review are still being considered and have not yet been implemented.
Several studies have examined the quality of racial and ethnic data from vital records. Hahn, Mulinare, and Teutsch (1992) used birth records linked with infant-death records to study the consistency of racial and ethnic reporting for infants who died before their first birthday. The study compared the race and ethnicity coded on the infant’s birth record (which is determined by the parents’ report of race and ethnicity) and the infant’s death record (which is usually recorded by observation from a funeral director or other certifier) and how these inconsistencies affected computa-
NOTE: It was recommended that a question about Hispanic ethnicity precede this series of choices. The census began prefacing race choices with a Hispanic ethnicity question in 2000.
SOURCE: National Center for Health Statistics (2000).
tions of infant mortality rates (IMRs). Results showed that 3.7 percent of the infants in the study sample were classified as having a different race at death from that recorded at birth. Inconsistencies in classification were more prevalent for nonwhites, and especially for infants of “other” race classifications. More infants were classified as white at death than at birth (i.e., there were fewer infants classified as black at death compared to the number at birth, and fewer infants classified as other race at death than at birth). Finally, the authors found that if consistent racial and ethnic coding is used to calculate IMRs,1 the IMRs for whites and non-Hispanic whites decrease compared with those for the same groups using standard race definitions. However, the IMRs for blacks, non-Hispanic blacks, and other race groups increase compared to IMRs calculated using standard race definitions. A similarly structured study in the state of Washington found that 61 percent of infants who died in the first year of life and were recorded as being of American Indian or Alaska Native descent on their birth records were coded as American Indian or Alaska Native at death (Frost and Shy, 1980).
Other studies have found high-quality information on race and ethnicity in vital records for most racial and ethnic groups. Baumeister and colleagues (2000) found that racial and ethnic information on birth records was quite similar to the respondent’s self-identified racial and ethnic information provided in face-to-face interviews. The one group for which birth records did not match self-reported racial and ethnic categorizations at a high rate was Native Americans. Sorlie, Rogot, and Johnson (1992), using data from the National Longitudinal Mortality Study, showed that classification of blacks and whites in death records is highly comparable to racial and ethnic classifications from survey data reports for blacks and whites. However, the classification of American Indians and Asian and Pacific Islanders on death records had more errors. Individuals from these two groups were often categorized as white on death records. The authors suggested that these misclassifications could result in underestimation of death rates for these groups. Frost and colleagues (1994) came to a similar conclusion regarding data on death certificates for American Indians and Alaska Natives in Washington state. This study found that 12.8 percent of individuals who appeared in the Indian Health Service (IHS) patient registry (that is, patients treated at IHS facilities, who must be a member or descendent of a member of a federally recognized tribe) in Washington state were not classified as American Indian or Alaska Native on their death
records. The authors concluded that death rates for this group may be underestimated.
HOSPITAL DISCHARGE ABSTRACTS
Thirty-seven states have legislative mandates to collect discharge abstracts on patients that use nonfederal hospitals. In most states without a mandate, providers may submit the data voluntarily to a private entity. These data contain a complete demographic, treatment, and financial record for each patient admitted to the hospital and are the only source of population-based health care data (Chapter 6 discusses these abstracts in more detail). States use the data to evaluate health care access, cost, and quality and to support policy and market-based decisions. Medicaid programs use the data for comparative norms for uncompensated care estimates. Most states that collect these data use a standard format, the Uniform Bill for Hospitals (UB92), a form that is used to pay claims. Although the core UB92 data elements do not include standard racial and ethnic data, 27 states include these data as part of their own requirements (see the paper by Geppert et al., in Appendix E).
The federal government collects and maintains hospital discharge data as an important source of information to study health care utilization and treatment outcomes. The government purchases the state hospital discharge data from state and private data organizations and hospital associations to create a national data set, the Healthcare Cost and Utilization Project (HCUP) maintained by the Agency for Healthcare Research and Quality (AHRQ). HCUP is a family of databases and products and includes the largest collection of longitudinal hospital care data in the United States, with all payer, discharge-level information since 1988. The AHRQ Quality Indicators provide information about health care cost, quality, and access derived from the discharge data. States can measure inpatient admissions to hospitals for preventable conditions such as diabetes or asthma. If the state captures data on race and ethnicity with its discharge data, admission rates for Hispanics and other minority groups can be measured and used to target interventions. Data from HCUP have also supported disparities research. Elixhauser and colleagues (2002) used data from the Healthcare Cost and Utilization Project, which is described further in Chapter 6, to examine disparities between Hispanic and non-Hispanic whites with cerebrovascular disease in the use of in-hospital diagnostic and therapeutic procedures. The study found that controlling for a hospital’s experience with Hispanic patients eliminated or reduced the magnitude of disparities in procedure use.
Because HCUP data are collected at the state level and states vary in
their data collection practices, many fields, such as race and ethnicity, are incomplete or inconsistent in their format.
Forty-five states collect data on cancer cases, including information about the occurrence of cancer, the type and location, the conditions of the cancer at diagnosis, and treatment. These data are used to study the causes of cancer and outcomes of treatment, and to target intervention and prevention programs. The registries are sponsored by either the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program or the CDC’s National Program of Cancer Registries (NPCR).2 Data come from hospitals, physicians’ offices, and laboratories and are provided to a central statewide registry, or, in the case of six SEER metropolitan sites and other SEER rural or special population sites, to local registries. The data are then shared with either the National Cancer Institute (SEER sites) or the CDC (NPCR sites) or, in some cases, both.
The CDC provides funding and technical assistance to NPCR participant states to collect the data and to improve their completeness, timeliness, and quality. To receive funding, states must implement quality and format standards that are reviewed and certified by the North American Association of Central Cancer Registries (NAACCR). These standards include prescribed formats for the collection of both racial and ethnic data.3 SEER abstracts racial and ethnic data that are as detailed as are available from the records used to identify individual cases of cancer. For example, if a record shows the individual is of Chinese descent, SEER will include that information in its record for that incidence of cancer. The SEER system also aggregates these racial and ethnic data into one of four broad race categories (black, white, Asian or Pacific Islander, and American Indian or Alaska Native) and into Hispanic or non-Hispanic ethnicity. In 2002, 28 states participated in NPCR and were certified as meeting the data standards requirements, 14 participated but were not certified, 4 states participated in
See the Web sites for SEER (http://seer.cancer.gov/) and NPCR (http://www.cdc.gov/cancer/npcr/).
The racial categories used are white, black, American Indian/Aleutian/Eskimo, Chinese, Japanese, Filipino, Hawaiian, Korean, Asian Indian/Pakistani, Vietnamese, Laotian, Hmong, Kampuchean, Thai, Micronesian, Chamorran, Guamanian, Polynesian, Tahitian, Samoan, Tongan, Melenisian, Fiji Islander, New Guinean, other Asian (including Asians, not otherwise specified [NOS] and Oriental), Pacific Islander NOS, other, and unknown. Ethnicity categories are non-Spanish/non-Hispanic, Mexican (includes Chicano), Puerto Rican, Cuban, South or Central American (except Brazil), other Spanish (includes European), Spanish NOS, Hispanic NOS, Spanish (surname only), and unknown whether Spanish or not.
both NPCR and SEER and were certified, and 4 participated in SEER alone.4
Cancer registries do not include much information on socioeconomic position. While work history is often contained in these records, usually to understand exposure to environmental factors, this information is often missing or incomplete (Swanson, Schwartz, and Burrows, 1984). However, these systems include geographical information on the patient that would allow for the geocoding of aggregate-level SEP data from the person’s record. Singh and colleagues (2003) linked SEER information to county and census tract-level information on poverty rates to study cancer incidence, diagnosis, and mortality across areas with differing poverty levels.
STATE HEALTH INTERVIEW SURVEYS
A number of states and local areas conduct their own surveys of individuals to collect data on health status and health care services. These surveys typically sample the population of the state or local area. State surveys are useful because they can be used for state-level estimates and allow for details on subgroups (by race and ethnicity and/or by geographic area such as county or region). For example, the Hawaii Department of Health has conducted the Hawaii Health Survey every year since 1968. It is a household survey of all persons living in noninstitutionalized housing units in Hawaii and collects demographic and health data to monitor the health, socio-demographic status, and other characteristics of the population of Hawaii. It also provides statistics for planning and evaluation of health services and programs, and identification of problems (see http://www.hawaii.gov/doh/stats/surveys/hhs.html). Based on this survey, the Hawaii Department of Health publishes annual reports on the general demographics, income, health insurance, and health conditions of the state’s population (see Hawaii Department of Health, 2003). Many health conditions are reported by ethnicity.
The state of California in 2001 sponsored the California Health Interview Survey, which collected extensive health-related data on individuals—demographic information, health status, health insurance coverage, eligibility and participation in public coverage and assistance programs, utilization of health services, and access to care. These data have been used, for example, to show that racial and ethnic differences exist in the percent of adults with diabetes in California who monitor their glucose levels at least once per day and in percentages for those with different health insurance
statuses (Diamant et al., 2003). This study found that American Indians and Alaska Natives were the most likely to monitor their glucose level, while Latinos were the least likely to monitor. Not surprisingly, those without health insurance were also least likely to monitor their glucose, whereas those with Medi-Cal (California’s Medicaid program) were the most likely.
King County (Seattle), Washington, conducted a survey modeled on the Behavioral Risk Factor Surveillance System called the Ethnicity and Health Survey, which oversampled seven racial and ethnic minority groups, including five different Asian ethnic groups with high concentrations in the Seattle area (Smyser, Krieger, and Solet, 1999).
Many states use surveys developed on a national level for implementation by states and localities that choose to use them. Examples of these surveys include the Behavioral Risk Factor Surveillance System (BRFSS) (described in Chapter 4) and the Pregnancy Risk Assessment Monitoring System (PRAMS) (described in Chapter 3), which collects information from women with new infants on their behavior before, during, and after pregnancy. Each of these surveys contains a core set of questions that can be supplemented by each state, with additional questions according to its data collection needs.
One model of a federal-state data partnership is the Health Resources and Services Administration’s (HRSA) State Planning Grants. HRSA provided funding to states to develop plans for providing access to affordable health insurance coverage to all citizens. In 2002, 32 states received funding to collect data for and analysis of the characteristics of the uninsured and to develop potential models to increase coverage.
MEDICAID AND SCHIP DATA
Medicaid provides health insurance coverage for low-income families with children, disabled individuals, and certain low-income elderly citizens. It is a program with shared state and federal funding that is coordinated by the Centers for Medicare and Medicaid Services (CMS) in DHHS. SCHIP is a federally funded, state-administered program that provides insurance coverage for low-income children from birth to age 18. States can choose to use federal SCHIP funds to expand their Medicaid programs or as separate state health insurance programs. SCHIP served 5.3 million children in 2002.
States collect information on Medicaid program enrollees to assess eligibility and administer the program. Until the Balanced Budget Act of 1997, states were required to report only aggregate data to the federal government, although they could voluntarily report individual-level data on enrollees. When the voluntary data collection ended (1998 was the last year), 32 states had submitted data on individuals (including racial and
ethnic data) to the federal government. The 1997 act mandated that states report data on eligibility and claims through the Medicaid Statistical Information System (MSIS) on a quarterly basis beginning in 1999. These data have not yet been extensively used for research purposes.5
The MSIS required data set includes information on eligibility (e.g., characteristics of Medicaid enrollees, including racial and ethnic data in addition to financial information) and on claims made on their behalf (e.g., utilization of health care services and payments). States have the discretion to collect racial and ethnic information as they see fit; for those that do, CMS, through MSIS, requests that the data be reported in a standardized format. Currently, racial and ethnic data are reported to CMS in a set of categories that combines race and ethnicity, with no separate question for Hispanic ethnicity. The categories of race and ethnicity collected by CMS from the states are: white; black or African American; Asian; Hispanic or Latino (no race information available); Native Hawaiian or other Pacific Islander; Hispanic or Latino and one or more races; more than one race (Hispanic or Latino not indicated); and unknown. States do have the option, however, of reporting race and Hispanic ethnicity in separate questions, indicating whether an enrollee is white or not, black or African American or not, American Indian/Alaska Native or not, Native Hawaiian or Other Pacific Islander or not, and Hispanic or not.6 Thus, if states record this information separately, information about Hispanic ethnicity can be obtained separately from racial status in the MSIS. In addition, if a state collects this information, the MSIS data include an option for multiracial status and so multiple races can be identified as well. To accommodate the additional data collection, however, each state program will have to implement form changes, computer system changes, and staff training.
CMS does not yet have any information on the quality of the racial and ethnic data collected through the MSIS. In fiscal year 2000, the race and ethnicity of 3 million (about 7 percent) of the total 44.5 million Medicaid enrollees were reported as “unknown.” Furthermore, CMS does not know how the racial and ethnic data categories collected through each state’s eligibility processes are translated into the MSIS categories for race and ethnicity. For example, if a state combines the categories Asian and Pacific Islander in collecting information on enrollees and potential enrollees, it is
not clear to CMS how the state reports the data for the MSIS categories, which are separate for Pacific Islander and Asian status.
Since Medicaid eligibility is based in part on low-income status, only those below the income eligibility or “medically needy” thresholds in each state qualify for the program and are included in the data system.7 Income and resource information is collected during the enrollment process and so MSIS includes a field for enrollees’ income and for information on resources if states have further resource tests for eligibility. However, these data are not consistently reported to the federal government. MSIS does not contain a field to collect information on language use.
There are no requirements to report individual data for SCHIP, although states are required to report aggregate information to the federal government. If states use SCHIP funds to expand their Medicaid programs, then individual enrollment data (including racial and ethnic data) are reported to the federal government through MSIS. While states are required to report race and ethnicity in the aggregate, only about one third of all states do so (personal communication with CMS staff). Because there are no standard racial and ethnic categories for the enrollment forms, the quality of the data aggregated at the state level is suspect. Increased cooperation between the Medicaid and public health agencies would provide many states with opportunities to coordinate data collection goals and interventions to improve disparities in the health status of racial and ethnic minority groups. SCHIP does collect income information on its enrollment forms because it is needed to determine eligibility. However, eligibility rules differ from state to state, and there are no standardized ways to collect the data. These data are not reported to the federal government either. No language or acculturation data are collected for federal reporting of SCHIP data.
State and local governments maintain a wide array of data systems that can be used to better understand the health and health care of their constituent populations. Many of these data systems also play a crucial role in providing data for national-level health and health care research. Because these state-based data systems provide a substantial portion of the data
available on health and health care services, they have great potential for furthering the understanding of disparities in health and health care and for the design and evaluation of interventions to eliminate them.
But the collection of data on race and ethnicity in these state-based systems is inconsistent. While the Medicaid program, through its new MSIS, will now collect data on race and ethnicity in a consistent way, SCHIP does not. Hospital discharge systems also do not uniformly collect racial and ethnic data across states; some states mandate the collection of these data as a part of their hospital discharge data reporting, others do not. Racial and ethnic data are not consistently collected in cancer registries, either. Both the SEER and NPCR systems rely on medical records for the information gathered in these systems. The racial and ethnic data that exist in these medical records are used to aggregate individuals into broad racial and ethnic classifications that are not consistent with the OMB standards. Thus, many opportunities to use these state-based systems for research on disparities in health and health care are missed because states do not require that the data be collected in these systems or that they be collected in a consistent manner.
State requirements to collect data on race and ethnicity in these systems could improve states’ abilities not only to identify and monitor disparities in health and health care but also, more importantly, to design and evaluate intervention programs to eliminate the disparities. States can tailor their own requirements to ensure that data are collected on specific populations of interest in their states. By requiring standardized reporting of racial and ethnic data, states may also improve the completeness and consistency of the data collected. For example, not surprisingly, states that mandate the collection of data on race and ethnicity on hospital discharge abstracts have more complete reporting of these data than states without mandatory reporting (97 percent of abstracts in states with mandatory reporting contained data on race and ethnicity, compared with 83 percent in states with nonmandatory reporting) (AHRQ, 1999).8
At the same time, state requirements for the collection of racial and ethnic data would benefit research on health and health care disparities at the national level. Data from hospital discharge abstracts and from cancer registries are important especially for research on disparities in health care, an area where federal sources of data are lacking. Data from Medicaid and SCHIP, if they can be linked with claims data as the MSIS hopes to do, can
also provide vital information on the health care of low-income individuals as well as for different racial and ethnic groups.
The collection of data on SEP and language use and acculturation is not standardized in state-based data collection systems and these data are, in fact, rarely collected at all. This is probably in part because the settings in which data from vital records, hospital discharge abstracts, and cancer registries are collected are often not well suited for collecting extensive amounts of data, especially for concepts as difficult to measure as economic resources and degree of acculturation. However, some simple measures of SEP, such as education level or occupation, could be collected fairly easily in these data systems. In addition, states should examine the feasibility and potential usefulness of collecting simple measures of language use and acculturation, such as place of birth or generation status.
RECOMMENDATION 5-1: States should require, at a minimum, the collection of data on race, ethnicity, socioeconomic position, and, where feasible, acculturation and language use.
The panel recognizes that there are barriers to states imposing data standards where none existed before. For example, there are costs to changing reporting and computer systems when new standards are imposed. Further, the actual recording of an individual’s racial and ethnic data is often done by those who are not survey interviewers by training but rather work in health care or program administration (e.g., medical records clerks, providers and health care workers, or funeral directors). Some training in recording these data is often warranted, however, so that they are consistently captured. For instance, admitting clerks or eligibility workers may code a client’s race and ethnicity through observation, making an assumption based on external characteristics such as surname, accent, skin color, or other features. Others may be trained to ask patients directly for their race and ethnicity, using a form or verbal interview. But training frontline workers is not a one-time effort—some of these positions experience high turnover rates so that the need for training is almost constant. Finally, getting political support for the collection of data on race and ethnicity during the health plan enrollment process or health care encounters requires education of both health system administrators and the general public about the benefits and risks of capturing and providing these data, and clear explanations of how they will be used and not used are essential. DHHS should therefore provide states with guidance and support to educate the public about the benefits of the data and to train administrative and medical records personnel to improve the recording of racial and ethnic data.
Many data collection systems have incomplete data because respondents refuse to answer questions about race, ethnicity, language, or SEP or
because recorders fail to request or ascertain the information. How these missing data should be handled in analysis (e.g., by imputation or by supplementation with other data) is an issue on which many states need guidance. States should in any case continue to implement the new OMB standards for the collection of racial and ethnic data as these standards provide both a framework for collecting the data and the flexibility to capture more detailed data.
That said, several technical issues arise from the implementation of the standards. First, states need to serve their own populations first and foremost and so the categories of race and ethnicity that they use may or may not need to be more specific than the minimum OMB categories. However, there are great benefits to implementing at least the minimum categories for promoting multistate and national-level analyses of these data. Second, many of the data systems maintained by states derive from federal programs (Medicaid and SCHIP) or are part of federal/state cooperative data collection programs (e.g., vital records, NPCR, and SEER). Medicaid, SCHIP, and SEER do not require states to report data on race and ethnicity according to the OMB standards. Third, in order to understand changes in health disparities over time, comparisons across different reporting category systems will be necessary. The process of bridging the new categories to the old categories to allow comparisons of races and ethnicities over time is a technical issue that will need to be addressed. Furthermore, the new OMB standards allow individuals to choose multiple races and ethnicities in responding to questions, and how these multiple responses are converted into single category responses is another challenge that states will face. Much work on these technical issues has already been conducted by federal statistical agencies (Parker et al., 2003). DHHS should draw on this work to develop guidance for states on how to address these issues.
The collection of racial and ethnic data is beneficial only to the extent that the data are used to improve quality of care, administration of state programs, and the broad health status of state populations. It is important for key stakeholders in the data collection process to see that the data are being used for important and appropriate purposes. Individuals providing data are likely to feel better about doing so if they see that the data are being used in beneficial ways. Medical records clerks and providers that collect the data will also want to see them used so that they know their efforts are worthwhile. For all these reasons it is important, for both state and federal government collection of racial and ethnic data, that the usage of these data in reports and policy analysis be publicized. The value of the data also increases if they are made widely available to those who can use them. Health departments and policymakers are obviously the first layer of users at the state level, but researchers and analysts from a variety of settings (academia, federal government, local governments, and advocacy
groups) could also use the data to effectively improve knowledge of health disparities. At the same time, privacy and confidentiality protections should be in place and enforced to guard against improper uses.
Some states have much experience in the collection, use, and dissemination of data on race, ethnicity, SEP, and language use and acculturation. Other states do not and could learn much from those that do. DHHS can exercise leadership in encouraging states’ collection and use of existing data. DHHS uses the data it collects to produce a number of reports on health disparities, including such efforts as Healthy People 2010 and the National Healthcare Disparities Report. These and other efforts by DHHS agencies to implement research programs to better understand health disparities can serve as useful illustrations of the importance of how data on race, ethnicity, language, and SEP can be used. The federal government can also facilitate the sharing of information about how different states that collect the data use them in policy environments and make them available for outside users. Through these examples as well as the convening of conferences and the provision of training and technical assistance, DHHS could exemplify beneficial uses of this type of data.
RECOMMENDATION 5-2: DHHS should provide guidance and technical assistance to states for the collection and use of data on race, ethnicity, socioeconomic position, and acculturation and language use.
DHHS can consider a number of ways to support states in their collection of these data. One possible way for DHHS to encourage states to collect and use racial and ethnic data is by providing incentives to states in the form of competitive matching grants that target data collection for interventions to minority groups. An example of one such federal-state partnership is the Title V Maternal-Child Health Program (MCH), which provides matching funds to states and requires quantifiable outcomes monitoring, supported by the state data systems. One of the core MCH measures required by Title V is the Black and White Infant Mortality Ratio Performance Measure. The Vital Statistics Cooperative Program is another example of a mechanism with which DHHS can encourage states to collect and use racial and ethnic data in a standardized way. In this program, states are funded by the NCHS to provide standardized data on basic data items cooperatively agreed to by states and the NCHS.
To address gaps in coverage in many administration records, some states link public health data sets. For example, researchers from the New Hampshire State Department of Health and Human Services link data from the New Hampshire State Cancer Registry to state hospital discharge data (Taylor and Liu, 2003). In Ohio, Koroukian, Cooper, and Rimm (2003) linked Medicaid claims data with state cancer registry data to study the accuracy of Medicaid claims data in identifying breast cancer cases. Other
states link Medicaid enrollment and claims data with hospital discharge data and cancer registries; Medicaid enrollment data with birth certificate data to understand prenatal and birth outcomes; immunization registry data with data from the Special Supplemental Nutritional Program for Women, Infants, and Children (WIC) and from birth records to fill in data gaps in the registry; and immunization registries with Medicaid and health insurance plan enrollment files for Health Plan Employer Data and Information Set measures. Although none of these linkages is done specifically to fill in gaps in information on race, ethnicity, SEP, or acculturation and language use, similar linkages could be attempted to use the strengths of some data sources in measuring these characteristics to supplement what is not collected in other data sources.
In order to meet the data collection and analysis needs for measuring and reducing disparities, the public and private sectors will need to coordinate efforts to develop state data and research agendas and design effective interventions. No sector can address this by itself.