Stroke in American Indians and Alaska Natives: A Systematic Review

Abstract

We conducted a systematic review of published studies on stroke epidemiology in American Indians and Alaska Natives (AI/ANs). We used MeSH terms and strict inclusion criteria to search PubMed, identifying a relevant sample of 57 refereed publications. We report a consensus view in which prevalent stroke is more common, and estimates of cerebrovascular risk factors are higher, among AI/ANs than among other US populations. Like other minority groups, AI/ANs suffer stroke at younger ages than do non-Hispanic Whites. However, data on AI/AN stroke mortality are significantly compromised by racial misclassification and nonrepresentative sampling. Studies correcting for these problems have found that stroke mortality rates among AI/ANs are among the highest of all US racial and ethnic groups. As with Black and non-Hispanic White stroke mortality, AI/AN stroke mortality varies by geographic region, with the highest rates in Alaska and the Northwest and the lowest in the Southwest. Our results underscore the need for a concerted national effort to collect accurate cross-sectional and longitudinal data on stroke in AI/ANs.

American Indian and Alaska Native (AI/AN) people experience significant health and socioeconomic disparities relative to other US populations, including elevated rates of obesity,¹ diabetes,² alcohol abuse,³ cigarette smoking,³ and poverty,⁴ as well as the second lowest rates of educational attainment in the nation, after Hispanics.⁵ All-cause mortality among AI/ANs is approximately 50% higher than among non-Hispanic Whites (hereinafter Whites),⁶ and life expectancy is shorter.⁷ Given these harsh disparities, it is not surprising that Native people have the highest or second highest prevalence of risk factors for cerebrovascular disease among US populations, as several recent studies have confirmed.^8–14

Yet the extant literature on AI/AN health presents a remarkable paradox: alongside findings of high stroke risk are many studies, old and new, concluding that AI/AN people have low stroke mortality.^10,15–21 Because of the need for reliable data to inform public health priorities and community interventions, an accurate understanding of cerebrovascular health and illness in this population is essential. Therefore, we undertook a systematic literature review²² to determine the current state of knowledge about the risk factors, incidence, prevalence, mortality, and outcomes of stroke among AI/ANs. In the process, we encountered widely recognized obstacles to obtaining epidemiological data on Native people.

METHOD

In July 2014, we searched the online PubMed database by using combinations of these MeSH terms: Indians, North American; Inuit; cerebrovascular disorders; and stroke. This strategy returned 138 unique articles published between 1966 and 2014. We limited our review to studies that provided population-level data on AI/ANs and were directly relevant to stroke epidemiology. Case studies, research on physiological mechanisms or prognostic factors for stroke, and studies of non-US populations were excluded. The last criterion disqualified articles on First Nations populations, an omission we justify by noting the divergent population histories and health systems of the United States and Canada.

To determine eligibility for review, we applied the following exclusion criteria in sequence to all 138 articles:

1. non-US sample (20 exclusions),

2. not about stroke (57),

3. no population-level data on AI/ANs (10),

4. limited to clinical prognostic factors (12), and

5. no data or discussion relevant to our epidemiological areas of interest (11).

This process reduced the initial list to just 28 titles, but searches of the bibliographies of the remaining publications identified 29 additional articles that met our criteria. Therefore, our final review includes 57 articles published between 1967 and 2014. The complete list is obtainable as an appendix (available as a supplement to the online version of this article at http://www.ajph.org).

RESULTS

Because many studies illuminate more than 1 area of interest, we have organized our results by topic: (1) stroke risk, (2) stroke prevalence, (3) stroke incidence and mortality, and (4) poststroke outcomes. Within each topic, we distinguish regional and national studies. We preface our results with a glance at 2 factors that affect assessments of epidemiological data on AI/ANs: pervasive racial misclassification in surveillance data and statistical challenges resulting from small population numbers.

Researchers have long warned that state- and national-level statistics on AI/AN health, especially data on mortality from any cause, are compromised by high rates of racial misclassification.^17,23–28 In most cases, AI/ANs are misclassified as White. Given the relative sizes of the 2 populations, these errors have a minimal effect on mortality rates reported for Whites, but a very large effect on those reported for AI/ANs. Recent studies have shown that as many as 45% of AI/AN decedents in specific regions are misclassified as another race on death certificates^28,29 and that correcting for misclassification can raise estimates of AI/AN cancer incidence, mortality from heart disease and stroke, and all-cause mortality by 8% to 43%.^25,29

No single, universally applicable definition can say who is or is not AI/AN. Although genetic markers of Native American ancestry are available,³⁰ genetic definitions of race are largely absent from epidemiological literature. One investigation of Texas residents who self-identified as Mexican American found that their DNA reflected a significant proportion of AI/AN ancestry.³¹ Nevertheless, the investigators argued that self-reported race/ethnicity was an important “proxy for lifestyle and other social factors”^31(p508) that trumped genetic markers as an epidemiological indicator, particularly in studies of stroke. This argument is persuasive insofar as the risk factors for stroke typically involve lifestyle behaviors. Meanwhile, the contribution of genetic factors to stroke risk is poorly understood,^32,33 both for AI/ANs and for other racial and ethnic groups.

A recent review article³⁴ addressed the complexities of defining AI/AN identity. Its authors noted that even self-identification, the method used by the US Census Bureau and preferred by many Native people, remains problematic. For example, before the 2000 Census, Americans were constrained to endorse a single racial category. Permitting endorsement of more than 1 race led to a dramatic increase in the number of people reporting AI/AN race, from 1.96 million in 1990 to 4.12 million in 2000.³⁵ The Census Bureau now recognizes 2 major classifications of Native heritage: AI/AN alone and AI/AN in combination with 1 or more other races. These changing definitions of AI/AN race make accurate comparisons between decennial counts impossible.¹⁷ The resulting absence of consistent information on the size of the AI/AN population over time compromises efforts to track changes in health and illness at the national level.

Regardless of their exact numbers, AI/ANs represent a small proportion of the US population, yet they are extremely diverse in culture, history, and geographic residence. For calculations of disease incidence and mortality, which rely on subsamples of the total population, smaller numbers lead to less precise estimates and larger standard errors.³⁶ For example, rates of mortality are commonly expressed per 100 000 population, whereas AI/AN samples used to estimate disease rates are typically smaller than this value by 2 or 3 orders of magnitude. Racial misclassification and nonrepresentative sampling methods further exacerbate the problem by undercounting the number of AI/ANs affected by a given disease.

Stroke Risk Factors

Regional studies have agreed that stroke risk factors in Native people are similar to those in the all-races population, although the relative contribution of individual risk factors likely differs. A clinic-based study of stroke patients in Arizona from 1990 to 1996 described comorbidities among 3 racial and ethnic groups: Whites, AI/ANs, and Hispanics. Race and ethnicity were self-reported by patients or their families. Chart review revealed that AI/AN patients with stroke often had hypertension, high cholesterol, or current smoking, all of which are widely recognized cerebrovascular risk factors.³⁷ In addition, AI/ANs were more likely than Whites and Hispanics in the sample to have a history of diabetes and heavy alcohol intake, defined as more than 24 ounces of beer, 12 ounces of wine, or 3 ounces of liquor per day. A decade later, a prospective, clinic-based study in Arizona found the same comorbidities in AI/AN stroke survivors (primarily Apache, Navajo, Hopi, and Pima), emphasizing hypertension, diabetes, previous stroke, and regular alcohol use.³⁸

Similarly, a telephone survey of residents of Montana found that self-reported cardiovascular risk factors were more prevalent in AI/ANs than in non-AI/ANs in 1999³⁹ and that, 4 years later, the prevalence of 2 or more risk factors in individual AI/AN adults had increased substantially, from about one third of adults in 1999 to 44% in 2003.⁴⁰ The 4 most commonly reported risk factors were obesity, smoking, hypertension, and high cholesterol.⁴⁰

The Strong Heart Study conducted prospective, population-based research among members of 13 AI tribes and communities residing in Arizona, Oklahoma, North Dakota, and South Dakota.⁴¹ Study investigators collected clinical data from 1989 to 2004 during study visits with 4507 tribal members, aged 45 to 74 years, who were free of stroke at baseline. Hypertension, diabetes, elevated fasting glucose, smoking, and albuminuria were all identified as risk factors for incident stroke.¹³

At the national level, a series of 4 studies^8,9,12,14 from the Centers for Disease Control and Prevention assessed self-reported health risk factors among people who self-identified as AI/AN. Three of them^9,12,14 compared AI/AN rates with those of other racial and ethnic minorities. In the first study,⁸ based on a random-digit-dialed telephone survey of AI/ANs conducted by the Behavioral Risk Factor Surveillance System (BRFSS) in 1997, 31% of respondents were current smokers, 22% reported diagnosed hypertension, 22% had body mass index consistent with obesity, 16% reported high cholesterol, and 7% reported diagnosed diabetes. Although 63% of this AI/AN population sample were younger than 45 years, and 90% were younger than 65 years, only 37% reported no cardiovascular risk factors.

The second study addressed minority communities in selected states, including Blacks, Asians, and Hispanics in addition to AIs (but not ANs).⁹ Self-reported data were collected from 2001 to 2002 by means of a telephone survey, supplemented in some locales by standardized interviews, conducted by the Racial and Ethnic Approaches to Community Health (REACH) project. Of the 4 groups studied, AIs of both sexes had the highest prevalence of obesity (men, 40%; women, 38%), current smoking (men, 43%; women, 37%), diagnosed diabetes (men, 17%; women, 20%), and diagnosed cardiovascular disease (men, 16%; women, 13%), and AI men had the highest prevalence of diagnosed hypertension (39%) and elevated cholesterol (37%). About 80% of AIs reported at least 1 risk factor, and one third reported 3 or more.⁹ In a follow-up study using REACH survey and interview data from 2009, AIs again reported the highest prevalence of obesity, current smoking, and diagnosed diabetes and cardiovascular disease, and diagnosed hypertension was again most prevalent in AI men.¹² Results were reported separately by sex and race/ethnicity in the form of median, high, and low percentages for each condition. For example, the median for hypertension among AI men was 43.9%.

Finally, a study conducted by BRFSS in 2003¹⁴ among AI/ANs, Asians, Blacks, Hispanics, and Whites used a random-digit-dialed telephone survey to assess the self-reported prevalence of 6 stroke risk factors: hypertension, diabetes, high cholesterol, obesity, current smoking, and lack of exercise. However, the associated publication reported only the prevalence of endorsing more than 1 risk factor, which was 26% among Asians, 36% among Whites, 40% among Hispanics, 47% among AI/ANs, and 49% among Blacks.¹⁴

Both regional and national studies have therefore agreed that stroke risk factors are highly prevalent among AI/ANs, and national studies have demonstrated that AI/ANs have the highest or second highest prevalence of stroke risk factors among all US racial and ethnic groups.

Stroke Prevalence

Information on stroke prevalence in AI/ANs has been slow to accumulate. A 1995 review of stroke epidemiology concluded that such data were simply lacking for this population.¹⁷ Since then, substantial data collection has shown that stroke is highly prevalent among AI/ANs^7,42–53 and, in particular, that Native people experience stroke at younger ages than Whites.^{25,37,54–56} Early onset of stroke might contribute to high prevalence rates, because younger stroke survivors are likely to live longer than older stroke survivors. However, most studies have been conducted at the national level; we found just 1 that addressed a geographically defined region.

In 2010, a population-based study of 499 AN men and women older than 45 years in the Norton Sound region of Alaska reported that 6.1% of men and 1.8% of women had a medically documented history of stroke.⁴² Most respondents were Inupiat, traditionally speakers of an Inuit language; their average age was 58 years. Despite the persistence of traditional dietary patterns, history of stroke was significantly associated with hypertension, hyperlipidemia, and physical inactivity. The authors could not determine whether adoption of nontraditional lifestyles was associated with incident stroke, nor did they compare recent stroke prevalence with past stroke prevalence.⁴²

Several national studies have reported similar findings. BRFSS data collected in 2005 indicated a self-reported stroke prevalence of 6% among single-race AI/ANs, the highest estimate among 6 populations evaluated.⁴³ A follow-up study using data from 2006 to 2010 found a mean value of 5.5% for AI/AN stroke prevalence over all 5 years of data collection, the highest among 5 population samples.⁴⁴ Data collected from 2004 to 2008 by the National Health Interview Survey revealed a broadly similar self-reported prevalence of 4.7% among AI/ANs, including 6.9% of men and 2.5% of women (although small numbers of women led to unreliable estimates).⁷ Overall, AI/ANs had the highest prevalence of the 5 populations studied; the next-highest estimates were for Blacks (3.7% overall, 3.5% of men, 3.8% of women) and Whites (2.4% overall, 2.5% of men, 2.4% of women).⁷

These national results appear alongside other, similar estimates in Table 1, which summarizes stroke prevalence data cited in a series of updates on cardiovascular disease from the American Heart Association, published annually in Circulation since 2006.^45–53 It is readily apparent that the 2 smallest populations, AI/ANs and Pacific Islanders, also have the least reliable estimates, in a classic illustration of the small-numbers problem. Nevertheless, in most years in this series, AI/ANs had the highest self-reported prevalence of stroke.

TABLE 1—

US Stroke Prevalence Among Those Aged 18 Years and Older With a Self-Reported History of Stroke by Race/Ethnicity: 2006–2014

Year/Source	AI/ANs, %	Asians, %	Pacific Islanders, %	Blacks, %	Hispanics, %	Whites, %
200645
NHANES	3.1	1.8	NA	3.5	2.2	2.3
NHIS	3.6	2.0	NA	3.3	NA	2.2
200746: NHIS or NCHS	5.1	2.4	8.1	3.2	2.8	2.5
200847
NHIS or NCHS	5.8a	2.0	b	3.4	2.2	2.3
BRFSS	6.0	1.6ac		4.0	2.6	2.3
200948
NHIS or NCHS	b	2.6	b	3.7	2.5	2.2
BRFSS	6.0	1.6ac		4.0	2.6	2.3
201049
NHIS or NCHS	3.9a	1.8	b	3.6	2.6	2.7
BRFSS	6.0	1.6ac		4.0	2.6	2.3
201150
NHIS or NCHS	b	1.3	b	3.8	2.0	2.5
BRFSS	6.0	1.6ac		4.0	2.6	2.3
201251
NHIS or NCHS	5.9a	2.0	10.6	3.9	2.6	2.5
BRFSS	5.8	1.4c		4.0	2.5	2.4
201352
NHIS or NCHS	4.6	2.7	b	4.5	2.8	2.3
BRFSS	5.9	1.5c		3.9	2.5	2.4
201453
NHIS or NCHS	b	1.8	b	3.9	2.7	2.5
BRFSS	5.8	1.9c		3.8	1.8	3.0

Note. AI/ANs = American Indians/Alaska Natives; BRFSS = Behavioral Risk Factor Surveillance System; NA = not applicable; NCHS = National Center for Health Statistics; NHANES = National Health and Nutrition Examination Survey; NHIS = National Health Interview Survey.

Source. Data were abstracted from the annual Heart Disease and Stroke Statistical Updates from the American Heart Association.

^aNote that these estimates are considered either unreliable or to be interpreted with caution.

^bNote that estimates are not reported because of large relative standard errors or unreliability.

^cThis estimate represents the combined prevalence for Asians and Pacific Islanders.

Stroke Incidence and Mortality

Although available data on stroke risk factors and prevalence in AI/ANs are based largely on self-report, most data on AI/AN stroke mortality are derived from death certificates. Assessments of decedents’ race on these documents are typically made by funeral directors, who often misclassify AI/ANs as White.^25,28 The statistical reliability of mortality data is further compromised by the small size and dispersed residence of the AI/AN population.^24,57,58 These shortcomings have brought us to the situation observed today, in which AI/ANs seem to inhabit 2 parallel but incompatible realities. In one, Native people have good cerebrovascular health, with relatively low stroke mortality. In the other, their disparities in cerebrovascular health are among the worst in the nation.

Table 2 summarizes the results of 20 studies that provided original data on Native stroke mortality. Some regional data go back as far as the 1950s, but most studies have concentrated on the period from 1980 to 2005. Immediately evident is the wide variability in reported mortality rates, even for Whites. Part of this variability might result from inconsistent determinations of the underlying cause of death in medical records and death certificates; misclassification of deaths from stroke has been shown to yield inaccurate mortality rates in the all-races population as well as among AI/ANs.⁷⁰ Although such variability complicates comparisons between studies,⁷¹ some regularities appear. All national studies of multiple populations found that Blacks and Whites had substantially higher rates of stroke mortality than Hispanics and Asians.^{10,18–21,25} In 3 studies that examined mortality before and after age 65 years,^10,25,62 stroke death before age 65 years was substantially more frequent in minority populations than in Whites. In studies of 3 or more racial and ethnic groups^10,15–21—except for a single report that explicitly corrected for racial misclassification²⁵—AI/ANs had the lowest or second lowest stroke mortality of all groups evaluated.

TABLE 2—

Stroke Mortality in American Indians and Alaska Natives: Regional and National Studies, United States, 1992–2014

Study	Sample	Data Source	Stroke Mortality	Remarks
Southwest
Sievers et al.59	5131 Pima Indians, 1975–1984	ICD-9 codes 431.0–434.9 and 436 on death certificates, medical records	AI 1.3/1000	Age- and sex-adjusted rates per 1000; racial misclassification unlikely.
Kattapong and Becker15	NHW, Hispanic, and AI residents of New Mexico (including Apache, Pueblo, Navajo, Zuni), 1958–1987	ICD-9 codes 430–438.9 on death certificates, medical records	AI/AN 52.2 in 1957; 31.3 in 1983	Age- and sex-adjusted rates per 100 000; in all years, stroke mortality was lower in AI than in Hispanic and NHW.
Sievers et al.60	5284 Pima Indians, 1965–1989	ICD-9 codes 431.0–434.9 and 436 on death certificates, medical records	AI 4.6/1000 in patients without diabetes, 1.3/1000 in patients with diabetes	Age-adjusted rates reported per 1000 in terms of sex and diabetes status; racial misclassification unlikely.
Hoehner et al.61	4623 Pima Indians, 1965–1998	ICD-9 codes 431.0–434.9 and 436 on death certificates, medical records	Rates reported separately for patients with and without diabetes in 8-y intervals	Racial misclassification unlikely; no calculation of overall stroke mortality.
Midwest and Mountain West
Reeves et al.62	AI/AN and NHW Wisconsin residents, 1984–1993	ICD-9 codes 430–438 on state death records	For men aged 45–65 y, age-specific incidence-density mortality rates per 100 000 were 40.9 for AI/AN, 27.4 for NHW; for men aged > 65 y, 463.2 for AI/AN, 396.9 for NHW	Rates presented separately by sex and age; no calculation of overall stroke mortality; likely racial misclassification acknowledged.
Harwell et al.55	AI/AN and NHW Montana residents, 1991–2000	ICD-9 codes 430–434 and 436–438 and ICD-10 codes I60–I69 on state death records	In 1991–1995, AI/AN 80, NHW 64; in 1996–2000, AI/AN 81, NHW 60	Age-adjusted rates presented per 100 000; age-specific results demonstrate much higher stroke mortality for AI/AN than NHW aged < 65 y.
Southeast
Sergeev16	Four racial/ethnic groups in Appalachia (undefined), 2000–2006	CDC, NCHS	AI 24.8, NHW 97.5, African American 100.2, Asian 18.7	Unadjusted rates presented per 100 000; very small AI sample, racial misclassification likely but not acknowledged.
Alaska
Schumacher et al.63	AN residents of Alaska, 1979–1998	Alaska Division of Public Health	In 1998, AN 38	Age-adjusted rates presented per 100 000; AN mortality higher than NHW since 1984.
Horner et al.64	AN and NHW residents of Alaska, 1984–2003	ICD-9 codes 430–434 and 436–438 and ICD-10 codes I60–I69 on death certificates	In 1984–1993, AN 71.2, NHW 64.1; in 1994–2003, AN 74.5, NHW 60.3	Age-adjusted rates presented per 100 000 for 2 different intervals.
National and multiregional
Welty et al.65	11 IHS regions nationwide, 1981–1983	IHS, NCHS; ICD-9 codes 430–438 on death certificates	32.3 across all IHS regions, ranging from 52.2 in Alaska and 50.9 in Nashville, TN, to 25.1 in Tucson, AZ, and 19.1 in Navajo Nation	Age-adjusted rates presented per 100 000; stroke mortality varied widely across IHS regions; racial misclassification on death records was acknowledged even in these regions.
Gillum17	National, 1988–1990	NHANES	Mortality substantially lower among AI/AN than among NHW and African American	No individual rates reported; racial misclassification documented in NHANES sample.
Lee et al.66	7327 AI in Arizona, Oklahoma, North Dakota, and South Dakota, 1984–1988	Strong Heart Study	4/1000	Five-year, age-adjusted rate for men and women aged 45–74 y; rates were highest in ND and SD; racial misclassification unlikely.
Ayala et al.18	National, 1995–1998	ICD-9 codes 433–434 and 436–438 on NVSS death certificates	AI/AN 48.6, NHW 73.7, African American 95.8	Age- and sex-adjusted rates presented per 100 000 for ischemic stroke; likely racial misclassification acknowledged.
Keppel et al.19	National, 1990–1998	ICD-9 codes 430–438 on NVSS death certificates	In 1998, AI/AN 19.1, NHW 23.3, African American 42.5, Hispanic 19.0	Age-adjusted rates presented per 100 000; likely racial misclassification acknowledged.
Casper et al.67	National, 1991–1998	ICD-9 codes 430–438 on NVSS death certificates	AI/AN 79	Age-adjusted rate per 100 000 for people aged ≥ 35 y; likely racial misclassification acknowledged.
CDC10	National sample < 75 y, 2002	ICD-10 codes I60–I69 on death certificates	AI/AN 37.5, NHW 54.2, African American 76.3, Hispanic 41.3, API 29.4	Age-adjusted rates presented per 100 000; stroke death before age 65 occurred disproportionately in minority populations; likely racial misclassification acknowledged.
Rhoades25	National, 1989–1998	IHS vital event data	For 1996–1998, AI/AN 26.2 and NHW 13.1 for ages 45–54 y; AI/AN 60.8 and NHW 36.9 for ages 55–64 y; AI/AN 346.4 and NHW 411.3 for ages ≥ 65 y	Mortality rates presented per 100 000, with correction for racial misclassification, for AI/AN, NHW, and all-races populations for 3 age categories and 4 reporting periods.
Mensah et al.20	National, 2003	CDC	AI/AN 41.3, NHW 56.4, African American 78.8, Hispanic 44.9, API 51.2	Age-adjusted rates presented per 100 000; likely racial misclassification acknowledged.
Lackland et al.21	National, 1999–2010	CDC	AI/AN stroke mortality < 40/100 000 since 2007	Stroke data summarized in a graph (their Figure 2); AI/AN mortality rates consistently lower than African American and NHW; no mention of racial misclassification.
Schieb et al.56	National, 1999–2009	IHS, NVSS	Overall AI/AN 115.5, with regional rates ranging from 144.3 in Alaska and 141.6 in Oklahoma, Kansas, and Texas to 78.3 in the Southwest	Age-adjusted mortality data per 100 000 from 637 counties designated by IHS as contract health service delivery areas, corrected for racial misclassification.

Note. AI = American Indian; AN = Alaska Native; API = Asian/Pacific Islander; CDC = Centers for Disease Control and Prevention; ICD-9 = International Classification of Diseases, 9th Revision⁶⁸; ICD-10 = International Classification of Diseases, 10th Revision⁶⁹; IHS = Indian Health Service; NCHS = National Center for Health Statistics; NHANES = National Health and Nutrition Examination Survey; NHW = non-Hispanic White; NVSS = National Vital Statistics System.

A closer look at a subset of these studies can shed light on this complex situation. Some of them foreground issues of high overall mortality and small numbers; others unambiguously indicate high AI/AN stroke mortality.

Southwest.

Although the most populous tribes in the Southwest are the Navajo, the Apache, and the Pueblo,⁷² the most thoroughly researched Southwestern tribe is the Pima of the Gila River Community in Arizona. Longitudinal epidemiological studies under way since 1965 have demonstrated that the Pima have an extremely high incidence of type 2 diabetes.⁷³ Two studies in our review assessed the association of diabetes with mortality from various causes in 6027 tribal members from 1975 to 1989.^59,74 Heart disease was a leading cause of death in diabetic participants, but was relatively rare in people without diabetes, whereas stroke mortality in both groups was lower than mortality from heart disease, infections, alcoholic liver disease, malignant neoplasms, and traumatic injury. Later studies with longer follow-up revealed a paradox: hypertension was associated with death by diabetic nephropathy among Pima with diabetes, but with death by stroke among Pima without diabetes.^60,61 The authors concluded that hypertension conferred a high risk of cardiovascular and cerebrovascular disease on all tribal members, except that patients with diabetes died of renal disease before they could develop life-threatening vascular disease.⁶⁰

Here we see a well-attested scenario in which low rates of stroke death actually indicate poor community health: high mortality in younger adults reduces the population vulnerable to stroke in later life. However, this scenario does not explain how the national AI/AN population can simultaneously have a high prevalence of stroke and stroke risk factors but a low rate of stroke mortality, because stroke prevalence and mortality were both low among the Pima.

Southeast.

A 2013 study of the Stroke Belt⁷⁵ used national mortality data from 2000 to 2006 on people aged 25 years and older to calculate stroke death rates for 4 racial groups in the Appalachian Mountain region (Table 2).¹⁶ Without comment, the author reported a rate of 24.78 per 100 000 for AI/ANs, about one quarter the magnitude of the rates reported for Whites and Blacks.¹⁶ Apart from likely issues of racial misclassification, this mortality inference was based on just 86 stroke deaths attributed to AI/ANs, who represented only 0.3% (49 573) of the total regional sample of 16.13 million. These results provide another instance of the small-numbers problem.

Midwest and Mountain West.

Two studies addressed stroke mortality in Wisconsin and Montana during the 1980s and 1990s.^55,62 Both found higher mortality for AI/ANs than for Whites, and analysis of annual trends for Montana⁵⁵ indicated declining mortality during the 1990s for Whites but not for AI/ANs. Both also reported significantly higher stroke mortality before age 65 years among AI/ANs. In Montana, for example, the proportion of stroke deaths before age 65 years among AI/AN men was triple that for White men, and the proportion for AI/AN women was quadruple that for White women.⁵⁵

Alaska.

By the 1980s, heart disease was a major cause of death for ANs. Available data show a steady increase in cerebrovascular mortality in this population through 1998, even as it declined among Whites in Alaska.⁶³ It is unclear, however, whether existing evidence supports the widely circulated hypothesis that cardiovascular and cerebrovascular disease were rare in ANs before 1950.⁷⁶ A study from 2003 used a literature review and reanalyses of mortality statistics on the Inuit and other AN groups to question the validity of this supposition. The authors concluded that rates of stroke mortality in recent decades were higher among ANs than among Whites, that AN stroke mortality continues to increase, and that claims of excellent vascular health among circumpolar tribes in the early 20th century are unfounded.⁷⁶

A 2009 study offered partial support for these conclusions.⁶⁴ By reviewing Alaska death certificates from 1984 to 2003, the authors determined that stroke mortality among ANs was high throughout this period, whereas it declined among Whites. Although they found no evidence for increasing rates of stroke among ANs, they noted that stroke mortality before age 45 years was substantially higher among ANs than among Whites, that AN women were at particularly high risk for stroke throughout adulthood, and that hypertension was the principal risk factor for stroke in ANs.⁶⁴

Multiregional data: the Strong Heart Study.

In 1998, researchers with the Strong Heart Study reported the results of a mortality survey for the period 1984 to 1988 for all members of participating tribes and communities aged 45 to 74 years.⁶⁶ All-cause mortality was higher in these samples than national rates in both White and all-races populations. Cardiovascular mortality varied significantly across regions, with the highest rates in North and South Dakota. Five-year, age-adjusted stroke mortality was calculated as 4 per 1000. The authors did not compare AI/AN stroke mortality rates with those of other races.

A 2008 publication reported stroke incidence and case fatality between 1989 and 2004 in 4507 Strong Heart participants with no previous history of stroke.¹³ Notably, this is the only study of AI/AN stroke incidence that we could identify. Stroke subtypes were similar to those in other large population studies in the United States: ischemic, 86%, and hemorrhagic, 14%. Mean age at first stroke was 66.5 years, the 1-year case fatality rate was 32%, and the age- and sex-adjusted rate of stroke incidence was 679 per 100 000. Overall stroke incidence was higher in this cohort than in national surveys of Whites and Blacks, and the case fatality rate was 1.5 times higher than rates calculated for large national studies.^13,46 The authors did not present overall estimates for stroke mortality or stroke prevalence, nor did they compare age at stroke onset for AI/ANs with results for other groups.

National data.

Regional variations in stroke mortality have been recognized among Whites and Blacks for many years.⁷⁷ Table 2 indicates notable regional variations among AI/ANs as well,^56,65,67 with an apparent trend in which mortality rates are higher west of Lake Erie and north of the 40th parallel.⁶⁷ Two studies in particular provide a useful overview.^56,65 Both used data from health centers operated by the Indian Health Service (IHS), supplemented by death certificate data, but they were conducted 20 years apart and used different approaches.

The 1993 study estimated age-adjusted annual stroke mortality in 11 IHS service areas across the United States and compared these rates with stroke mortality rates in the general US population.⁶⁵ AI/AN stroke mortality varied widely, with the lowest rates in the Southwest and the highest in Tennessee and South Dakota. The mean annual age-adjusted stroke mortality rate for all IHS areas was 32.3 per 100 000 from 1981 to 1983, lower than the all-races rate (35.8/100 000) in the same period.⁶⁵ However, the authors noted problems of racial misclassification even in IHS areas, and they cautioned that their estimates were lower bounds.

The 2014 study explicitly set out to test the hypothesis that pervasive racial misclassification invalidates most national reports on AI/AN stroke mortality.⁵⁶ Its approach updated the methodology used in the 1993 study⁶⁵ by examining data only from IHS contract service delivery areas (where racial misclassification in hospital records and death certificates is likely rare) and by using a linkage with IHS registration records spanning 2 decades to adjust for any misclassification that occurred even under these favorable circumstances. A systematic manual review further improved the study’s validity. In each contract service delivery area, mortality data on AI/ANs were compared only with mortality data on Whites residing in the same area. Like the 1993 study, this study found sharp regional variations, again noting that the Southwest had the lowest stroke mortality and Alaska and the Southern Plains (OK, KS, TX) had the highest.⁵⁶ In every region, stroke mortality was higher among AI/ANs than among Whites of the same age, and across all contract service areas, AI/ANs had a remarkably high rate of stroke death: 115.5 per 100 000. As the authors noted, this is twice the target for Healthy People 2010 and triple that for Healthy People 2020.^78,79 Consistent with the results of earlier studies, disparities in mortality were most severe for AI/ANs younger than 55 years.⁵⁶

Poststroke Outcomes

Few studies in our review offered information on poststroke outcomes among AI/ANs. Two studies using death certificate data found that among AI/ANs, as in the all-races population, hemorrhagic strokes are more often fatal than ischemic strokes,¹⁸ and women have higher mortality from hemorrhagic stroke than men.⁸⁰ Only the Strong Heart Study offered an estimate of case fatality rates, finding higher rates among AIs than those reported for Whites and Blacks in other cohort studies.¹³ A clinic-based study suggested an explanation for this situation: compared with other races, AI/AN patients delayed seeking therapy for acute stroke for longer periods of time, with barriers to care likely contributing to delays.³⁸ An Alaska-based study also noted serious geographic barriers to stroke care. Almost half of ANs live in rural villages in which specialty care requires air travel, and only 1 regional hospital in the whole state had a computerized tomography scanner in 2007.⁸¹

DISCUSSION

Our critical review of the literature on stroke among AI/ANs provides a context for the conflicting results published over the past 3 decades. Generally, we found that data on stroke prevalence and risk factors were more reliable than data on mortality, which were compromised by racial misclassification and nonrepresentative sampling, and that data on stroke incidence were limited to a single study.¹³

AI/AN Stroke Risk Factors and Prevalence

The consensus view was that stroke risk factors among AI/ANs are similar to those among other US populations, and many are related to lifestyle: physical inactivity, high-fat diet, smoking, and type 2 diabetes. The single most frequently cited stroke risk factor for AI/ANs was hypertension,^{13,37,38,42,64,66,67,82} and almost half of Native adults in national surveys reported the presence of 2 or more risk factors. Overall, the prevalence of stroke risk factors in AI/ANs was among the highest of all racial and ethnic groups.^9,41

We found a similar consensus on the prevalence of stroke. Numerous studies, mostly at the national level, found that prevalent stroke was significantly more common among AI/ANs than among Blacks, Whites, Hispanics, and Asians.^42,45–53 Recent studies have estimated AI/AN stroke prevalence at 5.5% to 6.9%.^7,44

AI/AN Stroke Mortality and Poststroke Outcomes

Despite contrary findings from national surveys, stroke mortality also appears to be extremely high among AI/ANs. Studies that found low rates of stroke death relative to other US racial and ethnic groups typically did not account for racial misclassification, and those few^25,56,64 that adequately controlled for this problem returned strikingly high rates—as high as 115.5 per 100 000.⁵⁶ The prevalence of stroke in a population is conditioned by 2 factors: rate of stroke incidence and rate of survival after stroke.⁴³ Low rates of stroke mortality are inconsistent with elevated stroke prevalence unless few strokes are fatal. Yet as numerous studies concurred, rates of stroke mortality among AI/ANs are higher at younger ages than among other racial and ethnic groups,^{25,37,54–56,82} and the single study that defined a case fatality rate found higher rates among AI/ANs than among other groups.¹³ Therefore, our review indicates that stroke deaths among AI/ANs have been undercounted because of racial misclassification, and the true number is not known.

Two other factors likely complicate analyses of AI/AN stroke deaths and contribute to unexpectedly low estimates of stroke mortality: high rates of all-cause mortality at all ages and nonrepresentative population sampling. As shown by the body of research on the Pima, a relatively unhealthy community may have low rates of cerebrovascular mortality simply because so many tribal members die before age 65 years. Studies of the all-races population have revealed that most stroke deaths occur after age 65 years,⁸³ yet people aged 65 years and older represent only 9.5% of the AI/AN population, compared with 12% for Blacks and 18.8% for Whites.⁷ We might thus be seeing survivorship bias in many tribal populations.⁸⁴ People who would be at high risk for stroke after age 65 years, such as men who drink and smoke, are removed from the population by premature death. The surviving population therefore has a better cerebrovascular risk profile than the decedents. This view is reinforced by a recent study comparing the 15 leading causes of death among AI/ANs with those among Whites.⁶ Stroke was lower in the rankings for AI/ANs than for Whites, and diabetes and chronic liver disease were much higher. Reducing the incidence of diabetes and heavy drinking among Native people would likely bring the mortality profile for AI/ANs closer to that for Whites.

As it is, stroke mortality shows wide variation across AI/AN communities,⁵⁶ making oversampling an essential strategy to produce a study population large enough to adjust for confounding by racial misclassification, geographic heterogeneity, and high all-cause mortality. Yet such national surveillance efforts as the BRFSS¹⁴ and REACH^9,12 studies did not oversample AI/ANs. The REACH survey was limited to AI/ANs in a few communities, and the BRFSS, which permits oversampling of minorities in individual states,⁸⁵ does not uniformly require that approach.⁸⁶

In addition to these methodological issues, we note a dearth of studies on poststroke outcomes in AI/ANs. Data on the US all-races population have shown that stroke is the leading cause of long-term disability nationwide. In all subpopulations, stroke survivors cope with significant physical, cognitive, and emotional impairments, including reduced mobility and inability to perform activities of daily living.⁸⁷ A substantial fraction of survivors are either partially or fully dependent on caregivers,⁸⁷ and many suffer depression or anxiety or both.^88,89 Given the high prevalence of stroke in Native communities, it is likely that many AI/AN stroke survivors need both specialty and informal care, increasing the burden on community resources.

Theoretical Implications

The results of our review speak to popular theoretical frameworks used to conceptualize population health. Over the past 5 decades, cerebrovascular disease in AI/ANs has often been discussed in terms of transitions. In the 1970s, Omran used historical demographic data from Europe to propose an “epidemiologic transition” in global population history.^90,91 In this model, high birth rates, high infant mortality, and high adult mortality as a result of famine and infectious disease—which were typical of Europe during the medieval and early modern periods—gave way during the 20th century to low birth rates, low infant mortality, and low adult mortality resulting primarily from chronic illness (heart disease, stroke, cancer).

Consistent with this model, the earliest studies in our review,^15,17,41,82 as well as the Heckler report from 1986,⁹² typically noted a recent or ongoing shift in AI/AN mortality profiles from infectious diseases to such “modern” ailments as heart disease and stroke, which were associated with urbanization and Westernization.⁹² An analogous epidemiological transition has been proposed by Diamond⁹³ for indigenous populations in Asia, Australia, and Africa. Diamond has argued that indigenous people have excellent health and long life spans as long as they follow traditional lifestyles, but they begin to suffer serious noncommunicable diseases (e.g., diabetes, heart disease, stroke, cancer) once they adopt a Western lifestyle. All of these arguments are consistent with the notion that AI/ANs are still catching up to Whites in terms of cerebrovascular disease trends, and all are consistent with the relatively low AI/AN stroke rates reported in the recent past.^9,65,94

Our review found no supporting evidence for such inferences.⁷² Indeed, 2 studies that attempted a systematic application of Omran’s⁹⁰ model to AI/AN communities concluded that it either required substantial modification⁹⁵ or simply did not fit well.⁹⁶ A glance at AI/AN population history can explain why. The people of North America entered an era of violent transition in the 16th century with the arrival of the first European colonists, who brought infectious diseases previously unknown in the Western Hemisphere. Successive epidemics dramatically reduced Native populations over the next 300 years.⁹⁷ The 17th century saw the first dispossession of North American tribes from their ancestral lands, a trend that accelerated westward through the 18th and 19th centuries.⁹⁷ Traditional economies and dietary practices were eradicated from the Atlantic to the Pacific, ensuring that many tribes either vanished or adopted lifestyles remote from those of their forebears. The American Indian boarding school program, which began in the 19th century and reached its peak in the 20th, forcibly removed Native children from their homes and compelled them to assimilate to White society, further accelerating the loss of cultures and lifestyles.^98,99 The distribution of commodity foods on reservations, institutionalized in the 1930s and continuing today,¹⁰⁰ fuels diets rich in salt, sugar, and fat that are completely alien to indigenous practices.^101,102 Given the specificity and long duration of this traumatic history, we caution against any assumptions that AI/ANs are just emerging from a premodern epidemiological regime.

Practical Implications

Our findings suggest future directions for public health research, especially if US agencies want to meet the goals set by Healthy People 2020.⁷⁹ First, more large-scale longitudinal studies of AI/AN populations, on the scale of the Framingham Study, are needed to collect adequate data on stroke incidence and mortality. Second, national surveillance efforts must oversample AI/ANs of all ages and in all regions, and make more strenuous efforts to classify race appropriately, to yield health data as sensitive and useful as those already available on other Americans. Third, a concerted effort is required to assess the number and needs of Native stroke survivors, a population likely to grow in the next decades. Finally, more effective and culturally congruent interventions are necessary to address stroke risk factors in AI/AN communities.

In the overall AI/AN population, we single out 2 neglected groups for special attention: urban AI/ANs and stroke survivors. Although Census data have for some decades shown that most AI/ANs have urban or suburban residence,^103,104 community-based research overwhelmingly focuses on rural reservation residents. This is also the population served by the IHS. Therefore, urban AI/ANs remain a largely invisible majority for whom comprehensive study is urgently needed. Regarding stroke survivors, we are ignorant of which communities or geographic regions have large concentrations of people living with the aftermath of stroke and what their needs might be regarding specialty health care or activities of daily living.

Strengths and Limitations

We note 2 limitations of this review. First, the PubMed search algorithm that we used was neither sensitive nor specific, because it missed many titles relevant to AI/AN stroke epidemiology and returned many others that we identified as not about stroke. Nevertheless, we believe that our scrutiny of bibliographies in the publications that met our inclusion criteria compensated for this shortcoming, especially because Native stroke epidemiology is a new field with far fewer studies than, for example, stroke in the all-races population.

Second, we engaged only in evidence synthesis and did not attempt to conduct a meta-analysis of the findings we reviewed. Given the defects in much of the existing data, however, we feel that this soft approach is appropriate to the current state of the field, and we look forward to analyses of more robust data in the future.

Conclusions

Despite limitations, our systematic review adds to the literature by highlighting key factors in AI/AN stroke epidemiology: the unique age structure⁷ and mortality profile⁶ of Native populations, both of which are skewed toward younger ages than in the all-races population; the high prevalence of risk factors and diagnosed stroke; the presence of widespread racial misclassification and undersampling in stroke mortality statistics; and the methods that have been used successfully to correct for these problems in regional population samples.⁵⁶ After correction, it appears that stroke mortality among AI/ANs is higher than that among the White and all-races populations and probably among the highest of all racial and ethnic groups. We offer these results in the hope that they will contribute to the state of knowledge on stroke in AI/ANs and, in particular, that they will help to correct widespread misconceptions about cerebrovascular health and illness in this unique population.