Cancer screening and risk factor rates among American Indians.

Objectives. We examined cancer screening and risk factor patterns in California using 4 different statistical tabulations of American Indian and Alaska Native (AIAN) populations.

Methods. We used the 2001 California Health Interview Survey to compare cancer screening and risk factor data across 4 different tabulation approaches. We calculated weighted prevalence estimates by gender and race/ethnicity for cancer screening and risk factors, sociodemographic characteristics, and access to care variables. We compared AIAN men and women with members of other racial groups and examined outcomes among AIAN men and women using the 4 tabulation methods.

Results. Although some differences were small, in general, screening and risk factor rates among American Indians/Alaska Natives were most similar to rates among Whites when the most inclusive multiracial tabulation approach was used and least similar when the more exclusive US census "single-race" approach was used.

Conclusions. Racial misclassification and undercounting are among the most difficult obstacles to obtaining accurate and informative data on the AIAN population. Our analysis suggests some guidelines for overcoming these obstacles.

Despite declines in mortality from cancer among the overall US population in recent years, mortality rates among American Indians and Alaska Natives have not declined significantly.(n1) Moreover, a persistent gap in self-reported health status remains between American Indians/Alaska Natives and non-Hispanic Whites.(n2) National health indicators for American Indians/Alaska Natives rank near the bottom among the major racial/ ethnic groups, including Whites, Blacks. Asian Americans, Native Hawaiians and other Pacific Islanders, and Latinos.(n3) For example, overall cancer mortality rates are rising among American Indians/Alaska Natives who are residents of California,(n4) even though their rates of Papanicolaou test and colorectal screening are among the highest of the major racial/ethnic groups.(n5)

A recent evaluation of REACH 2010 (Racial and Ethnic Approaches to Community Health 2010) showed that American Indians were more likely than other minority groups to use preventive services; at the same time. however, they exhibited higher levels of risk factors such as smoking and obesity.(n6) The study also showed that American Indians have reached or are close to reaching the mammography and Papanicolaou test objectives established by Healthy People 2010,(n7) an achievement the REACH authors credited to the commitment of communities, tribal corporations, public health authorities, and health care providers. However, as pointed out by the authors, the REACH 2010 study included only 2 American Indian communities and might not have been representative of American Indians/Alaska Natives from other areas.

We explored cancer screening and risk factor patterns among American Indian/Alaska Native (AIAN) women and men. Racial misclassification has been one of the most difficult obstacles to accurate and informative data collection and reporting for the AIAN population. The California Health Interview Survey (CHIS) of 2001 is the largest population-based cancer risk factor and screening data source on American Indians/Alaska Natives. The CHIS database offers a unique opportunity to compare several different tabulation methods for documenting the AIAN population surveyed because of the careful and collaborative manner in which the data were collected. To begin to fill the research gap suggested by the REACH 2010 evaluation, we examined whether the same cancer screening and risk factor patterns were evident in California as in the REACH sample and, more specifically, how these data compared when different statistical tabulations of the AIAN population were used.

CHIS 2001 was carried out between November 2000 and October 2001. In addition to the main random-digit-dialing (RDD) sample of more than 54000 adults, CHIS 2001 oversampled American Indians/Alaska Natives in an effort to include a sufficient number of respondents to allow an urban-rural comparison.(n8) The sources used in this over-sampling were a telephone database compiled from California Area Indian Health Service (IHS) clinics, urban Indian clinics and health organizations, and membership directories of AIAN social organizations. The information provided to CHIS was limited to telephone number and city and county of residence. Names and other personal identifiers were neither requested nor provided; the purpose was solely to develop a list of telephone numbers of households likely to include at least one AIAN resident.

The CHIS 2001 Adult Survey AIAN File used in this study pooled data from the RDD sample and the list oversample. RDD sample respondents were included if they reported their race as American Indian/Alaska Native. To be classified in this group, a respondent had to meet at least one of the following criteria: (1) be an enrolled member of a federally recognized or state-recognized tribe; (2) report only AIAN race; or (3) if more than one race/ethnicity was reported, the respondent most identified with American Indian or Alaska Native. No duplicate records were included in the Adult Survey AIAN File.

Because the AIAN population is small, it is particularly important to tabulate it accurately; otherwise, calculated rates can vary widely in the case of rare events such as cancer.(n9, n10) Although the Indian Self-Determination Act (Pub L 93-638) defines an American Indian as a member of an Indian tribe, federal agencies do not use this definition for their databases." Instead, federal surveys ask respondents to identify their race from among several options. In addition, for the First time. the 2000 US census provided the opportunity for respondents to identify multiple races (e.g., American Indian) combined with one or more other races.

Another issue affecting tabulation of the AIAN population is the fact that race and ethnicity are defined and measured separately in US census data. The 1997 revision of Office of Management and Budget Directive 15 distinguished between ethnicity and race, categorizing Latino/Hispanic heritage as ethnicity and White/Caucasian, Black/ African American, Asian/Asian American, American Indian/Alaska Native, and Native Hawaiian or other Pacific Islander heritage as Race.(n12) The revised Office of Management and Budget directive mandates that the US Census Bureau and other federal agencies collect ethnic data before collecting data on race and that these data be collected separately. This is important to recognize when tabulating the AIAN population in California, because a large proportion of American Indians are of Latino ethnic backgrounds or have Spanish surnames.

The standard questions used in the 2000 census, in which respondents were first asked about their Hispanic or Latino origin and then were asked their racial group (or groups), were used to collect race/ethnicity data in CHIS 2001. The survey also asked respondents who reported more than one race or ethnicity whether they primarily identified with one. Respondents had the option of reporting "both," "all," or "none."

The rich data set generated by the CHIS makes it possible to use several different tabulation approaches for classifying race/ethnicity (W. Yen, CHIS data manager, unpublished data, 2004). One approach is to group race data according to the "any mention" method wherein respondents are categorized in each of the race groups they report. This scheme results in replicate counts of individuals across groups (P. Ong, unpublished data, 2003). A second approach is to include individuals in a group only if they report a single race. This was the approach used in the 2000 census. In a third approach, used by the California Department of Finance (DOF) in making state population projections, ethnicity is treated as race. Anyone indicating Hispanic/ Latino origin is tabulated as Latino and not categorized according to race.

A fourth method, used in most reports published by the Center for Health Policy Research at the University of California, Los Angeles (UCLA), which coordinated the CHIS, also combines race and Hispanic/Latino ethnicity but classifies individuals in the group with which they most identify. Latino is defined as a race category: respondents who identify primarily with Latino rather than American Indian/Alaska Native are not included in AIAN sample and population estimates. However, in the UCLA method, those who report enrollment in a federally recognized or state-recognized tribe are included as American Indians/Alaska Natives, whether or not they report they most identify with the AIAN category. To compare the AIAN population with other racial/ethnic groups in California using CHIS data, we applied the UCLA definition because, until recently, this was the automatic default category for Ask CHIS, the online data query system for the survey. The UCLA definition codes each respondent into a single category.

The population estimates for California AIAN adults obtained from the 4 methods just described--any mention of AIAN background ("any mention method"), AIAN origin only ("census method"), non-Latino AIAN origin ("California DOF method"), and primary identification with AIAN category ("UCLA method")--were 413 493, 210937, 64 349, and 114362, respectively. The corresponding CHIS 2001 sample sizes were 3186, 1155, 612, and 1132. It is clear from these figures that the different tabulation approaches result in very different sample sizes and estimates, the any mention method resulting in the largest numbers and the California DOF method the smallest. The UCLA method yielded a smaller sample size and population than the census method because UCLA defines Latino as a race and excludes American Indians/Alaska Natives who identify more strongly with the Latino category, coding them as the latter.

Use of these different methods has implications beyond total population estimates. Analyses have shown that the "most identified" (UCLA) classification tends to generate higher prevalence rates of health conditions (e.g., asthma and diabetes) among American Indians/Alaska Natives than the other tabulation methods, whereas the any mention method tends to generate the lowest.(n13) The multiple tabulation methods available with the CHIS data provide more flexibility for data users, but they also make it important for users to understand the different ways in which American Indians/Alaska Natives can be classified so that they can determine which method is most appropriate for their research hypothesis.

We used CHIS 2001 to examine cancer screening and risk factor data and to compare estimates of these data obtained through the different tabulation approaches described earlier. Our data were based on responses provided by the 3186 adults (2835 from the RDD sample and 351 from the list oversample) included in the CHIS 2001 AIAN Supplemental File. We calculated weighted prevalence estimates according to gender and race/ethnicity for cancer screening and risk factors, sociodemographic characteristics, and access to care variables. First, using the UCLA method, we compared AIAN men and women with members of other racial groups. Then we examined outcomes among AIAN men and women using the 4 tabulation methods. When the 95% confidence intervals of 2 separate estimates did not overlap, we made the conservative assumption that the estimates were significantly different.

Census counts from the 2000 Summary File 2 for California were used to derive CHIS 2001 statistical weights for the AIAN file. To create these weights, the RDD and AIAN list samples were combined and weighted as a single sample through the use of dual frame methodology.(n14) The AIAN file weights were constructed so that the state population estimate matched the California 2000 census population estimates derived from the any mention and census methods. Thus, as a result of the method of weight construction employed, the prevalence estimates obtained from the CHIS 2001 supplemental sample were unbiased when either of these 2 definitions was used.

The CHIS AIAN statistical weights were not constructed explicitly to obtain estimates for either of the 2 other definitions (California DOF and UCLA), If prevalence estimates involving these definitions were of primary importance, a different weighting method would have been necessary. However, it would have been impossible to use results from the 2000 census for the UCLA definition, because the census did not determine "most identified" status. Weighted prevalence estimates for these 2 definitions might be biased solely as a consequence of the weighting methodology.

Tables 1 and 2, respectively, show comparisons of indicators for AIAN men and women and indicators for other major racial/ ethnic groups using the UCLA definition of race/ethnicity (category most identified with). Tables 3 and 4 offer comparisons of the same indicators for AIAN men and women, respectively, using the 4 different tabulation methods, (In the case of the definitions other than "any mention" in Tables 3 and 4, further data are available from the authors that provide rates for American Indians/Alaska Natives in combination with individuals of other racial backgrounds, making the other tabulations more comparable to the any mention method.)

Table 1 shows that Latino and AIAN men were least likely to have a college education and most likely to have family incomes below 200% of the poverty level. Eight percent of AIAN men reported an IHS clinic as their usual source of care. AlAN men were more likely than any group other than Latino men to have no health insurance coverage, and, among those 65 years or older, they were most likely to be at risk of inadequate coverage (Medicare only, "other" only, or no coverage). AIAN and Black men were more likely than men in other groups to have an impairment that prevented them from working for at least 1 year.

Smoking rates among AIAN men were higher than among any other group, with the possible exception of Native Hawaiians/other Pacific Islanders, for whom the confidence intervals were large. The rate of screening for colorectal cancer among AIAN men was similar to that among White and Black men, who had the highest screening rates. AIAN men were less likely than White men to undergo prostate-specific antigen screening.

Among women, fewer American Indians/ Alaska Natives were college graduates than any other group except Latinas, Fewer than 25% of White women had incomes below 200% of the poverty level, whereas rates of poverty among women of all other races and ethnicities were much higher. Among AIAN women, the rate was nearly twice as high (48%) as that among Whites. AIAN women were more likely to have a usual source of care than AIAN men. Similar percentages of AIAN and Asian American women younger than 65 years had no health insurance coverage (approximately 17% in both groups), and the percentages in these groups were surpassed only by that among Latinas (33%). AIAN women 65 years or older, like Latinas and Asians, were at risk of inadequate health insurance coverage. AIAN and Black women were more likely than women in other groups to have an impairment that prevented them from working for at least 1 year.

Smoking rates among AIAN women were higher than rates among women in any other racial/ethnic group, again with the possible exception of Native Hawaiians/other Pacific Islanders, for whom the confidence intervals were large. Larger percentages of AIAN, Latina, and Black women were obese. The rate of Papanicolaou testing among American Indians/Alaska Natives was similar to rates among Latinas, Blacks, and Whites. Rates of colorectal screening among AIAN women were similar to rates among Black and White women, and mammography rates in the AIAN group were lower than the rates among Blacks and Whites and similar to the rates among Latinas, Asian Americans, and Pacific Islanders.

Both AIAN men and AIAN women were more likely to live in rural areas than were those of other races, so we compared rural and urban AIAN men and women (data not shown). Rural men were more likely than urban men to be members of families whose incomes were below 200% of the federal poverty level. Rural men also were more likely to be covered by Medicaid. Urban AIAN women were more likely than rural women to have a bachelor's degree or higher and less likely to he living in poverty. Urban women aged 18 to 64 years were less likely to use an IHS clinic l3ut more likely to have a usual source of health care than their rural counterparts. Finally, more rural than urban women 65 years or older were covered by the IHS.

Tables 3 and 4 present results obtained with the 4 different AIAN tabulations. Several differences can be noted. For example, the overall population of AIAN men and women appeared to be younger when the census definition was used than when the other 3 definitions were used. There were some other statistically significant variations according to age group between the 4 definitions as well. For example, the any mention method yielded a smaller proportion of men aged 18 to 49 years than the census method and a larger proportion than the UCLA or California DOF method.

The census definition resulted in the fewest AIAN men and women being classified as having completed some college or more and the highest percentage of AIAN women being classified as having less than a high-school education. The any mention method yielded the lowest IHS coverage rates, and the any mention and census methods yielded the lowest rates of reliance on an IHS clinic as a usual source of care. The percentages of AIAN men and women born in the United States were highest when the California DOF was used. The different methods did not result in statistically significant differences in cancer screening or risk factor findings.

Although at times the differences between categories were small, in general, AIAN rates were most similar to rates observed among Whites when the any mention method was used and least similar when the census method was used. In terms of cancer screening, rates obtained with the census definition appeared to be lower than rates obtained with the other definitions, although differences did not reach levels of statistical significance.

The data available on the AIAN population are problematic for several reasons, including misclassification during data collection and, as the results presented here demonstrate, during tabulation of data. Accurate information is needed for planning and policymaking and for interventions designed to improve the health of this population. Misclassification because of a range of causes can result in underreporting of the AIAN population.(n15-24)

For example, many American Indians/ Alaska Natives, especially those residing in California and the Southwest, have Spanish surnames, often a historical remnant of ancestors who were slaves of missions, and they are frequently counted as Latino when surname is used as a proxy for race/ethnicity. Also, AIAN race may not be recorded on medical records (e.g., at hospitals or health clinics), possibly as a result of incorrect assignment by clerks or because there is no option for recording American Indian or Alaska Native on intake forms.

In addition, self-identifications may change when a tribe formerly "unrecognized" becomes federally recognized by Congress or if tribal enrollment ordinances change. Furthermore, errors are made on birth certificates or death certificates, both of which can be used to obtain population data.(n12, n25)

Racial misclassification is one of the most difficult obstacles to accurate and informative data collection and reporting on the AIAN population. The tabulation method used by the California DOF provides a good example of the reasons why large proportions of American Indians/Alaska Natives are undercounted in that state. The DOF employs the unusual procedure of classifying as Hispanic/ Latino anyone who reports Hispanic/Latino ethnicity or has a Hispanic surname. As a result, because Latinos were excluded, the percentage of American Indians/Alaska Natives born in the United States was highest and the percentage of Californians classified as American Indians/Alaska Natives was lowest when the DOF definition was used. The population estimate obtained with the California DOF method (261 non-Latino AIAN men and 351 non-Latino AIAN women) was by far the smallest of any of the methods used, yielding only about half as many American Indians/ Alaska Natives as the census definition.

Interestingly, the other key concern related to AIAN data is overcounting for tabulation purposes, and the any mention definition is illustrative of this problem. The any mention method may produce overcounts of American Indians/Alaska Natives because it includes many people who do not identify closely with AIAN culture and society. This is most sharply illustrated by IHS coverage rates, which were significantly lower among the any mention group than among the other groups. One explanation of why people mention AIAN background but do not identify themselves as only or mostly American Indian/ Alaska Native may be that, historically, tribes have required a minimum blood quantum for tribal enrollment and accompanying IHS health coverage.

Another possible explanation is that out-of-state tribally enrolled American Indians/ Alaska Natives are unlikely to be able to use California-based IHS services. Furthermore, although only a small percentage of California AIAN residents (fewer than 10%) are served by the IHS, the 4 different methods yielded different percentages in terms of use of the IHS as a usual source of care. For example, according to the any mention method, this rate was 2%, whereas the California DOF definition suggested a rate of nearly 10%.

The definitions have other implications for analyses of AIAN data. As mentioned, the overall population of AIAN men and women seemed to be younger when the census definition (which included Latinos) was used than when the other definitions were used; in addition, lower levels of education and higher levels of poverty were seen with this definition. Although the cancer screening rates yielded by the census definition appeared to be lower than those yielded by the other definitions, the differences were not significant. Generally, the any mention method appeared to move AIAN rates closer to the rates among Whites--that is, the rates usually used in study comparisons--and the census method moved these rates further away from those of Whites.

Because they were weighted to the 2000 California census, as described earlier, we have more confidence in the CHIS 2001 AIAN prevalence estimates obtained with the census and any mention methods. However, statistical weighting is a complicated procedure and depends heavily on accurate population totals obtained from a large survey or census. The US census serves as the benchmark for all federal surveys as well as for the CHIS. The timing differences between the 2000 census and CHIS 2001 were minimal. The 2000 census population totals reflected the US population on April 1, 2000, and the CHIS 2001 data collection period spanned November 2000 to October 2001. The modes of data collection were different, however. Data for the 2000 census were obtained via a mailed questionnaire; individuals who did not respond were queried by telephone or in person. All CHIS 2001 responses were obtained via telephone.

More single-race California AIAN residents were reported in CHIS 2001 than in the 2000 census. The California census estimated that the state's AIAN adult population represented 1.1% of the overall state population, whereas the CHIS estimate, adjusted to reflect geographical stratification, was 1.7%.(n26) Although this difference is small in absolute terms, it is large in percentage terms (54.5%). This large percentage difference requires explanation; however, the reasons for the discrepancy are not obvious, and it is unclear which estimate is more accurate. One possible explanation is that, in the census, one individual could respond for all members of the household (i.e., proxy responses were allowed). Additional research is needed to establish accurate numbers in this instance.

Categorizing American Indians/Alaska Natives with CHIS data is difficult, and there is no obvious "best" approach. However, our analysis suggests some guidelines, and we conclude by recommending what we think is a reasonable approach. We do not recommend using either the any mention or California DOF method for tabulating AIAN population data. The any mention method does not appear to be the best option given that it includes many individuals who do not strongly identify with AIAN race/ethnicity. The difficulty with the DOF method is that it excludes all American Indians/Alaska Natives with Spanish surnames and those who identify themselves as Hispanic/Latino.

This leaves us with the UCLA Center for Health Policy Research and US census definitions. The virtue of the UCLA approach is that it includes a question asking multiracial respondents to indicate their primary race. Despite this advantage, data analyzed via the UCLA definition cannot be compared with AIAN data from census, IHS, or tribal databases or with other national data. Therefore, we believe that--given its wide use and the fact that it can be readily incorporated into between-studies comparisons--the most straightforward approach is the census single-race definition.

The census definition is particularly useful in the case of studies in which the focus is on comparisons of racial/ethnic groups. When more screening and lifestyle data become available on American Indians in regions other than California, it would be interesting to compare results. We have found California to be unique in its heterogeneity of populations and expect that the data presented here do not generalize to all AIAN groups. However, variations in tabulating AIAN groups will occur across regions.

A report recently released by the US Commission on Civil Rights indicated that "persistent discrimination and neglect continue to deprive American Indians of a health system sufficient to provide health care equivalent to that provided to the vast majority of Americans."(n27) Accurate data collection is the first step in establishing baseline estimates of the health service needs of a given population. We have shown that methods of identification of American Indians/ Alaska Natives are not uniform across the data collection enterprise. Although there is no "best approach" to categorizing American Indians/Alaska Natives, it is clearly important to measure health disparities in this population. Because it is critical to develop consistent measurement criteria, we recommend using the census definition in comparisons of health indices across racial/ ethnic groups.

In 2003, the US Census Bureau conducted focus groups with representatives of tribal governments to plan for the 2010 census in tribal areas.(n28) Obviously, complex sociocultural and political concerns are associated with identifying oneself as--and being classified as--American Indian/Alaska Native, and these issues are beyond the scope of this article. However, we hope the present investigation provides guidance to researchers and policymakers in improving the data available on the AIAN population, an often forgotten and invisible minority group.

This article was accepted December 23, 2004.

J. Swan led the writing and supervised the study. N. Breen assisted with the writing and led the analysis. L. Burhansstipanov and D. E. Satter contributed to the discussion of misclassification and policy. W.W. Davis provided information on weighting and assisted with the analysis. T. McNeel performed computer analyses of the data. C.M. Snipp provided background information on population statistics related to American Indians and Alaska Natives.

The 2001 California Health Interview Survey was co-funded by the National Cancer Institute, the California Department of Health, the California Endowment, the Centers for Disease Control and Prevention, and the Indian Health Service.

The California Health Interview Survey is bound, by California law, to promises made to respondents and is constrained by university and government human participant protection committees to ensure that no personal information is released in a form that identifies an individual.

Percentage Distribution of California Women According to Sociodemographic Characteristics, Access to Care, Risk Factors, and Cancer Screening: California Health interview Survey (CHIS), 2001

Percentage Distribution of California American Indian/Alaska Native (AIAN) Men by Sociodemographic Characteristics, Access to Care, Risk Factors, and Cancer Screening: California Health interview Survey (CHIS), 2001

Percentage Distribution of California American Indian/Alaska Native (AIAN) Women According to Sociodemographic Characteristics, Access to Care, Risk Factors, and Cancer Screening: California Health Interview Survey (CHIS), 2001