| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RESEARCH AND PRACTICE |
Emily Puukka, Paul Stehr-Green, and Thomas M. Becker are with the Northwest Tribal Epidemiology Center, Northwest Portland Area Indian Health Board, Portland, Ore.
Correspondence: Requests for reprints should be sent to Emily Puukka, MS, Northwest Tribal Epidemiology Center, Northwest Portland Area Indian Health Board, 527 SW Hall, Suite 300, Portland, OR 97201 (e-mail: epuukka{at}npaihb.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Objectives. We compared the historical method of calculating cancer incidence rates with 2 new methods to determine which approach optimally estimates the burden of cancer among the Northwest American Indian/Alaska Native (AIAN) population.
Methods. The first method replicates the traditional way of calculating race-specific rates, and the 2 new methods use probabilistic record linkages to ascertain cancer cases. We indirectly adjusted all rates to the standard 2000 US population.
Results. Whereas the historical cancer incidence rates for all races are more than double those for the AIAN population, this apparent gap is considerably narrower when the all-race rates are compared with AIAN-specific rates calculated with probabilistic linkage methods. Similarly, there is no meaningful difference in incidence rates for selected site- and gender-specific cancers between the AIAN population and all races combined, and, in fact, some of these rates may be higher among the AIAN population.
Conclusions. Our results suggest that the burden of cancer among the AIAN population is considerably higher than was previously understood. We recommend that a standardized approach based on probabilistic linkage methods be adopted and that adequate financial and technical support be made available for conducting routine linkage studies throughout Indian communities.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
To design and target effective intervention programs and to justify increased resources to fund such activities, we must obtain health status data that are accurate and complete. This imperative demands resolution of misclassification of the AIAN population and also indicates the need for a standard method for calculating AIAN cancer estimates so that rates are comparable over time and from place to place. We conducted this analysis to compare historical methods of calculating AIAN cancer incidence with 2 new methods to determine which approach optimally describes the true burden of cancer among the Northwest AIAN population.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Cancer is a reportable disease in Idaho, Oregon, and Washington. All invasive, malignant neoplasms (except basal and squamous cell carcinoma of the skin) and all in situ carcinomas (except carcinoma in situ of the cervix uteri and squamous cell carcinoma in situ of the skin) are reported to the state cancer registries. Cases are reported by a variety of sources, including hospital cancer registries, ambulatory surgical centers, physicians’ offices, state vital statistics offices, pathology laboratories, and hospital medical record departments. The cancer registries of Idaho, Oregon, and Washington adhere to the standards set forth by the North American Association of Central Cancer Registries. Racial data reported to the state cancer registries are collected inconsistently, and there are no state laws requiring uniform collection of racial and ethnic data. Hospitals are covered by federal mandates that dictate the collection of these racial and ethnic data; however, hospital tumor data registry personnel are able to report to state registries only what is available in a medical chart (Recinda Sherman, certified tumor registrar, written communication, October 28, 2004).
The Tribal Registry is an enumeration of the AIAN population primarily from Idaho, Oregon, and Washington. Source data for the Tribal Registry come from the Portland area Indian Health Service (IHS) patient file, a compilation of patients’ demographic data from Indian-run health care facilities that use the Resource and Patient Management System and export patient data to the Portland area IHS office. These data represent individuals who received services from Northwest Indian-run health care facilities from the mid-1980s to the present. All individuals in the Tribal Registry are of proven AIAN ancestry and have accessed health services from an IHS or tribal health care facility during this time period.
For all record linkages, we used the software program Integrity (Version 11.0, Ascential Software, Westboro, Mass), which employs a probabilistic linkage algorithm to link the Tribal Registry to the state cancer data. The program is designed to link records in 2 different data files in which data on selected characteristics (e.g., name, date of birth, social security number) are contained in both. A valuable feature of the program is its ability not only to identify exact matches (i.e., records for which selected fields are identical) but also to calculate the probability of a correct match in situations in which minor differences exist between records in the 2 data files (e.g., transposed digits in a social security number).
Since January 2003, we have provided information to state cancer registry personnel identifying racially misclassified cases of cancer among the AIAN population, enabling them to correct racial classifications in their respective registries. Therefore, we could only use data obtained from linkages conducted before 2003 to calculate rates with the historical method (described later). For the purposes of our comparisons, we limited data to cancer cases occurring between 1996 and 1999. We excluded all in situ cases from our analyses (except for cases of in situ urinary bladder cancer) and indirectly adjusted all rates to the standard 2000 US population.
We calculated all-race rates by using all reported cases in the state cancer registries as numerators and total population counts from the US Census as denominators. We calculated AIAN-specific rates by each of the 3 methods described next (Figure 1
).
|
Linkage Method 1
In linkage method 1, we calculated rates with only cases in which AIAN race could be verified through the Tribal Registry. To ascertain the number of cancer cases among known members of the AIAN population, we conducted a record linkage with the Tribal Registry and the 3 state cancer registries to identify records common to both data sets. Unlike the historical method, linkage method 1 included cases for which a record existed in both the state cancer registry and the Tribal Registry (i.e., regardless of whether they were correctly coded as AIAN in the state registries) as the numerator; however, we did not include in these counts or in the rate calculations the cases in which race/ethnicity was coded as AIAN in the state cancer registries when there was no matching record in the Tribal Registry. To ensure that all individuals who appeared in the numerator were also included in the denominator, we used the Tribal Registry as the source of denominators for calculating cancer incidence rates by linkage method 1.
Linkage Method 2
In linkage method 2, we calculated rates with all cases of cancer among the AIAN population, with or without verification by the Tribal Registry. We derived numerators by summing cases in which race/ethnicity was coded as AIAN in the state cancer registries and for which there was a matching record in the Tribal Registry (as in linkage method 1) plus cases in which race/ethnicity was coded as AIAN in the state cancer registries but there was no matching record in the Tribal Registry.
In calculating rates with this approach, we needed to use a denominator that included both individuals in the Tribal Registry (comprising IHS or tribal clinic users) and individuals reported to the state cancer registries whose race/ethnicity was reported as AIAN (but for whom we could not validate AIAN race with the Tribal Registry). Before 2000, the US Census allowed the reporting of only 1 of 4 races, that is, AIAN, Asian or Pacific Islander, Black, or White13; however, beginning in 2000, respondents were allowed to self-report 1 or more races: AIAN, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White.14 Because prior data on race were collected with different categories, the National Center for Health Statistics developed a "bridging" method based on data from the 1997–2000 National Health Interview Survey that allocates individuals who self-selected more than 1 race on the 2000 census to the 4 single-race groups used before 2000; the National Health Interview Survey is currently the best source of data on multiple and primary race selection, because individuals who self-identify with more than 1 racial group also select a primary race.15 Regression models were fitted to 4 years of the National Health Interview Survey data, and allocation probabilities were derived and applied to the 2000 census data on the assumption that responses to the primary race question on the National Health Interview Survey corresponded to how respondents would have answered the racial questions with the pre-2000 single-race categories.16 We reasoned that this denominator met the criterion given earlier in this subsection and captured persons who identified as AIAN, whether exclusively or in combination with 1 or more other race/races (i.e., the same persons who would qualify to be included in the numerator). Bridged-race population estimates are available from the National Center for Health Statistics Web site.17
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
The average annual age-adjusted rate for cancer of all sites among the Northwest AIAN population was 201.1 per 100000 (95% confidence interval [CI] = 172.1, 230.0) when the historical method was used, compared with 332.9 (95% CI = 288.4, 377.4) and 409.0 (95% CI = 362.1, 455.9) with linkage methods 1 and 2, respectively (Table 1). In addition, the magnitude and direction of change between methods was similar for gender-specific all-site cancer rates. The gender-specific and overall (both genders) age-adjusted cancer incidence rates for the general population appeared to be more than double those calculated by the historical method for the AIAN population; this apparent gap was considerably narrower when we compared the all-race rates with AIAN-specific rates calculated with linkage methods 1 and 2.
|
and 3). Although the overall burden of cancer among the AIAN population appeared to be slightly lower than the overall burden of cancer in all races combined, there was no difference in incidence rates for selected site- and gender-specific cancers between the AIAN population and all races combined, and some of these rates may actually have been higher among the AIAN population.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Given this strong body of evidence, we conclude that cancer incidence rates estimated with the historical method have underestimated the true incidence of cancer. Furthermore, evidence also suggests that in official US Census enumerations before the 2000 census, the AIAN population was probably undercounted, at least within certain sociodemographic subgroups (e.g., AIAN persons living on reservations).20 Although the overall impact of these denominator under-counts may have been to raise historical incidence rate estimates closer to the "true" rates (at least within the sociodemographic subgroups most likely to have been under-counted), the uncertain match between numerators and denominators (i.e., the cases represented in the numerator may not always come from the population at risk represented in the denominator) diminishes our confidence in the interpretation of these historical rate estimates.
Fortunately, there is a growing body of literature that demonstrates the potential of the record linkage approach for improving our understanding of the burden of disease, including cancer, among the AIAN population and their concomitant risk and etiological factors.4–12 Such linkages improve ascertainment of AIAN cancer cases and have the potential to allow managers of public health data sets to reduce racial misclassification. In this regard, linkage method 1 is an improvement compared with the historical method used to calculate AIAN cancer incidence rates. Not only does linkage method 1 ensure the validity of the numerator by counting only cases that occur among known members of the AIAN population, but it also uses a denominator that is known to comprise only AIAN persons, thereby representing a better match between numerator and denominator (i.e., only those AIAN persons in the IHS-defined "user population" are counted in both the numerator and the denominator). Nonetheless, despite the improved validity of linkage method 1, the rates calculated with this method are still incomplete and may not even be representative.
Because the extant evidence suggests that most misclassification errors are one-way (i.e., AIAN persons are most often misclassified as non-AIAN, not vice versa), linkage method 1 almost certainly yields an incomplete count of cases in the numerators (e.g., it misses AIAN persons who do not receive health care through Indian-run health care facilities). Furthermore, there are known problems with the completeness and representativeness of the Resource and Patient Management System data (the source of the population-at-risk denominator). Our analyses of the Portland area patient files suggest that only about 75% of the Northwest AIAN population identified by the 2000 US Census receive regular health care services through Indian-run health care facilities, and that this ratio varies by age group (AIAN children born into successive birth cohorts after 1990 were increasingly less likely to be listed in the IHS-defined active user population) and geographic location (ratios of active users of Indian-run health care facilities to AIAN census enumerations varied from state to state and were lower in urban areas) (P. Stehr-Green and E. Puukka, unpublished data, Northwest Portland Area Indian Health Board, 2004).
Linkage method 2 may be an improvement over the historical method and linkage method 1 approaches. Because the numerator counts used in linkage method 2 include both "official" AIAN cancer cases (those whose race/ethnicity is recorded as AIAN in public health data sets) and previously misclassified AIAN cases identified with linkage studies, this approach provides the most complete count of cases in the numerator. Furthermore, at least theoretically, the denominator used in linkage method 2 represents the best available estimate of the true population at risk (people who consistently identify exclusively as AIAN or as AIAN in combination with 1 or more other races). However, even linkage method 2 is not without its flaws and uncertainties. Perhaps of most concern is the fact that the National Center for Health Statistics–developed bridging method is new and has been incompletely evaluated with regard to its representativeness and consistency over time in measuring the number and distribution of the AIAN population. And, as with the historical method, the numerators and denominators derive from slightly different populations; until the magnitude and details (e.g., sociodemographic subgroups most affected by such discrepancies) are better understood, incidence rates calculated with linkage method 2 should be interpreted cautiously. Although we believe linkage method 2 to be the most comprehensive method currently available for determining the cancer burden among the AIAN population, it may still result in underestimation, given that AIAN persons not recorded in the Tribal Registry and incorrectly coded as non-AIAN in the state cancer registries are not captured by it. In addition, specific subpopulations of the AIAN population may be more likely than others to be racially misclassified (e.g., urban AIAN persons), leading to bias.
For the future, we need to develop a standardized method for calculating incidence rates among the AIAN population; given our results, we favor the use of rates calculated with linkage method 2. Furthermore, we need to insure the availability of adequate financial and technical support for conducting routine linkage studies throughout Indian communities. The immediate benefit of these actions will be an enhanced ability to identify the magnitude and distribution of racial misclassification and to adjust statistics, past and present, for these misclassification errors. Eventually, the results of these linkages will allow individual occurrences of racial misclassification in official public health data sets to be corrected. Likewise, the accuracy and representativeness of the bridged AIAN population estimates must be evaluated, especially for the entire AIAN population and for key demographic subgroups, so that these estimates can be used as standardized denominators in rate calculations performed by IHS and other federal agencies, tribes and tribal organizations, state vital statistics and other official data set registries, and other interested parties.
Finally, previous authors have identified the nature and source of problems in official public health data collection systems that lead to racial misclassification.21,22 Misclassification of race occurs in health-related databases most often because the recorded information is based on observation by physicians, coroners or medical examiners, or other health care workers, rather than on self-reports or reports by close relatives. Public health professionals must devise corrective measures and effectively advocate for financial and organizational support to correct these problems.
Sound public health practice requires that we use estimates of the burden of disease among AIAN communities to inform public decisionmaking. The validity and accuracy of such health assessments are adversely affected when health or demographic data are not reported accurately or completely in vital records and disease registries (e.g., cancer registries). Ultimately, through the efforts we have suggested, we should derive an improved understanding of AIAN health status that will provide a strong foundation for programmatic activities to improve health and well-being in AIAN communities.
|
Acknowledgments |
|---|
The authors thank the following colleagues for assistance with this project: Joe Campo, MPH; Brenda Edwards, PhD; Joe Finkbonner, RPh, MHA; Katie Golub, CTR; Chris Johnson, MPH; Josh Jones, MD; Lori Lambert, MA; Dee Robertson, MD, MPH; Recinda Sherman, CTR; and Judith Swan, MPH.
Human Participant Protection
Approval for this study was granted by the Portland Area Indian Health Service institutional review board.
|
Footnotes |
|---|
Contributors
E. Puukka conducted the data linkages, completed the analyses, and wrote the article. P. Stehr-Green originated the study, assisted with the analyses, and wrote the article. T. M. Becker assisted with interpretation of findings. All authors helped to conceptualize ideas and to review drafts of the article.
Accepted for publication January 3, 2005.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
2. Ries LAG, Eisner MP, Kosary CL, et al., eds. SEER Cancer Statistics Review, 1975–2001. Bethesda, Md: National Cancer Institute; 2004. Available at: http://seer.cancer.gov/csr/1975_2001. Accessed January 23, 2005.
3. Miller BA, Kolonel LN, Bernstein L, et al., eds. Racial/Ethnic Patterns of Cancer in the United States 1988–1992. Bethesda, Md: National Cancer Institute; 1996. NIH publication 96–4104.
4. Frost F, Tollestrup K, Ross A, Sabotta E, Kimball E. Correctness of racial coding of American Indians and Alaska Natives on the Washington State death certificate. Am J Prev Med. 1994;10:290–294.
5. Sugarman JR, Hill G, Forquera R, Frost FJ. Coding of race on death certificates of patients of an urban Indian health clinic, Washington, 1973–1988. IHS Primary Care Provider. 1992; 17:113–115.
6. Frost F, Taylor V, Fries E. Racial misclassification of Native Americans in a surveillance, epidemiology, and end results registry. J Natl Cancer Inst. 1992;84:957–962.
7. Sugarman JR, Holliday M, Ross A, Castorina J, Hui Y. Improving American Indian cancer data in the Washington State cancer registry using linkages with the Indian Health Service and tribal records. Cancer. 1996;78(suppl 7):1564–1568.
8. Sugarman JR, Lawson L. The effect of racial misclassification on estimates of end-stage renal disease among American Indians and Alaska Natives in the Pacific Northwest, 1988 through 1990. Am J Kidney Dis. 1993;21:383–386.
9. Sugarman JR, Soderberg R, Gordon JE, Rivara FP. Racial misclassification of American Indians: its effect on injury rates in Oregon, 1989 through 1990. Am J Public Health. 1993;83:681–684.
10. Becker TM, Bettles J, Lapidus J, et al. Improving cancer incidence estimates for American Indians and Alaska Natives in the Pacific Northwest. Am J Public Health. 2002;92:1469–1471.
11. Stehr-Green P, Bettles J, Robertson LD. Effect of racial/ethnic misclassification of American Indians and Alaskan Natives on Washington State death certificates, 1989–1997. Am J Public Health. 2002;92:443–444.
12. Epstein M, Moreno R, Bacchetti P. The underreporting of deaths of American Indian children in California, 1979 through 1993. Am J Public Health. 1997; 87:1363–1366.
13. Race and Ethnic Standards for Federal Statistics and Administrative Reporting. Washington, DC: Office of Management and Budget; 1977. Statistical Policy Directive 15.
14. Office of Management and Budget. Revisions to the standards for the classification of federal data on race and ethnicity. Federal Register. October 30, 1997; 62:58781–58790.
15. Ingram DD, Parker JD, Schenker N, et al. United States Census 2000 population with bridged race categories. Vital Health Stat 2. 2003;No. 135:1–55.
16. Schenker N, Parker JD. From single-race reporting to multiple-race reporting: using imputation methods to bridge the transition. Stat Med. 2003;22:1571–1587.
17. National Center for Health Statistics. US Census populations with bridged race categories. Available at: http://www.cdc.gov/nchs/about/major/dvs/popbridge/popbridge.htm. Accessed March 19, 2004.
18. Poe GS, Powell-Griner E, McLaughlin JK, Placek PJ, Thompson GB, Robinson K. Comparability of the death certificate and the 1986 National Mortality Followback Survey. Vital Health Stat 2. 1993;No. 118:1–53. Available at: http://www.cdc.gov/nchs/data/series/sr_02/sr02_118.pdf. Accessed January 29, 2005.
19. Sorlie PD, Rogot E, Johnson NJ. Validity of demographic characteristics on the death certificate. Epidemiology. 1992;3:181–184.
20. Tribal Governments Liaison Program Handbook. Washington, DC: US Census Bureau; 1999.
21. Hahn RA, Wetterhall SF, Gay GA, et al. The recording of demographic information on death certificates: a national survey of funeral directors. Public Health Rep. 2002;117:37–43.
22. Burhansstipanov L. Office of Management and Budget racial categories and implications for American Indians and Alaskan Natives. Am J Public Health. 2000; 90:1720–1723.
This article has been cited by other articles:
|
|
 |
|
  T. Smyer and T. E. Stenvig Health Care for American Indian Elders: An Overview of Cultural Influences and Policy Issues Home Health Care Management Practice, December 1, 2007; 20(1): 27 - 33. [Abstract] [PDF]
|
|
|
|
 |
|
 M. Lillie-Blanton and Y. Roubideaux Understanding and Addressing the Health Care Needs of American Indians and Alaska Natives Am J Public Health, May 1, 2005; 95(5): 759 - 761. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
