Disparities in Life Expectancy of Pacific Northwest American Indians and Alaska Natives: Analysis of Linkage-Corrected Life Tables

Abstract

Objectives

American Indians and Alaska Natives (AI/ANs) experience a high burden of mortality and other disparities compared with the general population. Life tables are an important population health indicator; however, federal agencies have not produced life tables for AI/ANs, largely due to racial misclassification on death certificates. Our objective was to correct this misclassification and create life tables for AI/ANs who resided in the Pacific Northwest region of the U.S., making comparisons with the general population.

Methods

To correct racial misclassification, we conducted probabilistic record linkages between death certificates from three Northwest states—Idaho, Oregon, and Washington State—issued during 2008–2010, and AI/AN patient registration records. We calculated mortality rates and generated period life tables for AI/ANs and non-Hispanic white (NHW) Americans.

Results

Overall life expectancy at birth for Northwest AI/ANs was 72.8 years, which was 6.9 years lower than that of NHW Americans. Male AI/ANs had a lower life expectancy (70.9 years) than female AI/ANs (74.6 years). The disparity in life expectancy between AI/ANs and their NHW counterparts was higher for females (with AI/ANs living 7.3 years fewer than NHW females) than for males (with AI/ANs living 6.7 years fewer than NHW males). The greatest disparity in mortality rates was seen among young adults.

Conclusion

Data linkage with a registry of known AI/ANs allowed us to generate accurate life tables that had not previously been available for this population and revealed disparities in both life expectancy at birth and survival across the life span. These results represent an important tool to help AI/AN communities as they monitor their health and promote efforts to eliminate health disparities.

As of the 2010 Census, 5.2 million people reporting American Indian or Alaska Native (AI/AN) ancestry were living in the United States.¹ AI/ANs represent a heterogeneous group with diverse cultures and histories, coming from more than 560 federally recognized and 100 state-recognized Tribes, as well as numerous additional Tribes currently not recognized by U.S. governmental entities. The U.S. AI/AN population is younger and experiences lower educational and economic attainment than the majority population.² As a group, AI/ANs have a profoundly different experience of morbidity and mortality throughout their lives than the general population: rates of diabetes, injuries, and cancer mortality are substantially higher.^2–8

The Pacific Northwest states of Idaho, Oregon, and Washington State are home to approximately 7% of the nation's AI/AN population.⁹ Northwest AI/ANs also experience significant health disparities; for example, mortality rates for this population exceed rates for non-Hispanic white (NHW) Americans for each of the 10 leading causes of death, with the exception of Alzheimer's disease. This population is twice as likely to die from unintentional injuries, diabetes, chronic liver disease, and homicide compared with their NHW counterparts in the Northwest.¹⁰ These statistics are alarming, and yet they fail to adequately convey the profound effect such health inequities have on tribal communities, in which each untimely death impacts the entire Tribe.

Accurate and complete health data are crucial to addressing this health status gap. Health indicators collected by federal, state, and local governments are the basis of the surveillance and evaluation that inform health policy decisions; however, these data systems are seldom able to generate accurate health measures for the AI/AN population. Vital statistics and surveillance systems are known to contain racial misclassification, in which AI/ANs are coded as another race, resulting in underestimated morbidity and mortality measures.^11–16 The need to overcome these issues and provide high-quality health data to AI/AN communities is especially important in light of the disparities that have been documented, even using data with these limitations.

Racial misclassification in public health data sources has been well examined, and for AI/ANs, misclassification has been estimated to be as high as 60%.^11,17,18 Misclassification of AI/ANs on death certificates has been demonstrated through multiple methods.¹⁹ A recent national study comparing death certificate race coding with self-identified race on the Current Population Survey found only 55% agreement for AI/ANs.¹² Other studies have established that AI/ANs are more likely to be misclassified than other races,^11,14 yet misclassification of other races as AI/AN is not common. Thus, the result of this racial misclassification is nearly always an underestimation of the disease burden for AI/ANs.

Record linkage is a relatively inexpensive and efficient way to address the problem of racial misclassification in estimating morbidity and mortality for AI/ANs.^{13,14,17,18,20–22} Through linkage, race coding on death certificates can be assessed and, where necessary, corrected. The Northwest Portland Area Indian Health Board (NPAIHB), a health organization operated by the 43 federally recognized Tribes of Idaho, Oregon, and Washington State, established the Northwest Tribal Registry (NTR) Project in 1999 to maintain a registry of AI/AN individuals. Researchers at NPAIHB have used this registry to improve the accuracy of health data available to the Tribes through linkage with public health data systems.

Mortality rate estimates provide the basis for many health surveillance efforts, including the construction of life tables. Researchers engaged in public health surveillance efforts use life tables to measure the overall health of a population, providing a readily understandable estimate of life expectancy at birth. Life table comparisons may reveal disparities that can better direct public health efforts, and form the basis for other health assessment tools, such as survival analysis, quality adjusted life years, and years of potential life lost.

Until recently, life tables were not published for the AI/AN population, due in large part to the racial misclassification issues previously mentioned.^12,23 Arias et al. (2014) produced the first period life tables for U.S. AI/ANs, with death counts based on a linkage between Indian Health Service (IHS) patient records and the National Death Index to correct racial misclassification.⁵ To improve reliability of the death data, these analyses were restricted to the non-Hispanic AI/AN population residing in Contract Health Service Delivery Area (CHSDA) counties. CHSDA counties are generally rural and, as a result, the majority of the urban AI/AN population, which accounts for 73% of AI/ANs in the Northwest,²⁴ is not fully represented in these life tables. Additionally, the heterogeneity of AI/AN health outcomes by geographic region has been well documented;⁶ thus, national-level life tables may not reflect the AI/AN population in the Northwest.

To address these limitations, we constructed abridged life tables for the AI/AN population in Idaho, Oregon, and Washington State after conducting linkages to identify and correct inaccurate race coding for AI/AN decedents, using a registry that included both urban and rural AI/ANs. To our knowledge, this is the first time linkage-corrected, regional life tables have been produced for this population.

METHODS

Data sources

The NTR is a list of AI/ANs, alive or deceased, who have received health services from federally operated (IHS), tribal, or urban Indian health-care facilities (collectively known as I/T/U) in Idaho, Oregon, or Washington State. Data in the NTR are from two primary sources: patient registration information from the Portland Area IHS administrative office (1986–2010) and the Seattle Indian Health Board (SIHB, 2007–2010). The SIHB, the largest urban Indian health organization in the Northwest, is a nonprofit, multiservice community health center that serves urban AI/ANs in the Puget Sound region of Washington State. Eligibility for services from IHS is generally based on enrollment in a federally recognized Tribe; thus, AI/AN race is verified at registration. Race information for patients of the SIHB is collected by self-report at the point of care. Prior to linkage, we excluded patient records that indicated a race other than AI/AN, leaving 194,413 AI/AN records for linkage. These files represented the most comprehensive and up-to-date registration of the AI/AN population in the Northwest, representing an estimated 76% of the three-state AI/AN census population.²⁵

We obtained death certificate files from 2008 through 2010 from the Idaho Department of Health and Welfare's Bureau of Vital Records and Health Statistics, the Center for Health Statistics at the Washington Department of Health, and the Oregon Health Authority's Office of Disease Prevention and Epidemiology. Race information on death certificates is usually provided by coroners or funeral directors who may not consult the decedents' medical records or next of kin to ascertain race. Using a probabilistic linkage process, we compared the death certificate files with the NTR. We used the linkage software Registry Plus^® Link Plus,²⁶ and matched on first and last name, date of birth, social security number, and date of death. Two individuals on our research team reviewed possible matches and assigned match status to uncertain pairs by consensus. We also reported to state vital statistics staff any death certificate cases that matched records in the NTR file, to enhance the quality of race data in the state systems. We then obtained de-identified death certificate files, with a flag indicating match status, to use for analysis.

Analysis

We constructed sex-specific life tables for the AI/AN and NHW populations in each state. We computed mortality rates for the life tables using linked, race-corrected death certificate files from 2008–2010 in the numerator and bridged-race population estimates from the National Center for Health Statistics for 2008–2010 in the denominator.^27–29 We summed three years of death counts and population numbers to lessen the impact of annual variability. Following the computation of mortality rates, we created complete and abridged life tables using the methodology described in the most recent National Vital Statistic Report.³⁰ Our life table construction differed somewhat from the methodology used to create the national life tables. First, we were unable to correct age misreporting on death certificates by linkage with Medicare data. To mitigate inaccuracies in mortality rates for older age, we capped the life tables at age 85 years. Secondly, we chose not to use any graduation techniques to smooth mortality rates, as these models have not been assessed for the AI/AN population specifically and could potentially inflate life expectancy estimates.^31–34 To reduce stochastic fluctuations in the data due to the small population, we aggregated three years of data and present abridged tables. We calculated 95% confidence intervals (CIs) for life expectancy estimates using the method described by Silcocks et al.³⁵

RESULTS

Linkage

All calculations were made based on the life tables we generated (Table 1 and Online Table, which is available online only). Linkage with the NTR identified 296 matched records in the Idaho death certificate files, 699 matched records in the Oregon files, and 1,653 matched records in the Washington State files (Figure 1). Of the total 2,648 matched records, 423 were coded as non-AI/AN or missing all race information in the death certificate files, for an overall misclassification prevalence of 16%. The linkage with NTR increased ascertainment of AI/AN deaths by 10%. After adjusting for race coding based on the linkage, the death certificate files contained 4,595 AI/AN records and 247,361 NHW records (data not shown). For the life table numerators, we classified as AI/AN any records that were coded AI/AN on the original death certificate or had a match in the NTR. The addition of the matched records decreased AI/AN estimates of life expectancy at birth by 0.98 years in Idaho, 2.15 years in Oregon, and 1.19 years in Washington State (data not shown).

Table 1.

Abridged life tables for the AI/AN population of Idaho, Oregon, and Washington State, 2008–2010

^aAge-specific mortality rates are based on small numerators in some cases and are not recommended for use in analyses without standard errors. Contact the authors for further information about the mortality rates.

AI/AN = American Indian/Alaska Native

CI = confidence interval

Figure 1.

Data linkage and derivation of AI/AN deaths in Idaho, Oregon, and Washington State, 2008–2010

AI/AN = American Indian/Alaska Native

Life expectancy

Table 2 summarizes life expectancy at birth for AI/ANs and NHW individuals by sex and state. For AI/ANs, life expectancy across the three states was 72.8 years. There were marked regional differences, with a range of more than three years among the three states. Oregon AI/ANs had the longest life expectancy (74.8 years), followed closely by Idaho AI/ANs (74.5 years), while Washington State AI/ANs had the shortest life expectancy (71.4 years). Across the Northwest, female AI/AN life expectancy was 3.7 years longer than for male AI/ANs. The gender gap was smallest among the Idaho population (1.9 years) and largest in Washington State (3.7 years). Male AI/ANs in Washington State were also the only group to experience a life expectancy at birth of <70 years.

Table 2.

Life expectancy at birth by race and sex for AI/AN and NHW populations: Idaho, Oregon, and Washington State, 2008–2010

AI/AN = American Indian/Alaska Native

NHW = non-Hispanic white

CI = confidence interval

Compared with their NHW counterparts, life expectancy at birth was 6.9 years shorter for AI/ANs. This pattern held for all three states and both genders. The gap between races was greater for females than males: female AI/AN life expectancy was 7.3 years shorter than for NHW females, while male AI/AN life expectancy was 6.9 years shorter than for NHW males. Stratified by state, the largest difference was observed among Washington State females (8.8 years less for AI/ANs) and the smallest difference was between Idaho males (3.9 years less for AI/ANs).

Disparities across the life span

Figure 2 shows mortality rate ratios (RRs) comparing AI/ANs with NHW people across the life span. The AI/AN to NHW mortality RR was 1.74 (95% CI 1.43, 2.12) in the first year of life, and increased to a -maximum of 2.45 (95% CI 1.58, 3.78) at age 23 years. The RR approached 1.00 slowly after the mid-twenties, but never dropped below 1.54 (95% CI 1.21, 1.97). The peak of disparity was slightly earlier for male AI/ANs, reaching 2.36 (95% CI 1.03, 4.45) at age 21 years. Females experienced more disparity overall, with the greatest disparity occurring at age 27 years with a maximum RR of 3.15 (95% CI 1.16, 3.92).

Figure 2.

AI/AN to NHW mortality rate ratios^a by age, in Idaho, Oregon, and Washington State, 2008–2010

^aMortality rates were smoothed using an eight-parameter Heligman and Pollard model to aid in visual interpretation of the graph. Life expectancy estimates used for all analyses were calculated without smoothing.

AI/AN = American Indian/Alaska Native

NHW = non-Hispanic white

DISCUSSION

This study, the first of its kind for the Northwest AI/AN population, identified high rates of misclassification on death certificates, as well as disparities in life expectancy and mortality rates compared with the majority population. Left uncorrected, racial misclassification caused life expectancy estimates for AI/ANs to be inflated by more than two years. Given the size of the AI/AN population, even a small number of miscoded cases can result in large differences in rates, making correct racial coding especially important for this group. Federal and state efforts are underway to address the issue of incorrect race data coding at the point of collection.³⁶ However, until these efforts have gained traction, data linkage remains a viable means of moving toward accurate AI/AN health statistics.

The life tables presented in this article provide both a starting point from which to monitor disparities in mortality of the Northwest AI/AN population and a snapshot of the current situation. Overall, we found that AI/ANs in the Northwest have a life expectancy at birth of 72.8 years, about seven years fewer than their NHW counterparts. Some subpopulations had even larger gaps, with female AI/ANs in Washington State experiencing the largest disparity at 8.8 years fewer than their NHW counterparts.

Mortality patterns throughout the life span revealed that, during infancy and childhood, AI/AN mortality rates were about double those of NHW children. Mortality RRs indicate that the greatest disparity arises among young adults. In early adulthood, AI/AN disparities are primarily due to higher mortality rates from injuries. For NHW people, unintentional injuries drop from the leading causes of death by age 45 years as chronic diseases take the forefront. For AI/ANs, this transition occurs later in life.³⁷ AI/ANs in the Northwest also differ from NHW people in suicide mortality; while the highest rates of suicide occur in the oldest age groups for the majority population, suicides among AI/ANs peak in young adulthood.^10,38 RRs comparing AI/AN and NHW mortality rates due to motor vehicle crashes and accidental drug and alcohol overdoses are particularly high among those aged 15–64 years.¹⁰

Correction of race misclassification on death certificates has been undertaken in other indigenous populations, and after addressing miscoding, researchers have consistently found that mortality among indigenous populations is higher than the majority population. First Nations and Metis data from Canada show a life expectancy gap of more than five years between the indigenous and nonindigenous populations;³⁹ a similar study in Hawai'i showed a gap of nearly five years between Native Hawaiians and Caucasians in that state.⁴⁰ Statistics from New Zealand find a gap of more than seven years between Maori and non-Maori populations,⁴¹ and data from Australia estimate aboriginal life expectancy at about 17 years fewer than the total Australian population.⁴² Although these indigenous populations vary widely in culture and health-care systems, these studies demonstrate a consistent disparity in life expectancy compared with majority populations.

In the U.S., many potential underlying factors contribute to these higher rates of mortality. Individuals who are unable to access I/T/U health facilities and are uninsured face substantial barriers to accessing quality health care. For those AI/ANs who can feasibly access an I/T/U health facility, underfunding and high provider turnover remain challenges. For tribal and IHS facilities, access to specialty care must be obtained by outside providers through a contract and is limited by budget and geographic isolation.⁴³

Despite these powerful factors pulling AI/AN health status downward, AI/AN communities in the Northwest have shown remarkable resilience. Alongside an increased focus on tribal self-determination, there have been encouraging improvements in AI/AN health during the past several years, including the development of successful injury prevention and cancer navigator interventions.^10,44–47 Organizations such as NPAIHB and SIHB allow Tribes and urban AI/AN communities to engage in collaborative public health efforts, and attract more funding and resources than would be possible individually.

Limitations

Our study had several important limitations. While the NTR represents most of the AI/ANs in the Northwest, it does not capture the entire population. A recent evaluation comparing the NTR population with census data estimated that the registry represents approximately 76% of the AI/AN population in the Northwest.²⁵ Some subpopulations (especially AI/ANs living in urban areas) are more poorly represented in the registry. There is some evidence that racial misclassification is more likely among urban AI/ANs;^12,14,15,18 thus, this subpopulation is both more likely to have miscoded race and less likely to have been corrected by our linkage. Without more comprehensive correction for misclassification in urban AI/ANs, it is uncertain to what degree our results accurately describe the true mortality experience of this population.

Most AI/AN studies are limited by the small population size relative to the general population, making estimates unstable and increasing the margin of error. To address this issue, we combined three years of data and limited geographic comparisons to the state level. Our results were comparable with those published by IHS⁴⁸ and more recent national results;⁵ therefore, we believe the results to be stable. We also recognize that age misreporting on death certificates is a known issue, and national life tables are adjusted to correct for misinformation through linkage with Medicare data. Historically, Medicare records did not identify AI/AN race, so these linkages are not possible for this population. Thus, we were not able to make any adjustments for age misreporting in this analysis, but chose to cap the life tables at 85 years to minimize the effect of this problem.

CONCLUSION

The AI/AN-specific life tables presented in this article are the first of their kind published for the Northwest AI/AN population. They provide an important tool to aid Northwest AI/AN communities, and the broader public health community, in monitoring the health of this population. The results highlight the need for these data to be disseminated and used for health assessment, program evaluation, and funding prioritization. Future efforts to establish linkage-corrected life tables for both urban and nonurban AI/ANs in other regions and nationally would facilitate heath surveillance efforts for this population.