Race, Ethnicity, and Language Data: Standardization for Health Care Quality Improvement
2. Evidence of Disparities among Ethnicity Groups
Research studies help provide an understanding of the extent of the health and health care disparities experienced by different racial and ethnic groups. While the Office of Management and Budget (OMB) race and Hispanic ethnicity categories can reveal many inequities, they also mask important disparities in health and health care. More discrete ethnicity groups, based on ancestry, differ in the extent of risk factors, degree of health problems, quality of care received, and outcomes of care. More granular ethnicity data could inform the development and targeting of interventions to ameliorate disparities in health care that contribute to poorer health.
The Institute of Medicine's landmark report on racial and ethnic disparities in health care, Unequal Treatment: Confronting Racial and Ethnic Disparities in Healthcare, emphasizes the need for standardized collection and reporting of race and ethnicity data (IOM, 2003). While Unequal Treatment recommends the Office of Management and Budget (OMB) race and ethnicity categories as the minimum standard by which collected race and ethnicity data should be parsed and reported, the recommendations go further, calling for better data on racial and ethnic populations "to reflect the diversity within racial and ethnic populations (e.g., subgroups of Hispanics, African Americans, Asian Americans, etc.), particularly at the local level" (IOM, 2003, p. 233).
Since the release of Unequal Treatment, evidence of disparities in health and health care among racial categories at the broad OMB level (Black or African American, Asian, Native Hawaiian or Other Pacific Islander [NHOPI], White, and American Indian or Alaska Native [AIAN]) has continued to be documented. Similarly, distinct differences continue to be shown between the broad Hispanic and non-Hispanic ethnic categories. For example, there is more information on varying life expectancy (IOM, 2008) and mortality risks or rates for certain medical conditions (Murthy et al., 2005; Wang et al., 2006), along with knowledge of disparities in general health status, access to health care, and utilization rates of services among these larger population categories (AHRQ, 2008a; Cohen, 2008; Flores and Tomany-Korman, 2008; Kaiser Family Foundation, 2008, 2009; Ting et al., 2008). Even as quality-of-care indicators such as screening for colorectal cancer show improvement for the overall population, disparities persist among the OMB race and Hispanic ethnicity categories (AHRQ, 2008a, 2008b; Moy, 2009; Trivedi et al., 2005).
In contrast, systematic analysis of similar quality-related data as a function of more discrete ethnic groups within the OMB categories has hardly progressed. After defining the term granular ethnicity, this chapter summarizes the evidence showing health and health care disparities at more fine-grained levels of ethnic categorization. The literature has more to say about ethnicity and disparities in health than about ethnicity and disparities in health care; this is reflected in the balance of articles reviewed in this chapter. To complement the research studies, data are also presented for selected population characteristics that can place people at risk of disparities (e.g., low education levels, poverty, lack of facility with English among those speaking a non-English language at home, and place of origin).
This focus on literature with respect to more granular detail on subgroups is not to negate the important differences found among the OMB racial groups and for Hispanics compared with non-Hispanics, but to learn more about where to focus interventions when categorical differences are masked by the OMB categories. Being able to focus interventions at the more granular level has been posited as a way to use resources most efficiently to reduce disparities.
Awareness of health and health care disparities has been heightened through the release of multiple documents besides Unequal Treatment, including—Healthy People 2010 and the National Healthcare Disparities Reports (AHRQ, 2008a; HHS, 2000), and successful initiatives have addressed some disparities using a variety of approaches. For example, some successful initiatives have applied general quality improvement concepts and techniques, while others have developed and used culturally sensitive outreach and education materials for health plan members, and still others have involved training of staff in culturally competent communications. Common to virtually all successful projects are some fundamental steps, including the acquisition of data on race and ethnicity, the stratification of quality-of-care data by race and ethnicity, the use of race and ethnicity to identify members of a target population to whom elements of an intervention would apply, and reanalysis of stratified quality data to evaluate the impact of the activities. Data on race and ethnicity are a fundamental requirement for disparity reduction initiatives. Without these data, it is impossible to identify disparities and track the impact of initiatives over time, and it is difficult to target those aspects of interventions that involve direct contact with individuals. The presence of data on race and ethnicity does not, in and of itself, guarantee any subsequent actions in terms of analysis of quality-of-care data to identify disparities or any actions to reduce or eliminate disparities that are found. The absence of data, however, essentially guarantees that none of those actions will occur.
Defining Racial and Ethnic Populations in the United States
The United States is a diverse country whose composition is changing. Table 2-1 shows the results of Census 2000 on the size and percentage distribution of the total U.S. population primarily by the broad OMB racial and Hispanic ethnic groupings. The Black and Hispanic groups are of equivalent size; the Census has multiple check-off boxes for specific Hispanic groups (i.e., Mexican, Puerto Rican, Cuban, and a write-in option for other groups) that it routinely reports, but there are no such more specific check-off boxes under the Black or White races. Asians and Pacific Islanders have many specific groups listed on the Census form from which to choose as well. There are efforts to legislatively mandate expansions to the current Census categories (e.g., add Caribbeans in general and Dominicans specifically).1 The groups included in the OMB race and Hispanic ethnicity categories are defined in Chapter 1 (go to Table 1-1).
Ethnicity is a concept that the subcommittee, for standardization purposes, distinguishes from race. The term ethnicity represents a common ancestral heritage that gives social groups a shared sense of identity that exists even though a particular ethnic group may contain persons who self-identify with different race categories. The OMB categories use the term ethnicity only in conjunction with Hispanic ethnicity. The U.S. Census captures data on a few discrete ethnic groups both under the Hispanic ethnicity question, by having check-off boxes for some Hispanic groups (e.g., Puerto Ricans, Dominicans), and under the race question, by listing some groups of Asian and Pacific Islander heritage (e.g., Japanese, Samoan) and leaving an option for American Indian and Alaska Natives to indicate a tribal affiliation.
Where one is born can make a significant difference in access to and use of health care, but the subcommittee adopts the concept of ethnicity (equated with one's ancestry) as more encompassing than questions about country of birth or origin. A person born in the United States might identify culturally with a specific ethnicity in ways that can affect his or her health-related behaviors and approach to utilizing health services. Also the subcommittee prefers the use of ethnicity over questions such as national origin because inquiring about national origin could engender mistrust on the part of respondents that they are being asked about immigration status (Carter-Pokras and Zambrana, 2006).2
Defining Granular Ethnicity
Granularity means a fine level of detail; the greater the level of granularity, the more finely detailed the data category is. The subcommittee adopts the term granular ethnicity to describe groups at a more specific level of categorization than the broad OMB categories, such as the ethnic groups that the Census lists as subgroups in its Hispanic ethnicity and race questions. The subcommittee, as will be examined in Chapter 3, believes a separate question on granular ethnicity would complement the OMB categories for race and Hispanic ethnicity without further intermingling the constructs of race and ethnicity. Additionally, this approach would allow more discrete categorization of large groups of the population who now have the option only of White or Black on the race question.
The term granular has been used in describing more detailed categories in the Hospital Research & Educational Trust (HRET) Toolkit (Hasnain-Wynia et al., 2007), and the notion of the need for more detailed subgroup data has been raised in Unequal Treatment and by many others. Kaiser Permanente also uses the term granular ethnicity in describing its collection of more detailed information beyond the OMB categories (Tang, 2009). More detailed ethnicity categories provide a useful way of analyzing quality data about the populations served by providers, health plans, state and federal programs, and others to determine whether there are differential health needs and disparities in access to and use of appropriate health services. The level of detail for analysis for quality improvement can be influenced by the size of the ethnic population under study; the number or proportion of those ethnicities that might have a specific condition such as diabetes or be of an age at which immunization for pneumonia is needed; and the actual associations among ethnicity, other correlated factors (e.g., income, insurance coverage), and quality of care. While there are hundreds of possible ethnic categories, not all will have local relevance nor always have added value for designing targeted approaches to remediate health care needs. This report's recommendations are driven by a need to identify and address quality differentials not simply to collect information to classify and count people.
Overview of Differentials in Care and Potential Quality Improvement Interventions
Health is the physical, mental, and functional status of an individual or a population. Health has been shown to be the result of multiple factors, including nutrition, educational level, socioeconomic level, and lifestyle, and of the health care that the individual or population receives. Health care comprises the prevention, treatment, and rehabilitation interventions that are provided to an individual to maintain or improve health. Disparities in health care (e.g., in access, in the rate at which a treatment is provided when indicated, or in the incidence of adverse events in care) can be the cause of disparities in health (e.g., in the incidence or severity of a disease, in functional level, or in mortality rate). Therefore, analyses of disparities in health care can help identify opportunities for quality improvement in care provision that will reduce disparities in health. For the most part, entities use the same categories of race, ethnicity and language whether data are collected for health or health care purposes so the connections between health disparities and health care disparities can be drawn more easily.
Illustration of Differences Among Ethnic Groups Within Broad OMB Categories
A study by Blendon and colleagues (2007) illustrates the concept of differences among subgroups residing in the United States, even after controlling for demographic characteristics such as income, education, age, and sex. A number of differences in health care service utilization and satisfaction can be seen among more granular Black, Asian, and Hispanic ethnic groups. Blendon and colleagues' telephone survey of 4,157 randomly selected adults in the United States found that fewer Caribbean- and African-born Blacks received any care than U.S.- born African Americans in the past year but it was the latter group that rated their care more poorly than Whites. Certain Hispanic American groups (Mexican and Central/South American Hispanic) and Asian American groups (Chinese, Korean, and Vietnamese) also received significantly less health care in the last year compared with Whites, even though other ethnicities within these broad OMB race and ethnicity categories fared as well as Whites. Native Americans also received less care compared with Whites and less often rated their care as good or excellent—the lowest rating of any of the groups. Regressions that controlled for demographic characteristics reduced the number of groups receiving no care in the past year by half, but significant differences remained for African-born Americans, Mexican Americans, Chinese Americans, and Korean Americans compared with Whites that were independent of the demographic factors (Blendon et al., 2007). While for some groups the access and utilization issues may stem from economic challenges, the reality remains that there are differences among ethnic groups in utilization and ratings of caregiving within the broad OMB categories.
Potential Applications for Quality Improvement
Cooper and colleagues (2002) review a variety of successful interventions, and note that while there are many well-identified potential opportunities for certain conditions and services, there is a lack of information on "ethnic subgroups." They also stress the need to improve the science of evaluating interventions to reduce disparities now that there is widespread acknowledgment of the existence of inequalities. A fundamental component of improving quality is collecting reliable demographic data to use in focusing attention on where interventions might be best applied.
Fiscella also observes that, "because disparities in healthcare represent inequities in the process of healthcare, they are potentially addressable through interventions designed to impact health delivery" (Fiscella, 2007, p. 142). Entities that collect race and detailed ethnicity data might use them in various ways to examine whether there are differentials in health care needs and to plan targeted interventions. For example, having read in published research that certain ethnic groups are at higher risk for cancer mortality and delays in care, a health plan could target educational calls to persons of these ethnic groups to make screening appointments for different site-specific cancers rather than having to contact a much larger number of persons (Bates et al., 2008). Or a hospital could look at the characteristics of patients who did not receive care according to evidenced-based protocols for acute myocardial infarction. Then the hospital could assess whether there were specific barriers that interfered with the appropriate delivery of care to specific populations and make concerted efforts to remove those barriers. Or the hospital might also want to take what it learned from that effort to institute strategies that could be applied universally to ensure that all patients with that condition receive the right care at the right time. Another hospital might be experiencing a high readmission rate; analysis of its readmission data might reveal a higher than expected rate for a specific ethnic group. From there, the hospital could determine whether culturally specific interventions at discharge planning are necessary to prevent unnecessary readmissions, and whether this patient group needs access to regular primary care. Similarly, a health center might find that women of a certain group are not coming in for prenatal care until late in their pregnancy; this finding could lead the health center to send community health workers out into the community to change attitudes and practices related to seeking timely care. Physicians receiving feedback on their practice patterns might discover that they are not giving the same evidence-based care to all patients, even though they believe they are, and when this is called to their attention, their practice improves. Fiscella reviews a variety of quality improvement tools, including reminders, provider feedback, provider education, intensive outreach, practice guidelines, patient education, cultural competency training, and organizational change/practice redesign and community-based interventions, and concludes that "the elimination of healthcare disparities will require the development and implementation of tailored interventions directed at multiple levels. Success will depend on the vision, leadership commitment, and allocation of resources by government, health plans, hospitals, communities, and practices..." (2007, p. 164).
The following sections examine further evidence of differences within the aggregate OMB categories. These studies are illustrative of how more granular ethnicity data reveal more precise opportunities for targeting health care quality improvement initiatives.3 Notations are made when the studies are controlled for socio-economic factors when comparing health or health care differences among populations. Statistically significant associations and trends are emphasized.
Hispanic or Latino Groups
In Census 2000, 12.5 percent of the U.S. population (35.2 million people) self-identified as Hispanic, with persons of Mexican origin representing the largest ethnicity group at almost 60 percent of the Hispanic population (Ramirez, 2004). Hispanic is the one distinct ethnicity included in the OMB basic categories and is defined by the Census and OMB as a "person of Mexican, Puerto Rican, Cuban, South or Central American, or other Spanish culture or origin regardless of race" (OMB, 1997; Ramirez, 2004). The question about Hispanic ethnicity used by the Census includes additional labels, such as Latino and Spanish, to delineate more clearly who is included since different people identify with one of the terms but not the others.
This ethnic category usually has been subdivided in the literature according to ancestry or according to regional designations of South and Central America (Table 2-2).4 From this table, one sees that individual Hispanic groups5 have different characteristics with respect to U.S. nativity, proficiency with English, educational attainment, and risk of poverty?factors that have been shown to impact the quality of care those populations receive and their health outcomes. More than 40 percent of most ethnic groups who speak Spanish at home do not speak English very well, and some groups have almost twice the poverty rate of others (Ramirez, 2004).
Health-Related Differences Among Hispanic or Latino Groups
Differences in dimensions of health and health care among specific Hispanic or Latino populations in the United States have been identified and studied more extensively than other racial and ethnic populations. The available literature includes studies of health and health care disparities between Hispanic groups by overall self-rated health, access to care, mental health, cancer and cancer screening, low birthweight, asthma, and cardiovascular health.
Overall Self-Rated Health
In a national study comparing the overall mental and physical health of multiple Hispanic ethnicity groups, the Mexican group tended to have better scores on both components of the SF-12 than Whites and other Hispanic groups, whether those of Mexican ancestry were born in the United States or Mexico (Jerant et al., 2008). The study is based on cross-sectional analyses of linked data from the 1998-2004 National Health Interview Survey (NHIS) and the 1999-2005 Medical Expenditure Panel Survey (MEPS); the study population compared four Hispanic groups—Mexican (13,522 persons), Cuban (778), Puerto Rican (1,360) and Dominican (829) including persons born in the United States and elsewhere—with 45,422 English-speaking Whites born in the United States. After regressions adjusting for demographic and socioeconomic variables, those of Cuban ancestry had the worst mental health scores, while those of Puerto Rican heritage had the worst physical health scores; the scores for Cuban, Puerto Rican and Dominican groups on both components were worse than Whites. The authors' suggest that the "paradox" of better health status among the Mexican group even with low socioeconomic status can mask poorer health status of other smaller groups of Hispanics when the Hispanic data are examined as one group. The authors also underscored that the observed ethnic differences within the Hispanic groups on the mental health component met a criterion for clinical significance.
Access to Health Care Services
Shah and Carrasquillo (2006) used cross-sectional analyses of the Census Bureau's Current Population Survey (CPS) to examine differences in insurance coverage, focusing on Hispanic populations. As of 2004, those identifying with the Mexican ethnicity category had the highest rate of uninsurance (35.6 percent), and the Puerto Rican category the lowest rate (17.6 percent), with Cuban (22.1 percent), Dominican (25.3 percent) and other Hispanic groups (32.5 percent) having intermediate values (Shah and Carrasquillo, 2006). The socioeconomic profile of the groups did not always parallel the rate of uninsurance, for example the subgroups with the greatest proportion under 200 percent of poverty were Mexican and Puerto Rican. Weinick and colleagues (2004) using MEPS data similarly showed that persons identifying with Mexican ethnicity had higher uninsurance rates than Cuban and Puerto Rican groups, but persons with Central American and Caribbean ethnicities had even higher rates of uninsurance than the Mexican group.
Additionally, Weinick and colleagues (2004) examined differences in use of four health care services (ambulatory care visits, emergency department [ED] visits, prescription medications, and inpatient hospitalizations). After controlling for sociodemographics, including income and health insurance coverage, multivariate regression analyses of MEPS data showed that persons of Mexican and Cuban ancestry had lower rates of ED visits than other Hispanics. Additionally, more recent immigrants were less likely to have made any ambulatory care or emergency department visits in the past year. The English-speaking subgroups had a higher rate of ED visits and hospitalizations, and foreign-born Hispanics showed lower rates of ambulatory visits, ED visits, and prescription medications. Based on these results, the authors concluded that understanding disparities in health care utilization will require disaggregation of patient demographic data by ethnic groups, language, and length of U.S. residence (Weinick et al., 2004).
Alegría and colleagues (2007) examined the prevalence of depressive, anxiety, and substance use disorders among Hispanics living in the United States using data from the National Latino and Asian American Study (NLAAS).6 Weighted logistic regression analyses controlled for age. In terms of lifetime prevalence, compared with the comparable Puerto Rican gender group, those of Mexican ethnicity showed lower rates of depressive disorders whether male or female and lower rates of substance abuse disorders for women, and lower overall psychiatric disorders for men. Cuban men were less likely to suffer from anxiety disorders and overall psychiatric disorders. Puerto Ricans tended to have the highest rates of lifetime and past year depressive, anxiety, substance use, and overall psychiatric disorders. Looking at all Hispanic groups in combination, those with higher English proficiency were significantly more likely to suffer from overall lifetime or past year psychiatric disorders than those with fair or poor English skills.
Cancer and Cancer Screening
Gorin and Heck (2005) used the 2000 NHIS to examine data from 5,377 Latinos on the use in the past 12 months of Pap smears, mammograms, breast self-examinations, and clinical breast exams among women; prostate-specific antigen (PSA) tests among men; and fecal occult blood tests (FOBT), sigmoidoscopy, colonoscopy, and proctoscopy among both men and women. Cancer risk factors such as smoking varied by ethnic group (e.g., over 25 percent of Puerto Rican and "other" Hispanics smoked while 13.9 percent of Dominicans did). For persons of average risk for cancer (i.e., did not have a personal or family history of cancer), ethnic group variations were apparent in use of Pap smears and clinical breast exams, but differed less on some tests such as FOBT where use was low for all groups. Multivariate logistic regression analyses revealed that Dominican women were 2.4 times more likely to have had mammography than other Latino women. Puerto Rican and the Central or South American groups had half the rate of colorectal cancer screening by endoscopy of others. Cuban males were five times more likely to have had a PSA test. Additionally persons with health insurance were 1.5 to 2.2 times as likely to have screening tests compared with the uninsured. Having visited a doctor in the past year, increased the odds of having screening tests to a level similar to having insurance, with the exception of PSA screening where the odds were almost five-fold greater. Greater acculturation,7 visits to a primary care provider, and use of other screening tests, predicted the likelihood of Pap smear screening. Clinical breast exam rates were also predicted by greater acculturation, visits to a primary care provider in the last month, and use of other screening tests, along with having a bachelor's degree and a personal history of cancer (Gorin and Heck, 2005).
Using multiple logistic regression analyses of NHIS data pooled from 1990 and 1992, Zambrana (1999) compared the use of three cancer screening practices (Pap smear, mammogram, and clinical breast exam) for five categories of Hispanic women including women who identify as Mexican versus Mexican-American. While Mexican women were the least likely to have been screened in the past three years, no statistically significant differences were found in the rates between the Mexican-American (referent group) and any of the other Hispanic groups. In this study, access measures such as having a usual source of care and knowledge of other clinical cancer screening techniques were more strongly associated than ethnic or language factors with screening rates for the population studied (Zambrana et al., 1999). The authors posit that the higher than expected rates of screening in the sample population may be attributable largely to contemporaneous intervention strategies and community outreach to increase screening among Hispanic women, concluding that such efforts appeared effective and should be expanded.
The National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) data from 1992-1995 showed that while all Hispanic women had a greater likelihood of larger tumor size and advanced tumor stage than non-Hispanic Whites, women born in Latin America had higher odds of large tumors (e.g., larger than 1 cm and 2 cm) than Hispanic women born in the United States (Hedeen and White, 2001). The researchers were only able to identify the ethnic subgroup for 38 percent of the Hispanic women in the SEER database.
Logistic regressions on 2002 U.S. Natality Detail Data (n = 634,797) showed that after controlling for a variety of demographic, educational and clinical factors, foreign-born Latino mothers had a lower risk of having low-birth-weight infants compared with U.S.-born Latino women. However, nativity patterns among Mexican-origin women explained these overall trends among Latino women and infants. Foreign-born women with Mexican ethnicity had about a 21 percent reduced risk of low birthweight, but the same phenomenon was not observed for other Latino women who were born outside the continental United States (i.e., Puerto Ricans, Cubans, Central/South Americans) (Acevedo-Garcia et al., 2007). Across each of the three regression models, Puerto Rican women had higher odds than other Hispanic subgroups of having a low-birthweight infant. The regression models for this study did not control for income or insurance status.
Large differences also exist in asthma burden among Hispanic children. Based on weighted logistic regression analyses of merged 1997-2001 NHIS data, Puerto Rican children had the highest prevalence (26 percent) and rate of recent asthma attacks (12 percent) compared with children of Mexican heritage whose prevalence and recent attack rates were 10 percent and four percent, respectively (Lara et al., 2006). Rates for Cuban and Dominican ethnicities were intermediate and similar to Black children. Adjusted odds ratios followed the same relative pattern among Hispanic subgroups (e.g., lifetime odds of 2.3 for Puerto Rican children vs. 0.90 for Mexican children compared with the non-Hispanic White referent group). Birthplace influenced the association between ethnicity and lifetime asthma diagnosis differently for Puerto Rican and Mexican children. When both Puerto Rican children and their parents were born in the continental United States, the adjusted odds ratio (OR) was 1.95 (95 percent CI 1.48-2.57) but 2.5 (95 percent CI 1.51-4.13) for those who were island-born; the odds ratios were calculated using as the referent group U.S.-born non-Hispanic White children whose parents were born in the United States (Lara et al., 2006). In contrast, U.S.-born Mexican families had a higher adjusted OR for lifetime asthma diagnosis of 1.05 (95 percent CI 0.90-1.22) than the 0.43 (95 percent CI 0.29-0.64) for those born outside of the continental United States. Similar patterns were observed for recent asthma attacks. Birthplace was the only co-variant that affected the Hispanic subgroup results; numerous factors were considered including family income and insurance status. Overall Hispanic data mirror the Mexican ethnicity data, thus masking the results for Puerto Rican children.
Borrell and Crawford used NHIS data (1997-2005) to perform descriptive and logistic regression analyses assessing the strength of association between Hispanic ethnic groups and self-reported hypertension; self-report was based on the question of whether they had ever been told by a health professional that they had hypertension. Dominican ethnicity and non-Hispanic Black adults had an adjusted odds ratio of 1.67 and 1.48, respectively, compared with the referent group of non-Hispanic Whites. Results were adjusted for age, sex, marital status, survey year, U.S. region, nativity status/length in the United States, health insurance, education, income, and occupation. In contrast, persons of Cuban, Central or South American, Mexican (whether born in the United States or not), and other Hispanic groups all had lower odds than non-Hispanic Whites or Blacks or those of Dominican ethnicity (Borrell and Crawford, 2008).
Another study examined hypertension-related mortality rates among women of various Hispanic subgroups using data from the National Vital Statistics System's Multiple Cause Mortality Files and further tracked whether changes occurred over time (1995-1996 to 2001-2002). In 1995-1996, the age-standardized death rate per 100,000 for hypertension-related mortality was higher among the Puerto Rican group (248.5) than for non-Hispanic Whites (188.7), while Mexican American (185.4), and Cuban (139.7) rates were lower. Over time, the mortality rate decreased for Puerto Rican (215.5), non-Hispanic White (171.9), and Cuban American (104.6) women, with each group keeping their relative position. At the same time the rate for Mexican American women increased to 205.5, now making their risk higher than non-Hispanic White women. The authors suggest the need for strengthening interventions to reach these higher risk ethnicity groups and those who provide their care (Zambrana et al., 2007).
In the broad Hispanic ethnicity category, more granular ethnicities are associated with different levels on health indicators and access to and utilization of health care depending on ancestry. The authors of the studies reviewed in this section stress the importance of not viewing the Hispanic population as monolithic, and they point out the masking effect that the larger Mexican ethnicity group has on overall statistics when data are viewed to represent all Hispanic groups as one. Even after adjustment for factors such as insurance, education, and income, many ethnic differences were found to remain. The authors also comment on how Hispanic populations beyond Mexican, Cuban and Puerto Rican ethnicity are not well characterized, because in surveys their numbers are small resulting in heterogeneous groups being lumped into an "other" Hispanic category.