Stroke‐Related Mortality in the United States–Mexico Border Area of the United States, 1999 to 2018

Abstract

BACKGROUND

The United States (US)‐Mexico border is a socioeconomically underserved area. We sought to investigate whether stroke‐related mortality varies between the US border and nonborder counties.

METHODS AND RESULTS

We used death certificates from the Centers for Disease Control and Prevention Wide‐Ranging Online Data for Epidemiologic Research database to examine stroke‐related mortality in border versus nonborder counties in California, Texas, New Mexico, and Arizona. We measured average annual percent changes (AAPCs) in age‐adjusted mortality rates (AAMRs) per 100 000 between 1999 and 2018. Overall, AAMRs were higher for nonborder counties, older adults, men, and non‐Hispanic Black adults than their counterparts. Between 1999 and 2018, AAMRs reduced from 55.8 per 100 000 to 34.4 per 100 000 in the border counties (AAPC, −2.70) and 64.5 per 100 000 to 37.6 per 100 000 in nonborder counties (AAPC, −2.92). The annual percent change in AAMR initially decreased, followed by stagnation in both border and nonborder counties since 2012. The AAPC in AAMR decreased in all 4 states; however, AAMR increased in California’s border counties since 2012 (annual percent change, 3.9). The annual percent change in AAMR decreased for older adults between 1999 and 2012 for the border (−5.10) and nonborder counties (−5.01), followed by a rise in border counties and stalling in nonborder counties. Although the AAPC in AAMR decreased for both sexes, the AAPC in AAMR differed significantly for non‐Hispanic White adults in border (−2.69) and nonborder counties (−2.86). The mortality decreased consistently for all other ethnicities/races in both border and nonborder counties.

CONCLUSIONS

Stroke‐related mortality varied between the border and nonborder counties. Given the substantial public health implications, targeted interventions aimed at vulnerable populations are required to improve stroke‐related outcomes in the US‐Mexico border area.

Nonstandard Abbreviations and Acronyms

AAMR

age‐adjusted mortality rate

AAPC

average annual percent change

Clinical Perspective

What Is New?

• Between 1999 and 2018, residents of both border and nonborder counties in California, Texas, New Mexico, and Arizona experienced a decline in stroke‐related mortality, although the decline has slowed in the past decade.

• Stroke‐related mortality varied among older adults and non‐Hispanic White adults in the border versus nonborder counties.

What Are the Clinical Implications?

• Overall and demographic disparities related to stroke‐related mortality in border versus nonborder counties are concerning.

• Targeted interventions aimed at narrowing clinical and socioeconomic inequalities may diminish gaps in stroke‐related mortality in the US–Mexico border area.

The United States (US)‐Mexico border region stretches ≈2000 miles and covers 62 miles north and south of the international border.¹, ² A total of 44 counties in 4 US states (Arizona, California, New Mexico, and Texas) encompass 53% of ≈15 million people residing in the border region.¹ This is a culturally diverse area where different civilizations from the US and Mexico connect across geographical borders.³ The US border region has witnessed a significant population growth over the years, with southwest border counties exhibiting a ≈30% population increase in the 1990s.⁴ Moreover, this region faces medical and socioeconomic challenges, demonstrating a wider socioeconomic gap between border counties and the rest of the US.¹, ²

Stroke remains the fifth leading cause of mortality in the US and the third leading cause of death in the border area.⁵, ⁶, ⁷ Besides cardiovascular disease burden, social disparities influence stroke‐related morbidity and mortality.⁸, ⁹ Stroke has shown to correlate with social determinants of health, including but not limited to low education, socioeconomic depression, healthcare access, unemployment, and social isolation.⁸, ⁹ Because social and health inequities in the region may influence the incidence and prevalence of stroke, it is imperative to investigate stroke‐related mortality trends in border counties compared with nonborder counties. A detailed assessment of the epidemiological profile of stroke‐related mortality may inform policymakers and healthcare professionals to improve care for the socially disadvantaged population residing in this dynamic area. Consequently, we compared demographic and geographical trends in stroke‐related mortality in the border versus nonborder counties in 4 border states (Arizona, California, New Mexico, Texas) using a national database of death certificates.

METHODS

Data Availability Statement

The Centers for Disease Control and Prevention Wide‐Ranging Online Data for Epidemiologic Research data sets used in this project are publicly available and are easily replicable from the methods described in the article.¹⁰

Data Source

We used the Centers for Disease Control and Prevention Wide‐Ranging Online Data for Epidemiologic Research—Multiple Cause of Death database for this analysis.¹⁰ The Multiple Cause of Death database is composed of death certificates for US residents. Each death certificate contains data on a single underlying cause of death, up to 20 additional causes, and demographic characteristics of the decedents. We identified natural deaths attributed to stroke—defined by the World Health Organization as the disease or injury that initiated the events leading directly to death as entered by the physician on the death certificate¹¹—using International Classification of Diseases, Tenth Revision (ICD‐10) codes I60 to I69. We focused on stroke‐related mortality in counties located in California, Texas, New Mexico, and Arizona further stratified into border counties (counties within 100 km [62 miles] of the US–Mexico border defined by the 1983 La Paz Agreement¹²) and nonborder counties.

This study did not require institutional review board approval because we analyzed government‐issued public use data without individual identifiable information.

Data Extraction

We abstracted the data on stroke‐related mortality in both border and nonborder counties in California, Texas, New Mexico, and Arizona (Table S1). We abstracted the number of stroke‐related deaths and population sizes from 1999 to 2018. We abstracted the data on age, sex, ethnicity/race (non‐Hispanic White, non‐Hispanic Black, non‐Hispanic American Indian/Alaskan Native, non‐Hispanic Asian/Pacific Islander, and Hispanic), and location of death. We grouped non‐Hispanic American Indian/Alaskan Native and non‐Hispanic Asian/Pacific Islander adults in “other” because of the low death counts in these groups. We grouped age into young (<45 years), middle aged (45–64 years), and older (≥65 years) adults. Location of death was categorized as home, hospital (inpatient, outpatient, or emergency room), hospice, nursing home/long‐term care, and others.

Statistical Analysis

We calculated crude death rates for individual years between 1999 and 2018 by dividing the number of stroke‐related deaths by the total corresponding population. We applied direct standardization for age‐adjustment of mortality rates using the 2000 US standard population.¹³ We examined mortality trends to identify changes in a slope using Joinpoint Regression Program version 4.7.0.0, which models consecutive linear segments on a log scale connected by joinpoints where the segments meet.¹⁴ Annual percent change (APC) with 95% CIs in age‐adjusted mortality rates (AAMRs) were estimated for the line segments linking joinpoints.¹⁴ The weighted average of the annual percent change was estimated to calculate the average annual percent change (AAPC) for the entire study period—with the weights equal to the length of the annual percent change interval.¹⁴

We applied the following settings to the Joinpoint Regression Program for the analyses: (1) grid search method, 2, 2, 0; (2) number of joinpoints, 0 to 3; (3) model selection method, permutation test; and (4) annual percent change/AAPC/tau 95% CI estimation, parametric method. For interpretation, slopes were considered increasing or decreasing if the estimated slope differed significantly from zero.¹⁵, ¹⁶ We applied a specific procedure‐comparability test to determine whether 2 regression mean functions are parallel because of different intercepts (test of parallelism).¹⁷ For all analyses, statistical significance was set at 5%.

RESULTS

Between 1999 and 2018, 56 019 stroke‐related deaths occurred in the border counties (147 408 326 patient‐years), corresponding to an overall AAMR of 39.3 (95% CI, 39.0–39.6) per 100 000 patient‐years. In comparison, 516 329 stroke‐related deaths occurred (1 243 119 793 patient‐years) in nonborder counties, corresponding to an overall AAMR of 45.2 (95% CI, 45.0–45.3) per 100 000 patient‐years. Overall, stroke‐related mortality was higher in nonborder than border counties in older adults, men, and non‐Hispanic Black adults versus their counterparts (Table 1).

Table 1.

Stroke‐Related Mortality in Counties Located in the US–Mexico Border Area Versus Nonborder Area, 1999 to 2018

	Border Counties				Nonborder Counties
	Events	Patient‐Years	Crude Mortality Rate (95% CI)	AAMR (95% CI)	Events	Patient‐Years	Crude Mortality Rate (95% CI)	AAMR (95% CI)
Overall	56 019	147 408 326	38.0 (37.7–38.3)	39.3 (39.0–39.6)	516 329	1 243 119 793	41.5 (41.4–41.6)	45.2 (45.0–45.3)
Age, y
<45	1447	96 687 275	1.5 (1.4–1.6)	1.7 (1.6–1.8)	12 453	804 641 323	1.5 (1.4–1.6)	1.7 (1.6–1.8)
45–64	6757	32 549 26	20.8 (20.3–21.3)	20.0 (19.5–20.5)	64 322	294 174 671	21.9 (21.7–22.0)	21.1 (20.9–21.2)
≥65	47 815	18 171 784	263.1 (260.8–265.5)	267.2 (264.8–269.6)	439 554	144 303 799	304.6 (303.7–305.5)	311.8 (310.9–312.7)
Sex
Female	32 010	74.239.985	43.1 (42.6–43.6)	38.5 (38.1–38.9)	304 129	625 194 548	48.6 (48.5–48.8)	44.6 (44.5–44.8)
Male	24 009	73 168 341	32.8 (32.4–33.2)	39.6 (39.1–40.1)	212 200	617 925 245	34.3 (34.2–34.5)	44.9 (44.7–45.1)
Ethnicity/race
Non‐Hispanic White	33 016	54 104 402	61.0 (60.4–61.7)	38.6 (38.2–39.0)	352 560	582 042 221	60.6 (60.4–60.8)	44.7 (44.5–44.8)
Non‐Hispanic Black	1722	5 117 360	33.7 (32.1–35.2)	57.4 (54.6–60.3)	50 220	106 388 091	47.2 (46.8–47.6)	66.8 (66.2–67.5)
Hispanic	18 557	78 661 800	23.6 (23.3–23.9)	37.8 (37.3–38.4)	73 775	423 864 173	17.4 (17.3–17.5)	39.8 (39.1–39.6)
Other	2724	9 524 764	28.6 (27.5–29.7)	37.7 (36.3–39.2)	39 774	130 825 308	30.4 (30.1–30.7)	37.4 (37.1–37.8)

We grouped non‐Hispanic American Indian/Alaskan Native and non‐Hispanic Asian/Pacific Islander adults in “Other” because of the low death counts in these groups. AAMR indicates age‐adjusted mortality rate.

Between 1999 and 2018, AAMR reduced from 55.8 (95% CI, 53.8–57.8) per 100 000 to 34.4 (95% CI, 33.0–35.4) per 100 000 in the border counties (AAPC, −2.70; 95% CI, −3.24 to −2.14), and 64.5 (95% CI, 63.7–65.2) per 100 000 to 37.6 (95% CI, 37.2–38.1) per 100 000 in nonborder counties (−2.92; 95% CI, −3.36 to −2.48; Table 2). The annual percent change in AAMR initially decreased, followed by stagnation in both border and nonborder counties (Figure 1).

Table 2.

Trends in Stroke‐Related Mortality in Counties Located in the US–Mexico Border Area Versus Nonborder Area, 1999 to 2018

	Border Counties		Nonborder Counties		Test for Parallelism
	AAMR (1999–2018)	AAPC (95% CI)	AAMR (1999–2018)	AAPC (95% CI)	P Value
Overall	55.8–34.4	−2.70 (−3.24 to −2.14)	64.5–37.6	−2.92 (−3.36 to −2.48)	0.60
Age, y
<45	1.7–1.4	−1.00 (−2.49 to 0.51)	1.8–1.4	−1.15 (−1.93 to −0.36)	0.42
45–64	24.1–19.4	−1.69 (−2.98 to −0.38)	25.5–20.2	−1.25 (−1.83 to −0.67)	0.04
≥65	390.4–230.5	−2.91 (−3.52 to −2.31)	455.9–254.9	−3.14 (−3.64 to −2.64)	0.33
Sex
Female	53.9–33.3	−2.80 (−3.39 to −2.21)	63.3–36.3	−3.00 (−3.48 to −2.52)	0.77
Male	57.3–34.6	−2.59 (−3.23 to −1.95)	64.9–38.4	−2.99 (−3.40 to −2.57)	0.50
Ethnicity/race
Non‐Hispanic White	55.4–34.0	−2.69 (−3.36 to −2.02)	63.5–37.3	−2.86 (−3.31 to −2.40)	0.02
Non‐Hispanic Black	91.7–54.6	−3.03 (−5.34 to −0.66)	91.0–54.4	−2.93 (−3.66 to −2.19)	0.20
Hispanic	51.6–33.3	−2.25 (−3.08 to −1.42)	53.8–34.7	−2.55 (−3.08 to −2.02)	0.13
Other	50.1–28.5	−3.29 (−5.20 to −1.34)	57.8–31.4	−3.16 (−3.69 to −2.62)	0.95

We grouped non‐Hispanic American Indian/Alaskan Native and non‐Hispanic Asian/Pacific Islander adults in “Other” because of the low death counts in these groups. AAMR indicates age‐adjusted mortality rate; and AAPC, average annual percent change.

Figure 1. Age‐adjusted stroke mortality rates in the border and nonborder counties in the United States, 1999 to 2018.

Figure illustrates observed and model‐adjusted mortality rates with APC (95% CI). APC indicates annual percent change.

State Stratified Analyses

Stroke‐related mortality varied across states in relation to border versus nonborder counties. California’s border counties had the highest mortality rates, whereas those located in Texas had the lowest mortality rates (Table S2). In contrast, Texas’ nonborder counties had the highest mortality rates, and those located in Arizona had the lowest mortality rates (Table S3).

Between 1999 and 2018, AAPC in AAMR decreased in all 4 states encompassing border and nonborder areas (Tables S2 and S3). In the border area, after the initial decline, the annual percent changes in AAMRs stalled in Arizona since 2014, New Mexico since 2015, and Texas since 2004, but increased in California’s counties since 2012 (3.94; 95% CI, 1.25, 6.71; Figure 2). In the nonborder areas, after the initial decrease, the annual percent change in AAMR stalled in nonborder counties in Arizona, California, and New Mexico since 2012 and Texas since 2011.

Figure 2. Age‐adjusted stroke‐related mortality rates in the border and nonborder counties, stratified by states in the United States, 1999 to 2018.

Figure illustrates observed and model‐adjusted mortality rates with APC (95% CI). APC indicates annual percent change.

Age‐Stratified Analyses

The age‐specific mortality rates increased exponentially with age for both border and nonborder counties (Figure S1). The AAPCs in AAMRs for all age categories are reported in Table 2, showing a significant difference in the middle‐aged group among border and nonborder counties. In border counties, the annual percent change in AAMR remained stable for young adults during the study period but decreased in middle‐aged adults since 2002 (−1.24; 95% CI, −1.88 to −0.60). For nonborder counties, after the initial decrease, the annual percent change in AAMR stalled in young adults (−0.73; 95% CI, −2.57 to 1.14) and middle‐aged adults (2.28; 95% CI, −0.45 to 5.09) since 2011 and 2014, respectively.

The annual percent change in AAMR decreased for older adults between 1999 and 2012 for border (−5.10; 95% CI, −5.62 to −4.57) and nonborder counties (−5.01; 95% CI, −5.44 to −4.57), followed by the increase in the border (1.99; 95% CI, 0.19–3.83) and stagnation in nonborder counties (1.02; 95% CI, −0.47 to 2.53).

Sex‐Stratified Analyses

Between 1999 and 2018, the AAPC in AAMR decreased for both sexes across the border and nonborder counties (Table 2). After an initial decline, the annual percent change in AAMR increased in men (1.91; 95% CI, 0.02–3.85) and stalled in women (1.22; 95% CI, −0.53 to 3.00) since 2012 in border counties, whereas the annual percent change in AAMR showed arrest for both sexes in nonborder counties between 2012 and 2018 (Figure 3).

Figure 3. Age‐adjusted stroke‐related mortality rates in the border and nonborder counties, stratified by sex in the United States, 1999 to 2018.

Figure illustrates observed and model‐adjusted mortality rates with APC (95% CI). APC indicates annual percent change.

Ethnicity/Race‐Stratified Analyses

The AAPC in AAMR decreased for all ethnicities/races; however, it differed significantly for non‐Hispanic White adults in border (−2.69; 95% CI, −3.36 to −2.02) and nonborder counties (−2.86; −3.31 to −2.40; Table 2). In border counties, the annual percent change in AAMR initially decreased for all ethnicities/races, followed by stagnation in non‐Hispanic Black and Hispanic adults (Figure 4). However, after an initial decrease, the annual percent change in AAMR increased in non‐Hispanic White adults in both border (2.98; 95% CI, 0.97–5.03) and nonborder counties (1.38; 95% CI, 0.04–2.74) since 2012.

Figure 4. Age‐adjusted stroke‐related mortality rates in the border and nonborder counties, stratified by ethnicity/race in the United States, 1999 to 2018.

Figure illustrates observed and model‐adjusted mortality rates with APC (95% CI). APC indicates annual percent change.

Location of Death

Between 2003 and 2018, the AAPC in the proportion of deaths from stroke in the border counties increased at home (3.15; 95% CI, 1.74–4.58), but decreased at hospice facilities (−4.21; 95% CI, −6.72 to 1.64), inpatient settings (−1.11; 95% CI, −1.60 to −0.62), and nursing home/long‐term care facilities (−1.90; 95% CI, −3.48 to −0.29; Figure S2).

Similarly, the AAPC in the proportion of deaths from stroke in the nonborder counties increased at home (4.39; 95% CI, 3.13–5.67), but decreased at hospice facilities (−6.22; 95% CI, −8.24 to −4.16), inpatient settings (−1.74; 95% CI, −2.39 to −1.08), and nursing home/long‐term care facilities (−2.03; 95% CI, −3.43 to −0.61; Figure S3).

DISCUSSION

Stroke‐related mortality varied between the US–Mexico border and nonborder counties. Overall, nonborder counties had higher mortality rates than border counties, and demographically, older adults, men, and non‐Hispanic Black adults had higher mortality rates than their counterparts. After the initial downtrend, mortality decline has stalled in both areas since 2012; however, there was considerable heterogeneity in mortality trends across border states and demographic subgroups. California’s border counties demonstrated a rise in mortality since 2012, whereas mortality decline has stalled in all other states. Non‐Hispanic White adults of border counties experienced a significant increase in mortality than those living in nonborder counties during the second half of the study. Finally, a higher number of individuals died at home, whereas deaths decreased at hospice facilities, hospitals, or nursing home/long‐term care facilities in both border and nonborder counties.

Contrasting mortality trends between the border and nonborder counties can be elucidated based on the heterogeneities related to demographic characteristics, cardiovascular risk burden, socioeconomic challenges, and limited access care among residents of border areas.¹⁸ In 2008, nearly 1 in 2 residents in border counties were non‐Hispanic individuals.¹⁸, ¹⁹, ²⁰ The border population is aging, and individuals aged ≥65 years may increase by 18% in the 4 border states by 2030.¹⁹, ²⁰ Hence, a rise in stroke‐related deaths in older adults may predict a concerning upsurge in total mortality burden in the future. From a socioeconomic perspective, as per 2007 estimates, the annual income per capita was $26 842 in the border counties compared with $39 013 in the border states and $38 839 in the United States, translating into more than twice the poverty rate (25%) in the border area than the national level (13%).²⁰, ²¹ Of the 10 most impoverished counties in the US, 3 (Starr, Maverick, and Hudspeth) belonged to Texas.²⁰ Of a total of 44 border counties, 48% were designated as economically distressed counties.²⁰ Between 2000 and 2003, about 23% of the border residents lacked health insurance coverage compared with 15% health insurance coverage nationally.²² In 2008, the unemployment rate was 11.5% in the border region and 5.6% in the US.²³

On the same note, the disproportionate burden of cardiometabolic risk factors in the border counties may further contribute to stroke‐related mortality. For instance, the prevalence of diabetes mellitus in the border area was ≈9.5% compared with 8.0% in the overall US in 2007.²² The hospital discharge rates of diabetes mellitus among border county residents were higher than their nonborder counties (16.6 per 10 000 versus 14.9 per 10 000).²⁴ A county‐level survey (1999–2008) showed suboptimal blood pressure control among men in counties along the US–Mexico border in Texas, New Mexico, and Arizona.²⁵ A significantly higher age‐adjusted prevalence of obesity (22%) and physical inactivity (43%) was reported among border residents.²² The combination of suboptimal cardiometabolic profile and less favorable social determinants of health in the border population contribute to health disparities.

As per the US–Mexico Border Health Commission 2010 update, the hospital discharge rate for stroke was significantly lower for border residents than for nonborder residents (28.0 per 10 000 versus 34.0 per 10 000).²² The lower hospital discharge rate may be attributed to a greater likelihood of border residents to die from stroke without being hospitalized, or a higher likelihood to die during or following hospitalization, thus eliminating rehospitalization for future treatment of the same condition.²⁶ Moreover, certain ethnic/racial minorities might prefer to die at home. Our findings complement these observations demonstrating a more significant number of individuals dying at home than in the hospital setting in the border area.²⁷

Several limitations of this study need to be acknowledged. Because vital statistics and census population data rely on death certificates, potential coding errors can exist.²⁸ Inaccurate ascertainment of demographic data is also subject to misclassification. We could not generate subgroup analyses based on stroke subtypes. We lacked data on stroke incidence and pertinent clinical and socioeconomic variables; therefore, we could not examine the association of clinical or social determinants of health with stroke‐related mortality. Finally, vital statistics records deaths to the state of residence at the time of death and does not factor in migration across the US–Mexico border. For instance, according to Pew Hispanic Center estimates, a total of 3 million Mexican citizens immigrated to the United States between 1995 and 2000, followed by a decline to 1.4 million between 2005 and 2010.²⁹ Meanwhile, nearly 1.4 million Mexican citizens moved from the United States to Mexico between 2005 and 2010.

Moreover, recent changing patterns of border enforcement and characteristics of return migrants can further influence this area’s population composition.³⁰ Because migration is strongly implicated in the intertwined demographic and health transitions, these aspects of demographic transition influence socioeconomic indicators, comorbidity burden, and subsequent mortality.

In summary, in the US–Mexico border counties, the initial decline in stroke‐related mortality has stalled during the past decade. Our findings illustrate the demographic and state‐level disparities regarding stroke‐related mortality across the border and nonborder counties. The Million Hearts initiative targets to prevent 1 million heart attacks and strokes by 2022.³¹ Future policy efforts should advocate for integrating social determinants of health into existing cardiovascular care paradigms to identify vulnerable populations that could benefit from targeted interventions and mitigate the burden of stroke‐related mortality in the unique US–Mexico border region.

Sources of Funding

makeadent.org's Ram and Sanjita Kalra Aavishqaar Fund at Cleveland Clinic Akron General in Akron, Ohio funded the open access fee for this study.

Disclosures

Dr Kalra reports being Chief Executive Officer and Creative Director of makeadent.org. Dr Virani reports grant support from the Department of Veterans Affairs, World Heart Federation, and Tahir and Jooma Family and honorarium from the American College of Cardiology (Associate Editor for Innovations; acc.org) and is a steering committee member of the PALM (Patient and Provider Assessment of Lipid Management) registry at the Duke Clinical Research Institute (no financial remuneration). Dr Shapiro reports scientific advisory activities with Alexion, Amgen, Esperion, and Novartis. Dr. Kalra reports being Chief Executive Officer and Creative Director of makeadent.org. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S3

Figures S1–S3

(J Am Heart Assoc. 2021;10:e019993. DOI: 10.1161/JAHA.120.019993.)