Mortality, Ethnicity, and Urbanization Among Children Aged 1-4 Years on the US-Mexico Border

Abstract

Objectives:

Little is known about the mortality of children along the US-Mexico border. The objective of our study was to determine whether mortality rates among Hispanic children along the border (“border Hispanic children”) exceeded mortality rates among non-Hispanic white children along the border.

Methods:

We examined mortality rates from 2001-2015 for children aged 1-4 years in US-Mexico border counties and in the United States overall. We compared mortality rates among Hispanic and non-Hispanic white children by county urbanization level (large central, medium, and small metropolitan; micropolitan nonmetropolitan; and noncore nonmetropolitan).

Results:

During 2001-2015, 1811 children aged 1-4 years died in the border region. The mortality rate per 100 000 children among border Hispanic children (28.3; 95% confidence interval [CI], 26.8-29.9) exceeded the mortality rate of US Hispanic children (24.7; 95% CI, 24.3-25.1) and border non-Hispanic white children (23.2; 95% CI, 20.8-25.6). When stratified by county urbanization level, however, mortality rates of border Hispanic children were not significantly different from mortality rates of US Hispanic or border non-Hispanic white children. Mortality rates in noncore nonmetropolitan counties were twice those in large central metropolitan counties, with injury mortality accounting for most of the excess. Mortality rates increased in nonmetropolitan border counties after 2010.

Conclusions:

Increased risk for injury and disease in noncore nonmetropolitan counties might be related to poverty, reduced access to care, or poorer quality of care. Future research should identify the remediable risk factors in such communities as the next step in preventing deaths among children aged 1-4 years.

The US-Mexico border region is defined as the land within 100 km (62 miles) of the international border.¹ In practice, the US border region is synonymous with the 44 counties in the border states of Arizona, California, New Mexico, and Texas that are all or partly inside the 100-km limit. The population of approximately 7.9 million people residing within the region in 2015 was 56% Hispanic and 34% non-Hispanic white. Most Hispanic border residents are Spanish-speaking Mexican Americans.² The Hispanic border population is economically disadvantaged; 32% of all residents and 37% of children aged <18 live below the federal poverty level.²

To date, most research on health inequities in the border region has focused on reproductive health outcomes, perhaps in part because Hispanic women in this region have higher birth rates than those of US Hispanic and non-Hispanic white women.³ Compared with other US Hispanic women, Hispanic women on the border have higher rates of preterm birth, late entry into prenatal care, cesarean delivery, and adolescent birth.^3-6 Studies of infant mortality in the border region, conducted as part of Healthy People 2010 reviews, showed higher mortality in Arizona⁷ and lower mortality in California.⁸ Otherwise, little is known about the mortality of border children.

Healthy People 2020 added a focus area on early childhood that was not included in Healthy People 2010.⁹ In 2015, the US-Mexico Border Health Commission updated its own set of objectives for the border region, Healthy Border 2020.¹ Among other objectives, Healthy Border 2020 aims to reduce childhood mortality due to injury and congenital anomaly, both of which are prominent causes of death among children aged 1-4 years.¹⁰ Many deaths among children aged 1-4 years are preventable, but to our knowledge, no studies have examined recent surveillance data, the foundation for the design of prevention programs, on the border for this age group.

The objective of our study was to describe the mortality of children aged 1-4 years on the US-Mexico border, with particular attention to ethnic or geographic differences, such as the types of differences shown for reproductive outcomes. To accomplish this objective, we compared mortality rates of Hispanic children aged 1-4 years on the border with mortality rates of non-Hispanic white children on the border, US Hispanic children, and US non-Hispanic white children aged 1-4 years. We also examined whether county urbanization could explain any observed ethnic differences in mortality in the region. Finally, we examined trends in mortality by urbanization levels.

Methods

The target population was children aged 1-4 years residing in the 44 US counties that compose the US-Mexico border region (Figure 1). In 2015, the border population in this age group was 467 938,⁹ of whom 68.5% were Hispanic. The most populous county was San Diego, California, which accounted for 36.5% of the border population.¹⁰

Figure 1.

The 44 counties of the US-Mexico border region.

We used 2001-2015 mortality data from the US National Vital Statistics System (NVSS), accessed through the Centers for Disease Control and Prevention’s WONDER database.¹⁰ NVSS classifies deaths by the decedent’s place of residence rather than place of death. We compared 4 groups: border Hispanic, border non-Hispanic white, US Hispanic, and US non-Hispanic white children. Data on race/ethnicity in NVSS are derived from death certificates.

WONDER provides data on the urbanization of all US counties based on 6 urbanization levels: large central metropolitan counties in metropolitan statistical areas (MSAs) with populations of ≥1 million; large fringe metropolitan counties or suburbs of large central metropolitan counties; medium metropolitan counties in MSAs with populations of 250 000-999 999; small metropolitan counties in MSAs with populations of <250 000; micropolitan nonmetropolitan counties with urban core populations of 10 000-50 000; and noncore nonmetropolitan counties, mostly rural counties not in micropolitan statistical areas.¹¹ Of the 44 border counties, 1 county (San Diego) was classified as large central, none as fringe metropolitan, 6 as medium metropolitan, 4 as small metropolitan, 12 as micropolitan nonmetropolitan, and 21 as noncore nonmetropolitan counties. Because no border county was classified as fringe metropolitan, we analyzed border mortality by using 5 urbanization levels. We also examined mortality by state and by selected causes of death by urbanization level for the entire border population aged 1-4 years, because the number of deaths was too small to conduct this analysis by ethnicity. Finally, to measure progress toward national⁹ and regional¹ health objectives, we analyzed trends in border region mortality by urbanization level in 5-year time periods.

We tested for trends in rates by using the extended Mantel-Haenszel χ² test for linear trend, with P < .05 considered significant. We calculated rates and 95% confidence intervals (CIs) in the WONDER system.¹⁰ We considered rates based on <20 deaths to be unreliable. Rates based on <10 deaths were suppressed by WONDER. The institutions with which the authors are affiliated did not require human subjects review because the data are publicly available and do not include personal identifiers.

Results

During 2001-2015, 1811 border residents of all races/ethnicities aged 1-4 years died, of whom 1281 (70.7%) were Hispanic and 363 (20.0%) were non-Hispanic white. The mortality rate per 100 000 children among border Hispanic children (28.3) was 22.0% higher than that of border non-Hispanic white children (23.2) and 14.6% higher than that of US Hispanic children (24.7) (Table 1). In general, mortality rates increased with declining urbanization for all 4 groups (tests for trend, P < .001). The mortality rates in noncore nonmetropolitan counties among border Hispanic (42.8) and border non-Hispanic white (57.2) children were 33% and 58% higher, respectively, than those among US Hispanic (32.1) and US non-Hispanic white (36.3) children. Conversely, the mortality rates in large central metropolitan counties among US Hispanic (22.9) and US non-Hispanic white (22.0) children were 15% and 24% higher, respectively, than those among border Hispanic (19.9) and border non-Hispanic white (17.7) children. For all 5 urbanization levels, the mortality rate among border Hispanic children was not significantly higher than the mortality rate among border non-Hispanic white children.

Table 1.

Mortality rates among US children aged 1-4 years, by race/ethnicity and urbanization level,^a United States and US-Mexico border region, 2001-2015^b

Race/Ethnicity and Urbanization Level	No. of Deaths	Population, No.	Mortality Rate per 100 000 Children (95% Confidence Interval)
US-Mexico border region
Hispanic	1281	4 529 062	28.3 (26.8-29.9)
Large central metropolitan	217	1 088 213	19.9 (17.3-22.6)
Medium metropolitan	751	2 558 227	29.4 (27.3-31.5)
Small metropolitan	150	458 279	32.7 (27.5-38.0)
Micropolitan nonmetropolitan	121	326 272	37.1 (30.5-43.7)
Noncore nonmetropolitan	42	98 071	42.8 (30.9-57.9)
Non-Hispanic white	363	1 558 222	23.2 (20.8-25.6)
Large central metropolitan	163	922 920	17.7 (14.9-20.4)
Medium metropolitan	130	433 912	30.0 (24.8-35.1)
Small metropolitan	43	127 193	33.8 (24.5-45.5)
Micropolitan nonmetropolitan	15	53 215	28.2c (15.8-46.5)
Noncore nonmetropolitan	12	20 982	57.2c (29.6-99.9)
United States
Hispanic	14 099	56 996 130	24.7 (24.3-25.1)
Large central metropolitan	6090	26 615 524	22.9 (22.3-23.5)
Large fringe metropolitan	2353	10 473 415	22.5 (21.6-23.4)
Medium metropolitan	3285	12 113 020	27.1 (26.2-28.0)
Small metropolitan	953	3 476 365	27.4 (25.7-29.2)
Micropolitan nonmetropolitan	975	2 938 691	33.2 (31.1-35.3)
Noncore nonmetropolitan	443	1 379 115	32.1 (29.1-35.1)
Non-Hispanic white	33 869	131 101 321	25.8 (25.6-26.1)
Large central metropolitan	6102	27 700 303	22.0 (21.5-22.6)
Large fringe metropolitan	7497	36 163 698	20.7 (20.3-21.2)
Medium metropolitan	7623	28 491 141	26.8 (26.2-27.4)
Small metropolitan	4185	14 304 333	29.3 (28.4-30.1)
Micropolitan nonmetropolitan	4770	14 282 790	33.4 (32.4-34.3)
Noncore nonmetropolitan	3692	10 159 056	36.3 (35.2-37.5)

^a Urbanization levels: large central metropolitan counties in metropolitan statistical areas (MSAs) with populations of ≥1 million; large fringe metropolitan counties or suburbs of large central metropolitan counties; medium metropolitan counties in MSAs with populations of 250 000-999 999; small metropolitan counties in MSAs with populations of <250 000; micropolitan nonmetropolitan counties with urban core populations of 10 000-50 000; and noncore nonmetropolitan counties, mostly rural counties not in micropolitan statistical areas.¹¹ No large fringe metropolitan counties exist in the US-Mexico border region; as such, this category was omitted.

^b Data source: Centers for Disease Control and Prevention WONDER.¹⁰

^c Unreliable rate based on cell count <20.

Of note, 16.9% (217/1281) of border Hispanic children deaths occurred among residents of central metropolitan San Diego County, whereas 44.9% (163/363) of border non-Hispanic white children deaths and 43.2% (6090/14 099) of US Hispanic children deaths occurred among residents of central metropolitan counties. As a result, a larger proportion of deaths among border Hispanic children than among border non-Hispanic white children or US Hispanic children occurred in less urban counties, where mortality rates were generally higher. In contrast, a higher proportion of deaths that occurred in large metropolitan counties were among border non-Hispanic white children (44.9%) than among US non-Hispanic white children (18.0%).

By state, the degree of urbanization declined from west to east along the border. Both of California’s border counties, 3 of Arizona’s 4 border counties, 1 of New Mexico’s 6 border counties, and 5 of Texas’s 32 border counties were in the 3 metropolitan categories. Border mortality rates per 100 000 children were significantly lower in California (20.3) than in Arizona (34.0), New Mexico (31.2), or Texas (30.6). Each state’s mortality rate—California (23.3), Arizona (35.3), New Mexico (36.8), and Texas (31.3)—was significantly higher than its border mortality rate.

Mortality rates for selected causes of death showed an urban-rural gradient (Table 2). Congenital anomalies, the leading cause of death, were responsible for 254 of 1811 (14.0%) deaths among border children aged 1-4 years. The highest calculable mortality rate for congenital anomalies (4.9) was in micropolitan nonmetropolitan counties, but the test for trend was not significant. Neoplasms were the cause for 196 (10.8%) deaths. The highest rate for neoplasms (4.1) was in micropolitan nonmetropolitan counties, but none of the rates were significantly different from one another, and the test for trend was not significant. Transport accidents, including pedestrian, cyclist, and motor vehicle crashes, led to 253 (14.0%) deaths among border children aged 1-4 years. The transport accident mortality rate in noncore nonmetropolitan counties (11.6) was significantly higher than the mortality rates of large central and medium metropolitan counties, and the test for trend was also significant (P < .001), as were the trends in non–transport-related injury and all other noninjury causes of death (P < .001). Together, the transport and other injury categories accounted for 18.3/100 000 (68.5%) of the 26.7 per 100 000 population overall difference between large central metropolitan and noncore nonmetropolitan counties, with diseases accounting for the remaining 31.5% of the difference. By comparison, transport accidents and other injury categories accounted for only 40.2% of the 1811 deaths.

Table 2.

Mortality rates among US children aged 1-4 years, by urbanization level^a and cause of death,^b US-Mexico border region, 2001-2015^c

Urbanization Level	Mortality Rate per 100 000 Children (95% Confidence Interval)
	Congenital Anomaly	Neoplasm	Transport Accident	Non–Transport- Related Injury	All Other Causes of Death	Total
Large central metropolitan	3.3 (2.6-4.1)	2.9 (2.3-3.7)	1.6 (1.1-2.2)	4.9 (4.0-5.8)	6.8 (5.8-7.9)	19.6 (17.9-21.4)
Medium metropolitan	4.0 (3.3-4.7)	2.8 (2.2-3.4)	4.6 (3.8-5.3)	7.8 (6.8-8.7)	10.8 (9.7-12.0)	29.9 (28.0-31.9)
Small metropolitan	3.8 (2.4-5.6)	3.4 (2.1-5.2)	4.6 (3.0-6.6)	11.9 (9.4-15.0)	9.6 (7.3-12.4)	33.4 (28.8-37.9)
Micropolitan nonmetropolitan	4.9d (2.9-7.6)	4.1d (2.3-6.6)	7.4 (5.0-10.6)	7.9 (5.4-11.2)	12.3 (9.0-16.3)	36.5 (30.5-42.5)
Noncore nonmetropolitan	Suppressede	Suppressede	11.6d (6.3-19.4)	13.2d (7.6-21.5)	14.1d (8.2-22.5)	46.3 (35.0-60.2)
P value, test for trendf	.14	.30	<.001	<.001	<.001	<.001

^a Urbanization levels: large central metropolitan counties in metropolitan statistical areas (MSAs) with populations of ≥1 million; medium metropolitan counties in MSAs with populations of 250 000-999 999; small metropolitan counties in MSAs with populations of <250 000; micropolitan nonmetropolitan counties with urban core populations of 10 000-50 000; and noncore nonmetropolitan counties, mostly rural counties not in micropolitan statistical areas.¹¹ No large fringe metropolitan counties exist in the US-Mexico border region, so this category was omitted.

^b Causes of death were defined in the Centers for Disease Control and Prevention (CDC) WONDER database according to the following International Classification of Diseases, Tenth Revision codes: congenital anomaly (Q01-Q99), neoplasm (C00-D48), transport accident (V01-V99), and non–transport-related injury (W00-Y89).¹²

^c Data source: CDC WONDER.¹⁰

^d Unreliable rate based on cell count <20.

^e Suppressed rate based on cell count <10.

^f Extended Mantel-Haenszel χ² test for linear trend, with P < .05 considered significant.

We found a downward time trend in mortality in the border region for metropolitan counties (Figure 2) but not for counties in the 2 nonmetropolitan categories, micropolitan and noncore. As a result, the urban-rural disparity grew over time: the noncore nonmetropolitan mortality rate was 2.3 times the large central metropolitan rate (50.0 vs 22.0) during 2001-2005 and 2.8 times the large central metropolitan rate during 2011-2015 (47.0 vs 16.6).

Figure 2.

Mortality rates among US children aged 1-4 years, by urbanization level and time period, US-Mexico border region, 2001-2015. Urbanization levels: large central metropolitan counties in metropolitan statistical areas (MSAs) with populations of ≥1 million; medium metropolitan counties in MSAs with populations of 250 000-999 999; small metropolitan counties in MSAs with populations of <250 000; micropolitan nonmetropolitan counties with urban core populations of 10 000-50 000; and noncore nonmetropolitan counties, mostly rural counties not in micropolitan statistical areas.¹¹ Extended Mantel-Haenszel χ² test for linear trend, with P < .05 considered significant. P values for test for trend: large central metropolitan (P = .01), medium metropolitan (P = .01), small metropolitan (P = .21), micropolitan nonmetropolitan (P = .45), and noncore nonmetropolitan (P = .78). Includes the 44 border counties in Arizona, California, New Mexico, and Texas. Data source: Centers for Disease Control and Prevention WONDER.¹⁰

Discussion

What at first appeared to be an ethnic mortality disparity, with higher mortality rates among Hispanic children than among non-Hispanic white children, was in fact a wide disparity between urban and rural counties—one that is growing over time.

The border mortality rate among children aged 1-4 years is an anomaly in that overall mortality rates in the US Hispanic population were about 25% lower than mortality rates in the US non-Hispanic white population in 2013.¹³ This phenomenon is termed the Hispanic paradox¹⁴ because socially disadvantaged populations in the United States typically have higher mortality rates than socially advantaged populations. The paradox is thought to be largely the result of lower smoking rates among Hispanic people than among non-Hispanic white people.¹⁵ Because smoking is not a factor that applies to children aged 1-4 years, the Hispanic population mortality advantage would not likely figure into a comparison between non-Hispanic white and Hispanic children in our study.

The relationship between levels of urbanization and risk of death among children in our study has also been found among adults in the United States,¹⁶ and the association we found between nonmetropolitan counties and potentially excess mortality has also been found elsewhere.¹⁷ In general, fringe metropolitan counties have the lowest mortality rates, and the most urban or most rural counties have the highest mortality rates, depending on age group and region.¹⁶ For people aged 1-24, the highest mortality rates are in the most rural counties for most US regions. The mortality rates for unintentional injuries, including motor vehicle–related deaths, are consistently highest in the most rural counties,¹⁶ and cancer mortality rates are highest in nonmetropolitan counties.¹⁷ Higher mortality rates in nonmetropolitan areas have been attributed to poverty, less access to emergency and specialty care facilities, and limited emergency medical service capability in nonmetropolitan communities.¹⁷

Border Hispanic people are more likely than non-Hispanic white people to live in poverty. Spanish-speaking Hispanic children residing in small rural communities were more likely than urban children to live below the federal poverty level in 2010.¹⁸ The median household income in the US Hispanic population is roughly two-thirds that of the US non-Hispanic white population,¹⁹ and the border Hispanic population has lower household incomes than the US Hispanic population.² However, the rurality-mortality gradient persists even after adjusting for poverty level on the national level.²⁰ Therefore, other explanations for these findings should be considered.

Health care might be another key factor to explain the mortality pattern in this study. Metropolitan counties tend to have more health care providers per capita than nonmetropolitan counties. Residents of suburban areas tend to have higher-quality health care and better outcomes than residents in rural areas. Rural residents often live farther away from health care resources than residents of suburban areas. In the event of trauma or acute serious illness, residents of rural counties are more likely to be seen in standard emergency departments than in the specially equipped trauma centers that are more prevalent in urban areas.²¹

The relationship between transport accident mortality and urbanization was likely related to multiple factors. Both environmental and behavioral factors have been suggested as explanations for the higher motor vehicle crash mortality rates among rural adults than among urban adults²² and the higher incidence of motor vehicle crash injury among rural children than among urban children.²³ A higher rural mortality rate among children in our study might be a function of both the incidence of injury and case-fatality rates. Factors that might contribute to a higher rural case-fatality rate include lower overall rates of seat belt use in rural areas²² and lack of emergency medical services and trauma center services. A lack of emergency medical services might be the more important factor because excess traffic deaths in rural counties are the result of deaths occurring before the arrival of emergency medical services rather than deaths occurring in hospitals.²⁴

Interpreting the higher rural mortality rates for all other noninjury causes of death is difficult because of the small number of deaths in some disease categories. The mortality rates in the most rural counties for the largest disease category, congenital anomalies, were suppressed, and the numbers were insufficient to confirm a trend toward higher mortality rates in more rural counties. Overall, however, surveillance systems²⁵ and the published literature provide little empirical evidence to support the idea that the rate of defects that are likely to cause death in children aged 1-4 years is higher in rural counties than in urban counties. Many serious central nervous system, chromosomal, and cardiac anomalies that might cause death after infancy can be detected through prenatal screening; yet some evidence suggests that pregnant women in rural counties are less likely to be screened than pregnant women in urban counties and, therefore, are less likely to abort affected fetuses.²⁶ If this finding is true on the border, the net effect would be a higher congenital anomaly mortality rate in infancy and childhood in rural counties than in urban counties.

Mortality rates in nonmetropolitan counties among children aged 1-4 years have declined in the United States since 2010 for both Hispanic and non-Hispanic white people¹⁰; as such, the stagnant or increasing mortality rates in nonmetropolitan counties of the border region might reflect local problems related to poverty and lack of medical services. Infant mortality fell during 2014-2016 in states opting for Medicaid expansion after the Affordable Care Act,²⁷ but children aged 1-4 years in rural border counties largely missed out on any possible benefit from such a change because Texas, home to 68% of children in nonmetropolitan counties and 39% of children in metropolitan counties, did not opt to expand Medicaid.²⁸ The persistently high mortality rates in rural border counties provide additional impetus for reducing income disparities and improving health care to improve the health of children in these counties.

Limitations

This study had several limitations. First, small numbers of deaths in border subpopulations limited our ability to study less frequent causes of death and trends for cause-specific mortality. Second, mortality rates might be overestimated if undocumented Hispanic border residents were recorded as US residents on death certificates but were missed in censuses, thereby producing an underestimate of the border population. Undocumented immigrants represent a larger proportion of the population in border states than elsewhere in the United States.²⁹ Finally, border rates might be underestimated if deaths of border residents that occurred in Mexico or elsewhere outside the United States were never reported to US vital statistics registrars or if decedents were counted as residents in the censuses but not recorded as US residents on the death certificate. Misclassification of Hispanic people as non-Hispanic people on US death certificates occurs in roughly 3% of Hispanic deaths and, therefore, was unlikely to be an important factor in this analysis.³⁰

Conclusions

More information about economic drivers, policies, access to care, and other contextual factors is needed to better understand the high early childhood mortality rates in rural border counties. County-specific analyses of law enforcement and disease registry data could help determine whether disparities in incidence and mortality for injuries and disease exist in the border region. Lack of use of restraints (eg, car seats) and other factors associated with motor vehicle fatalities (eg, type of crash) can be examined by county in the Fatality Analysis Reporting System.³¹ In regions with child death review committees (https://www.childdeathreview.org), data by county should be examined.

Future analyses should examine the health of the border population stratified by county characteristics given the heterogeneity of outcomes across the region. Future analyses and interventions should include both the Hispanic and non-Hispanic border populations and both rural and urban counties. Future research should also identify the important, remediable risk factors in these communities as the next step in preventing deaths among children aged 1-4 years.