ISSN 2691-7475 Health Policy Brief November 2020 Limited Access to Health Data on American Indian and Alaska Natives Impedes Population Health Insights Tara L. Becker, Susan H. Babey, Riti Shimkhada, AJ Scheitler, and Ninez A. Ponce ‘‘ The lack of reporting on AIAN populations and their health SUMMARY: American Indian and Alaska Native (AIAN) people are underrepresented and often invisible in public health data and research. AIAN health data capacity is impeded by the quality of information collected, released, and reported on AIANs in population-based surveys. AIANs are either put in a residual “other” category To promote more accurate insights on the health of the AIAN population, we examined eight population datasets that focus on health, health status, health behaviors, and health access. For each survey, we assessed whether it is possible to identify AIAN respondents from (1) readily accessible public-use data files and (2) restricted issues challenges or, typically, depicted as single-race non-Latinx AIANs. Though the figure varies substantially data files. We also assessed the extent to which data users can distinguish three major subgroups policymakers across federal surveys, fewer than two in five within the AIAN race category from these data charged with AIANs report as single-race non-Latinx AIANs. Drawing population inferences based on single- files: single-race non-Latinx AIAN, single-race Latinx AIAN, and AIAN of more than one race. addressing In this policy brief, we discuss the implications ’’ race AIANs also fails to capture the considerable disparities. segment of the AIAN population that comprises of the limited accessibility of AIAN information individuals of more than one race. in population datasets and provide some recommendations that may help improve the availability of AIAN health information. I mproving the health of an estimated 5.7 million American Indians and Alaska Natives (AIANs) in the nation1 is hampered to community distrust. Although protecting respondents is of primary importance for survey managers, one result of this limitation by population data systems that hide or is that conducting research that focuses on obscure their representation in federal data AIAN populations is often not possible systems. In most national, population- using publicly available data. The lack of based survey data, the number of AIAN reporting on these populations and their respondents interviewed is too small to allow health issues challenges policymakers and the inclusion of race/ethnicity information program developers charged with addressing that reports AIAN as a separate category in disparities. the publicly available data files. Doing so presents a disclosure risk, because AIAN Individuals who identify as AIAN are respondents could potentially be identified likely to also self-identify as another race by data users. The public and survey and are more likely to identify with Latinx respondents expect privacy, and survey ethnicity. (In this report, “Latinx” is used to managers ensure this commitment to guard denote those who are Latino/Latina as well against disclosure breaches, which could lead as those who are Hispanic or of Spanish- 2 UCLA CENTER FOR HEALTH POLICY RESEARCH ‘‘ In public-use data files, often only single- race non- Latinx AIAN speaking origin.) The most common race- classification systems in public health and most other population-based data combine those who report more than one race into a single category, which makes it impossible to identify AIANs who report being two or more races. In addition, most public health accessible, and (2) restricted-use data files. We also assessed the extent to which each data file allows users to distinguish three important subgroups within the AIAN race category: single-race non-Latinx AIAN, single-race Latinx AIAN, and AIAN of more than one race. Exhibit 1 provides information are classified datasets contain at least one measure that about the eight surveys evaluated. as AIAN, combines Latinx ethnicity with the five major Office of Management and Budget race substantially categories: non-Latinx American Indian or SURVEY ACRONYMS reducing the Alaska Native, non-Latinx Native Hawaiian/ BRFSS Behavioral Risk Factor Surveillance System sample size of Pacific Islander, non-Latinx Asian, non-Latinx ’’ black or non-Latinx African American, and CHIS California Health Interview Survey the population. non-Latinx white. This means that when MCBS Medicare Current Beneficiary Survey AIAN respondents can be identified in public-use data files, often only single-race NHANES National Health and non-Latinx AIAN are classified as AIAN, Nutrition Examination Survey substantially reducing the sample size of the NHIS National Health Interview Survey AIAN population. NSCH National Survey of Children’s Health This policy brief describes the accessibility NSDUH National Survey on Drug Use of information on AIAN in each of eight and Health (before 2002: National population health surveys (see box, this page). Household Survey of Drug Abuse) For each survey, we assessed whether it is PATH Population Assessment of Tobacco possible to identify AIAN respondents in and Health (1) public-use data files, which are readily Exhibit 1 Description of Eight Population Health Surveys Survey Time Period of Survey Survey Population Primary Source for... Longitudinal? Noninstitutionalized Began 1984 in 15 states; State-level population BRFSS civilian adults in the U.S. No nationwide since 1993 health information and territories Noninstitutionalized Population health Biennial 2001–2009; CHIS civilian adults, adolescents, information for the state of No continuously since 2011 and children in California California Medicare cost and MCBS Continuously since 1991 Medicare enrollees Yes utilization information Began 1960s; Noninstitutionalized NHANES Physical examination data No continuously since 1999 civilian residents of the U.S. Noninstitutionalized National population health NHIS Continuously since 1957 No* civilian residents of the U.S. information Noninstitutionalized National and state-level 2003, 2007, 2011–2012; NSCH children under age 18 population health No annually since 2016 living in the U.S. information on children Began 1979; annually since Noninstitutionalized Detailed substance use NSDUH 1990; all 50 states + D.C. civilian residents of the U.S. No information since 1999 ages 12 and over Noninstitutionalized Detailed tobacco use and PATH Began 2013 civilian residents of the U.S. Yes cessation information ages 12 and over *Subsample of respondents followed longitudinally through the Medical Expenditures Panel Survey UCLA CENTER FOR HEALTH POLICY RESEARCH 3 Accessibility of AIAN Subgroup Information in Public-Use Data Files by Survey Exhibit 2 Available in Public-Use Files (PUF)? Where to Access Single-race Single-race Restricted Data AIAN More Survey All AIAN Non-Latinx Latinx if Not in PUF? Than One Race AIAN AIAN CHIS Yes* Yes* Yes* Yes* CHIS Data Access Center BRFSS 2001–2012 from 2001 from 2001 2001–2012** FSRDC, NCHS RDC NHIS No from 1997 from 1997 No** FSRDC, NCHS RDC NSDUH † No Yes No No FSRDC, NCHS RDC NSCH No No‡ No‡ No FSRDC, NCHS RDC Limited dataset obtained MCBS No No No No from CMS with data use agreement NHANES No No No No FSRDC, NCHS RDC PATH No No No No ICPSR Virtual Data Enclave * Available for all age groups in two-year combined public-use Note: FSRDC = Federal Statistical Research Data Center files, not in single-year public-use files NCHS RDC = National Center for Health Statistics ** Respondents who reported more than one race were asked for Research Data Center a single main (NHIS) or preferred race (BRFSS), and those CMS = Centers for Medicare and Medicaid Services who selected AIAN as a main/preferred race can be identified ICPSR = Inter-university Consortium for Political and in the PUF data through 2019. Social Research † Public-use data files are available only from 2002. ‡ It is possible to identify single-race AIAN (both non-Latinx and Latinx) who live in the subset of states in which AIANs ‘‘ make up 5% or more of the child population, but not all single-race AIAN. We selected these surveys based on three use and cessation efforts. Each of these surveys Only two of the primary criteria: (1) how often the survey is used in published research studies, (2) the evaluates collected race information in a way that allows researchers to identify respondents eight surveys uniqueness of the information contained in of more than one race, and each collects Latinx allow users to the survey, and (3) the collection of detailed ancestry information separately from race. identify all racial/ethnic information that allows the identification of Latinx AIANs and AIANs Both of these steps are necessary to allow identification of the total AIAN population four AIAN who report more than one race. and the three AIAN subgroups. groups in public-use Using these criteria, BRFSS and NHIS were selected based on their extensive use by public health researchers. The remaining six AIAN Information in Population Health Surveys AIAN Information in Public-Use Data Files ’’ data files. surveys were selected based on the uniqueness Only two of the eight surveys—CHIS and of their content or their focus on a specific BRFSS—allow users to identify all four AIAN subpopulation. MCBS is the primary source groups in public-use data files in any time of information on Americans enrolled in period (Exhibit 2). The remaining datasets Medicare. The California Health Interview present limited or no identifiable information on Survey (CHIS) is the largest continuously AIANs in the public-use data files. collected state health survey, and California contains a sizable AIAN population. NSCH •In CHIS, it is possible to identify all three is the largest public health survey targeted AIAN race subgroups (single-race non- specifically to children. NHANES is unique Latinx AIAN, single-race Latinx AIAN, and for its collection of both survey and biological AIAN reporting more than one race) for all data. NSDUH contains detailed population ages by using the public-use data file that health data on drug use, and PATH contains combines two years of data. detailed population health data on tobacco 4 UCLA CENTER FOR HEALTH POLICY RESEARCH ‘‘ Restricted- use data files contain more detailed race information •Before 2013, BRFSS included in public- use files detailed information on all races reported by the respondent, as well as a separate indicator for Latinx ancestry. Beginning with the 2013 data, this detailed information is no longer available; single- race AIAN respondents can be identified. Access to Restricted-Use Data Files Each of the eight datasets has restricted-use data files that contain more detailed race information and allow researchers to identify each of the three AIAN subgroups (Exhibit 2). Five of these datasets now make their restricted data available through either a federal statistical and allow However, BRFSS includes a “preferred race” research data center (FSRDC) or a National researchers to measure, in which respondents who report more than one race choose which race they Center for Health Statistics research data center identify each of most identify with. Using this measure, (NCHS RDC). Three of these (NHANES, NHIS, and NSCH) have been available this the three AIAN it is possible to identify a subset of AIAN way for several years; access to BRFSS and subgroups.’’ respondents of more than one race who most identify as AIAN. •Beginning with the 1997 data, NHIS NSDUH was added more recently. The FSRDCs are located throughout the public-use data files have included a race- country. We accessed the FSRDC located only measure that identifies single-race at UCLA. NCHS RDCs open to nonfederal AIAN respondents, but combines those researchers are located in two locations in the who report AIAN and another race with greater Washington, D.C., metro area. Direct other respondents who report two or more access using the NCHS RDC is advertised races.1 When combined with the indicator as taking from 8 to 12 weeks. However, for Latinx ancestry, NHIS data identify both acquiring access to an FSRDC can take more single-race non-Latinx AIAN and single-race than three months, due to NCHS proposal Latinx AIAN. Though it is not possible to review and federal background checks needed identify all respondents who are more than for acquiring Special Sworn Status through one race, up until 2018, NHIS asked those the U.S. Census Bureau. Researchers can incur reporting more than one race to identify substantial fees and user costs (the costs for our which one best represented their race; those project were $4,500 for access to four NCHS who reported AIAN as their main race can datasets, and $15,000 for FSRDC access). thus be identified in the public-use data for The NCHS fees are based on the number of each year through 2018. datasets and years requested. The FSRDC fees support the operations of the local FSRDCs, •In the NSDUH public-use data files, only and the cost varies depending on the staffing single-race non-Latinx AIAN respondents can needs for the project. Once approved, a be identified, because all Latinx AIAN are researcher can conduct analyses in the secure classified as Latinx, while AIANs reporting site of the FSRDC, then submit the output to more than one race are classified with all other NCHS. After disclosure review by NCHS and respondents who report two or more races. the U.S. Census, NCHS emails the output to •In NSCH public-use data files, it is possible the researcher. to identify single-race AIAN (Latinx and non- Latinx) only within a subset of U.S. states in Three of the restricted datasets we used were which at least 5% of the population is AIAN. accessed through other means. Researchers who would like to use the CHIS restricted •In MCBS, NHANES, and PATH, AIANs data can do so only by submitting a research cannot be identified in the public-use data application to the CHIS DAC and paying files. UCLA CENTER FOR HEALTH POLICY RESEARCH 5 Weighted Percentage of AIANs by AIAN Subgroup and Survey Exhibit 3 Single-race Non-Latinx AIAN Single-race Latinx AIAN Multiracial AIAN NHANES 2011–2014 0.53% 0.06% 1.70% NHIS 2013 0.57% 0.32% 1.79% NSDUH 2014 0.54% 0.63% 1.21% PATH 2013–2014 0.33% 0.71% 1.54% BRFSS 2013 1.06% 0.83% 0.85% NSCH 2011–2012 0.93% 0.43% 1.95% CHIS 2011–2014 0.43% 1.36% 1.22% MCBS 2010–2015 0.96% 0.13% 2.13% 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% Note: Analyses used weights provided by administrators to account for the design of each survey. a setup fee of $500 and a user fee that on average ranges from $1,000 to $3,000. Although they are not granted direct access to the CHIS data, researchers can submit analysis requests or statistical analysis programs to a CHIS statistician, who will then produce the requested output. less impact on our understanding of health- related outcomes for AIAN populations. However, this is not the case. Using restricted- access files for all datasets except CHIS and BRFSS, Exhibit 3 shows the percentage of the population that falls into each of the AIAN subgroups within each survey. ‘‘ AIANs reporting more than one race make up the largest Researchers who want to use the MCBS Despite differences across surveys in the size proportion of limited dataset (restricted data) can request the data from CMS. Once CMS has approved of the overall AIAN population, we can see the overall the request, researchers enter into a data use a clear pattern. In six of the eight surveys, AIAN agreement (DUA) with CMS. In addition, effective this past June, researchers incur a fee of $600 per year of MCBS data requested. AIANs reporting more than one race make up the largest proportion of the overall AIAN population. Reflecting the high percentage ’’ population. Once fees are paid, the restricted MCBS of Latinx residents in California, single-race data are provided to the researcher. Access to Latinx AIAN respondents in CHIS comprise restricted-use data for PATH can be obtained the largest AIAN subgroup. In BRFSS, through the ICPSR Virtual Data Enclave. single-race non-Latinx respondents comprise the highest percentage of the overall AIAN Implications of Inaccessibility of AIAN population. Further, as seen in most population Information in Health Datasets surveys, the population of AIANs reporting Most often, when researchers can identify more than one race is substantially larger than any AIAN respondents in the public-use that of single-race non-Latinx AIANs. data, the only AIAN subgroup that can be identified is single-race non-Latinx. If most AIANs fell into this category, this would have 6 UCLA CENTER FOR HEALTH POLICY RESEARCH Exhibit 4 Weighted Percentage of AIANs Who Are Single-Race Non-Latinx AIAN by Survey NHANES 2011–2014 23.0% NHIS 2013 21.4% NSDUH 2014 22.9% PATH 2013–2014 12.9% BRFSS 2013 38.5% NSCH 2011–2012 28.1% CHIS 2011–2014 14.2% MCBS 2010–2015 29.8% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0% ‘‘ Note: Analyses used weights provided by administrators to account for the design of each survey. The practice The practice of focusing on single-race non- information about this population. When Latinx AIAN respondents significantly reduces these datasets are the primary source of of focusing on the size of the AIAN population. This can be information on a particular topic (such as single-race non- seen in Exhibit 4, which shows the percentage PATH’s wealth of information on tobacco Latinx AIAN of the overall AIAN population in each survey use) or are of a specific type (such as the representing single-race non-Latinx AIAN. NHANES biological data collection), respondents Though the number varies substantially across researchers’ ability to use the data to assess significantly surveys, fewer than two in five AIANs report these health measures for AIAN populations reduces the size as single-race non-Latinx AIAN. In Wave 1 of PATH, which was collected in 2013–2014, is nearly foreclosed. of the AIAN The convention of tabulation and racial/ ’’ only 12.9% of AIANS reported as single- population. race Latinx AIAN. In comparison, in 2013 ethnic reporting in publicly available data BRFSS, just under two-fifths (38.5%) of all is intended to (1) report single-race non- AIANs reported as single-race non-Latinx Latinx and (2) assign all respondents of more AIAN. Although single-race non-Latinx than one race to one multiracial category. As AIANs comprise less than 40% of the total shown in this brief, this convention obscures AIAN population nationally—and in some insights about the population that has surveys, less than 20%—this is the AIAN substantial subpopulations who are Latinx subgroup most often available to researchers. and are of more than one race. As a result, this is the AIAN subgroup most often used when making inferences about Restricted data generally provide the the health of the AIAN population. disaggregation needed for the AIAN population, but access to restricted data Conclusions and Recommendations is often difficult, expensive, and time- Few population health datasets include consuming. This discourages the use of measures that allow researchers to identify such data, thus significantly limiting access any AIAN populations in their public-use to information about the health-related data. Population-level public health research disparities faced by the AIAN population on AIANs remains scarce, in part because of and rendering their needs invisible in public the difficulties researchers face in accessing health policy. UCLA CENTER FOR HEALTH POLICY RESEARCH 7 There may be significant health-related one race into one category. This indicator differences across the subgroups we would allow continued access to data on examined; the presence of such differences AIAN populations when disclosure concerns would suggest that not accounting for the prevent the analysis of AIAN subgroups. experience of all three of these subgroups of The inclusion of at least the overall AIAN the AIAN population leads to an incomplete population category would allow researchers understanding of the health-related challenges interested in the AIAN population to use faced by AIANs. Unless there is better the dataset. measurement of the barriers to accessing data •Note when conclusions about AIAN on the AIAN population, and a lowering of health are based solely on single-race non- these barriers, these subgroups will continue to Latinx AIANs. When analyzing data in be underrepresented in public health research which only single-race non-Latinx AIANs and public health programs. are identifiable as AIAN, researchers should make it clear that results pertain only Survey leaders and researchers should consider to this population, not to AIANs more the following to improve measurement and generally. Much of what we know about the availability of information about the health of health of the AIAN population is based on the AIAN population: single-race non-Latinx AIANs. However, •Oversampling methods can be used to this subgroup comprises less than about increase the number of AIAN respondents 40% of the total AIAN population in to mitigate disclosure risk. This is BRFSS and only about 13% in PATH. expensive, but it is still an obvious and •Reduce cost/time for accessing restricted important strategy for improving federal data. If it is not possible for AIAN health data on the AIAN population. information to be included in public-use •Release pooled multiyear public-use files datasets for some surveys, the accessibility that include more detailed race/ethnicity of that survey’s restricted data becomes data. Pooling multiyear data reduces the extremely important for obtaining The California Health disclosure risk that may be associated with information about the AIAN population. Interview Survey (CHIS) providing more detailed race data. For covers a wide array of This is especially critical for surveys that health-related topics, example, CHIS provides two-year public- are the primary source of information on a including health insurance use data files that include sufficient racial/ health topic, such as NSDUH and PATH. coverage, health status ethnic detail to identify all of the AIAN and behaviors, and access subgroups. Author Information to health care. It is based The authors of this policy brief hold the following on interviews conducted •Include information on the overall AIAN positions at the UCLA Center for Health continuously throughout population. When sample sizes prohibit Policy Research: Tara L. Becker, senior public the year with respondents the inclusion of any of the AIAN subgroups, administration analyst; Susan H. Babey, PhD, senior from more than 20,000 datasets should include information on the research scientist; Riti Shimkhada, MPH, PhD, California households. overall AIAN population. It is preferable senior research scientist; AJ Scheitler, EdD, director to include the overall AIAN category along of stakeholder relations; Ninez A. Ponce, MPP, PhD, CHIS is a collaboration professor at the UCLA Fielding School of Public among the UCLA Center with the single-race AIAN category to Health and director of the UCLA Center for Health for Health Policy Research, expand tabulation options for the AIAN Policy Research. California Department of population. In cases where disclosure risk Public Health, California prevents the release of measures identifying Funder Information Department of Health Care any AIAN subgroup, such as single-race non- This policy brief was made possible through Services, and the Public funding from the Office of the Assistant Secretary of Health Institute. For more Latinx AIAN, survey administrators should Planning and Research (ASPE), U.S. Department of information about CHIS, consider publicly releasing at a minimum please visit chis.ucla.edu. Health and Human Services. an overall AIAN indicator that combines single-race AIAN and AIAN of more than UCLA CENTER FOR HEALTH POLICY RESEARCH 10960 Wilshire Blvd., Suite 1550 Los Angeles, California 90024 Acknowledgments For their thoughtful review and feedback, the authors wish to thank Delight Satter, MPH (Confederated Tribes of Grand Ronde), senior The UCLA Center for Health Policy Research health scientist, Office of Tribal Affairs and is part of the Strategic Alliances, Center for State, Tribal, UCLA Fielding School of Public Health. Local, and Territorial Support at the Centers for Disease Control & Prevention (CDC); Jennifer Parker, PhD, director, Division of Research and Methodology at the National Center for Health Statistics; and Randall Akee, PhD, associate professor, UCLA American Indian Studies and Luskin School of Public Affairs. The analyses, interpretations, conclusions, and views expressed in this policy brief are Suggested Citation those of the authors and do not necessarily Becker TL, Babey SH, Shimkhada R, Scheitler represent the UCLA Center for Health Policy Research, the Regents of the University AJ, Ponce NA. 2020. Limited Access to Health of California, or collaborating Data on American Indian and Alaska Natives organizations or funders. Impedes Population Health Insights. Los Angeles, PB2020-8 Calif.: UCLA Center for Health Policy Research. Copyright © 2020 by the Regents of the University of California. All Rights Reserved. Endnote Editor-in-Chief: Ninez A. Ponce, PhD 1 U.S. Census Bureau. American Community Survey, 2018 1-Year Estimates, Table B02010. Generated by Ninez Ponce using data.census.gov (2 August 2020). Phone: 310-794-0909 Fax: 310-794-2686 Email: chpr@ucla.edu healthpolicy.ucla.edu Read this publication online