Refined, regionally-specific data standards reveal heterogeneity in Hispanic death records

Abstract

Hispanic ethnicity can be captured with differing levels of granularity using various data standards, including those from the Office of Management and Budget, Health and Human Services and National Academy of Medicine. Previous research identified seven subgroups of Hispanics in New Mexico using open-ended interviews and information about the culture/history of the state. We examined age and manner of death to determine whether differences among subgroups are hidden by less-refined categorization. Significant differences in the mean age at death were found between some groups, including Spanish and Mexican Americans. We found an association between specific manners of death codes and subgroups. However, significance disappeared when manners of death were grouped (e.g. accident, homicide, etc.). This indicates that while certain manners of death are associated with group membership, overall types of death are not. Data descriptors for Hispanics should reflect more refined, regionally relevant groups, in order to unmask heterogeneity.

Introduction

Race and ethnicity are common data elements used in the Census, healthcare, research, and education. However, how race and ethnicity are classified can differ depending on the standard implemented. The Office of Management and Budget (OMB), after extensive testing and public engagement, concluded that Hispanic ethnicity should be captured as "Hispanic or Latino" and "Not Hispanic or Latino."¹ A great deal of research on health disparities in Hispanics and Latinos/as continues to use these overarching categories.^2,3 There also has been considerable research on health and mortality in subgroups of Hispanics.^4-8 Hummer et al.⁷ found differences in the age at death among Mexican Americans, Cuban Americans, Central and South American and other Hispanics. They also discovered that mortality differed between the subgroups using age, sex and cause of death. Weinick et al.⁴ found differences in emergency room visits, inpatient visits and prescription usage among Mexicans, Cubans and Puerto Ricans. Fenelon et al.⁵ found mortality differences between 12 subgroupings of Hispanics using region of origin and nativity. Additional studies have found differences in fertility, birth rates, life expectancies and morbidities.^6,8 These examples point to the importance of considering the heterogeneous nature of Hispanic subgroups for understanding health and mortality disparities.

Health and Human Services (HHS) has suggested the use of categories that can be "rolled up" into the OMB standard when needed. These groups are "Cuban," "Mexican and Mexican American," "Puerto Rican," and "Another Hispanic group."⁹ For the 2010 and 2020 censuses the groupings were "Mexican, Mexican Am., Chicano," "Puerto Rican," "Cuban,", and "another Hispanic, Latino, or Spanish origin."¹⁰ These groups have specific cultural and historical significance in some states, such as New Mexico, Florida and New York, but not in others. In many cases, researchers collect data on these relatively refined categories, but then use the "rolled up" groupings as the sample sizes for refined groups can be small, reducing the likelihood of achieving statistical significance in analyses.

The National Academy of Medicine (NAM) report of 2009 suggested using 36 subgroups of the group, "Hispanic," with a greater range of relevance in various regions of the country, in conjunction with the overarching OMB standard¹¹ (see Table 1). Here, we consider the Hispanic population in New Mexico, using locally relevant ethnic subgroup categories suggested by previous research.¹² We used a list of seven subgroups within the New Mexican Hispanic population: "Chicano/a," "Hispanic," "Latino/a," "Mexican," "Mexican American," "Nuevomexicano/a," and "Spanish."¹² Unlike the HHS or OMB standards, these categories are specific to the culture and history of New Mexico. They also match the NAM categories with the addition of "Nuevomexicano/a" and "Hispanic." Hispanic is added to the list of subgroups because it is the most common term chosen by New Mexicans. Note that "Hispanic" describes both a subgroup and a catch-all term for all subgroups combined. Nuevomexicano/a is a regionally and temporally specific term used in New Mexico; individuals who identify using this term have similar migration histories to individuals who identify as Spanish.^12,13

Table 1. National Academy of Medicine subgroups of Hispanic Ethnicity.

Andalusian	Catalonian	Ecuadorian	Mexican American	Salvadoran
Argentinean	Central American Indian	Gallego	Mexicano	South American
Asturian	Chicano	Guatemalan	Nicaraguan	South American Indian
Balearic Islander	Colombian	Honduran	Panamanian	Spaniard
Bolivian	Costa Rican	La Raza	Paraguayan	Spanish Basque
Canal Zone	Criollo	Latin American	Peruvian	Uruguayan
Castilian	Cuban	Mexican	Puerto Rican	Valencian
				Venezuelan

Understanding variation among subgroups such as these may be useful in unraveling the "Hispanic Paradox," in which, in the US at large, Hispanics have better health outcomes than non-Hispanic Whites (NHW,) despite lower socioeconomic standing.^14-19

Background

The New Mexico Decedent Image Database (NMDID) was created in 2020, and houses 15,243 full-body computed tomography (CT) scans and associated health and lifestyle data.²⁰ Contained within this database are the fields of race, ethnicity, and Hispanic subgroup. Next of kin were contacted to collect data not available in the medical examiner's database, including Hispanic subgroup. "Hispanic" is a variable in the ethnicity, race, and subgroup fields, reflecting responses from next of kin.

NMDID also contains information regarding age at death and manner of death. Age at death is the actual age each decedent was when they died. It provides a proxy for life expectancy, which can be used to compare among groups. Manner of death is one of five broad categories: accident, homicide, natural, suicide, and unknown. Manner of death code provides more specific information within one of these broad categories, such as pneumonia (natural) or automobile accident (accident). The decedents represented in NMDID with manner of death information (n=15,236) include 38.6% accidents, 34.6% natural, 7.4% homicides, 15.4% suicides, and 4% undetermined. Overall, 11% of deaths in New Mexico from mid-2010 to mid-2017 are included in the database. Excluded from the sample are most expected deaths, such as elderly persons in nursing homes, and physician-attended deaths. Thus, the analyses and results presented here are biased such that they underrepresent the very oldest in the living population, and over represent younger individuals.

Using this information, we have three aims: 1) determine whether there is a significant difference in age at death among the seven subgroups of Hispanics, 2) determine whether there is an association between manner of death and each Hispanic subgroup, and 3) determine whether there are significant associations between manner of death codes and Hispanic subgroups within the NMDID sample.

Materials and Methods

Only de-identified data from decedents was used to create NMDID, so no Institutional Review Board approval was required for this research. Approval from the research committee of the New Mexico Office of the Medical Investigator, where the data originate, was obtained in 2015.

NMDID was queried to find all decedents with a Hispanic subgroup identified. Hispanic subgroup identification was provided by the next of kin when contacted by NMDID staff for additional information. If next of kin identified the decedent as Hispanic in either race or ethnicity, they were asked how the decedent would have identified their own subgroup. Subgroup names were provided to next of kin, and included: Chicano/a, Hispanic, Latino/a, Mexican, Mexican American, Nuevomexicano/a, and Spanish. There are 639 decedents of Hispanic ethnicity or race that also have Hispanic subgroup information, comprising 13% of Hispanics in the database. The subgroups and counts are available in Table 2. In addition, data were queried for age at death in years, manner of death, and manner of death code.

Table 2. Distribution of subgroups in the New Mexico Decedent Image Database.

Hispanic Subgroup	Count
Chicano/a	99
Hispanic	196
Latino/a	46
Mexican	71
Mexican American	76
Nuevomexicano/a	28
Spanish	123
Total	639

The mean ages at death in years were compared among subgroups using ANOVA. We then ran a post hoc Tukey analysis to determine which groups were significantly different from one another. In addition, mean age between NHW and the subgroups were compared using the same methods.

Each decedent was assigned both a category manner of death (Accident, Homicide, Natural, Suicide, or Unknown) and one of 60 specific manner of death codes (such as pneumonia or automobile accident) by a forensic pathologist. We next determined whether there are associations between manner of death, both category and specific code, and subgroup membership. To account for the large sample size and high number of specific manner of death codes, we performed Monte Carlo simulations of the Fishers exact test with N=100,000 iterations. All analyses were performed in SAS Studio.²¹

Results

Age at death

The mean age in years for all 639 decedents is 45.45 years (SD:18.91), with range 0-99 years (with 12 decedents less than 10 years of age). The mean and standard deviation of age at death in years for the seven subgroups are listed in Table 3. For comparison, among the entire New Mexican population, mean age of death in 2017 was 81.2 years for women and 75.3 for men, demonstrating a bias toward younger age of death for this sample.²² Within NMDID, the NHW (n=8,767) average age at death is 52.69 years (with 232 decedents under the age of ten).

Table 3. Mean age at death in years, with standard deviations, for the seven Hispanic subgroups in New Mexico.

Hispanic subgroup	Mean age in years	Standard deviation
Chicano/a	48.52	16.55
Hispanic	43.29	19.20
Latino/a	42.91	17.46
Mexican	43.05	20.33
Mexican American	41.58	21.00
Nuevomexicano/a	48.90	15.69
Spanish	50.34	18.18

A significant difference was found in the mean age at death in years among the seven Hispanic subgroups (p=0.0029). See Figure 1A for the distribution of ages at death among the subgroups. There are significant differences in the Spanish-Hispanic (p=0.019) and Spanish-Mexican American comparisons (p= 0.0236), with Spanish dying seven years later than Hispanics and eight years later than Mexican Americans. When the seven subgroups were compared to NHW in NMDID, a significant difference was found (p<0.0001) in NHW-Hispanic (p<0.0001), NHW-Latino/a (p=0.0143), NHW-Mexican (p=0.0008), and NHW-Mexican American (p<0.0001) comparisons. Only Chicano/a, Nuevomexicano/a, and Spanish did not differ significantly from NHW.

Figure 1.

A: Boxplots of age at death between the 7 Hispanic subgroupings in New Mexico. B: Boxplot of age at death between Hispanic subgroups when similar groups are combined.

Hunley et al.,¹² describes similarities of history and migration among some of these subgroups. Following these descriptions and to further explore the importance of subgroup specificity, Mexican American and Mexican subgroups were combined, as were Nuevomexicano/a and Spanish. Differences in mean age at death are even more clear when these subgroups are considered together (p=0.0006). Significant differences are found between Spanish/Nuevomexicano/a and Mexican/Mexican American (p= 0.0028) and Spanish/Nuevomexicano/a and Hispanic (p= 0.0065). See Figure 1B for the distribution of age at death when similar subgroups are combined.

Manner of Death

There are fewer homicides among Spanish, and fewer suicides among Chicano/as than expected. However, there is no significant relationship between subgroup and manner of death, when all subgroups are considered independently (p=0.3465) or related subgroups are combined (p=0.2223). See Table 5 for the contingency table of manner of death in all groups.

Table 5. Manner of death groupings by Hispanic subgroupings in New Mexico. Frequency of each cell is above, with expected count below.

Manner of Death	Hispanic subgroup
Frequency (n) Expected	Chicano /a	Hispanic	Latino /a	Mexican	Mexican American	Nuevomexicano /a	Spanish	Total
Accident	54	83	17	28	33	13	54	282
43.69	86.498	20.3	31.333	33.54	12.357	54.282
Homicide	13	21	5	11	9	1	6	66
10.225	20.244	4.7512	7.3333	7.8498	2.892	12.704
Natural	23	61	15	20	21	10	42	192
29.746	58.892	13.822	21.333	22.836	8.4131	36.958
Suicide	6	25	6	11	8	2	18	76
11.775	23.311	5.471	8.4444	9.0391	3.3302	14.629
Undetermined	3	6	3	1	5	2	3	23
3.5634	7.0548	1.6557	2.5556	2.7355	1.0078	4.4272
Total	99	196	46	71	76	28	123	639

Manner of death code

There are 60 manner of death codes used in this sample, ranging from natural, various automobile accidents, drug and alcohol use, to falls (see Table 4 for list of specific codes and total counts in sample). A significant association was found between cause of death code and subgroup (p=0.0364). In contingency table (not shown), more Chicano/as died of drug use, more Spanish and fewer Mexicans died of prescription pill overdose, more Spanish died of exposure, and fewer Spanish died of gun shots than expected. When similar subgroups are combined, the significance increases (p=0.0123).

Table 4. Specific Manner of Death Codes in Sample.

Manner code	Count
ASCVD/COPD/Seizure disorder	1
Accident-specify	3
Asphyxia/airway obstruction/suffocation	5
Beaten by assailant(s)	9
Bitten/mauled/stung/kicked by (bee, dog, snake, horse, (name agent))	1
Blunt trauma/multiple injuries/subdural hematoma	1
Choked on (bolus of food, toy, etc.)	2
Contacted electrical current via (outlet, telephone lines, ungrounded chain saw)	1
Crushed/suffocated by (car falling from jack, plastic bag over head)	6
Cyclist accident	3
Driver of auto in collision with (auto, pickup, truck, minivan, ATV, motorcycle, (other motor vehicle type))	10
Driver of auto that left roadway (and overturned and/or became pinned underneath or in auto)	3
Driver of motorcycle (explain circumstances briefly, e.g. left roadway and struck tree, overturned in roadway, etc.)	5
Driver of motorcycle in collision with (auto, pickup, truck, minivan, ATV, motorcycle, (other motor vehicle type))	6
Driver of pickup that left roadway (and overturned and/or became pinned underneath or in auto)	5
Driver of truck that left roadway (and overturned and/or became pinned underneath or in truck)	1
Drowned in (tub, arroyo, pool, (this includes all non-recreational water accidents))	5
Drowned while swimming (this includes recreational swimming and rescue attempts)	2
Fall from (chair, table, mesa, cliff)	8
Fall from height/same height	1
Fall from standing height	8
Gunshot wound	2
Hanged self	28
Homicide-specify	3
Ingested alcohol (ethanol)	9
Ingested and/or injected illicit drug(s) - (in combination with ethanol)	94
Ingested and/or injected prescription medications	54
Ingested or injected medication	5
Ingested, injected or inhaled non-prescription medication (illicit, volatiles)	4
Inhaled toxic substance (toxic substances inhaled accidentally)	3
Inhaled toxic substance - (toxic substances abused to achieve intoxication)	2
Jumped from	1
Natural	192
Neglect/Starvation	1
Passenger .in (airplane, balloon, hang glider, (other aircraft type) that crashed) - (Also parachutist)	2
Passenger in auto in collision with (auto, pickup, truck, minivan, ATV, motorcycle, (other motor vehicle type))	7
Passenger in auto in collision with (tree, embankment, rock, wall, (other fixed object))	1
Passenger in auto that left roadway (and overturned and/or became pinned underneath or in auto)	6
Passenger in pickup in. collision with (auto, pickup, truck, minivan, ATV, motorcycle, (other motor vehicle type))	1
Passenger on motorcycle in collision with (auto, pickup, truck, minivan, ATV, motorcycle, (other motor vehicle type)	1
Pedestrian homicide (ie, Struck with auto by assailant(s)	1
Pedestrian struck by (auto, pickup, truck, train, (other motor vehicle type))	17
Pedestrian struck by (bicycle, (other non-motor vehicle))	2
Pedestrian struck by Motor vehicle	2
Pilot of (airplane, balloon, hang glider, (other aircraft type) that crashed)	1
Received blow/collided with	1
Remained outdoors exposed to (cold, heat) - ((while intoxicated))	6
Shot by assailant(s) with firearm	42
Shot self with firearm	34
Skeletal, mummified or decomposed remains	4
Slashed with	1
Stabbed by assailant(s)	7
Strangled by assailant(s)	3
Suffocated self with	1
Suicide-specify	1
Suicide as pedestrian	1
Undetermined after autopsy and/or toxicology	4
Undetermined-specify	3
Victim of (car bomb, letter bomb, (type of device)) explosion	1
Victim of (house, car, trailer, open range, (other site)) fire	5

Discussion

Hispanic is a panethnic term that is used to describe individuals with origins from the Caribbean, Spain, South and Central America, etc. As such, it includes individuals with a wide variety of geographical, economic and social backgrounds.^14,23,24 Hispanics are often grouped together for biomedical research, but do not represent a homogenous ethnicity.¹⁴ In this study we used age at death, manner of death, and manner of death code to determine whether there are differences among culturally and historically relevant subgroups in New Mexican Hispanics. Previous studies have shown differences in health, mortality, fertility and birth rates using Hispanic subgroupings with vocabulary standards defined by the HHS or a select subset.^5-8

Results show a significant difference in age at death between the Spanish and Mexican American and Spanish and Hispanic subgroups, with the Spanish dying seven to eight years later than these comparison groups, respectively. These differences are even more notable when culturally and historically similar groups are combined. These results indicate the value of understanding of the culture and history from which ethnic terminology standards develops. For example, individuals may use the term Nuevomexicano/a to indicate their long history in the state and their Spanish decent.²⁵ Given this background, it is not surprising that differences in age at death are made clearer when Spanish and Nuevomexicano/a groups are combined. This is because "Spanish" as a self-identifier is used by older individuals and "Nuevomexicano/a" by younger people with similar backgrounds.¹³ It appears that, while ethnic terminologies change over time, the challenges affecting age at death remain consistent.

The age at death of NHW drawn from the same medicolegal sample is significantly higher than that of the seven Hispanic subgroups, with Hispanics, Latino/as, Mexicans, and Mexican Americans all dying younger than NHWs. Spanish and Nuevomexicano/a, groups that are traditionally considered privileged in the state, do not die significantly younger than do NHW.^25,26 These results do not lend support for the Hispanic Health Paradox, as Mexicans and Mexican Americans, who have been shown to be more likely born outside of New Mexico or descended from more recent immigrants, are dying at a younger average age than NHWs.¹³

While there are no significant associations between manners of death and subgroups there is an association between specific manner of death codes and subgroups. More Chicano/as die of drug use, more Spanish and fewer Mexicans die of prescription pill overdose, more Spanish die of exposure, and fewer Spanish die of gun shots than expected if manner of death codes were unrelated to subgroup membership. Recognition of this heterogeneity can be used for targeted interventions, related to licit and illicit drug use, alcohol use, and gun safety.

These results make clear that OMB and HHS terminology standards are not specific nor culturally relevant enough for research into health and mortality research, at least not among Hispanics in New Mexico. The NAM standard, although not used widely, identifies and provides terminology for more specific regional variation. If the OMB or HHS standards had been used in this analysis, we would have missed heterogeneity important for understanding public health, especially between the combined Spanish and Nuevomexicano/a versus Mexican and Mexican American groups. These results indicate that researchers should strive to capture more regionally relevant ethnicity data on subjects. This will likely require large samples and the introduction of ethnological methods into the development of ethnic terminologies and vocabulary standards.

Conclusion

There are significant differences in ages at death between Spanish and Mexican American subgroups of Hispanics in New Mexico. In addition, specific manner of death codes are associated with seven ethnic subgroups. Without terminology to reflect the variation in ethnicity collected in this database, important mortality data is lost. The subgroup terms used in this study come from an understanding of the culture and history of place, and reflect different migration histories among the different subgroups.¹³ The results indicate that it is of utmost importance to define the subgroupings of Hispanic beyond that suggested by HHS and OMB, when studying health and mortality. These results support the contention made by several other authors that important data regarding health patterns can be lost if Hispanics are treated as a homogeneous group.^5-8

The results also argue against the Hispanic Paradox, as groups with more recent migration histories (Mexicans and Mexican Americans) die at younger ages than those with long histories in the United States (Spanish and Nuevomexicano/a). If subgroups are used when studying Hispanic health, it may elucidate different outcomes than has been captured with a panethic term, clarifying the "paradox."

How can culturally relevant, specific group differences be captured across the United States? Some vocabulary standards, like HHS, include the most common the subgroups within major regions of the United States. The NAM report suggested even further specificity within Hispanic ethnicity, naming 36 different groups, all with the common thread of having a Spanish-speaking background. The current study included additional subgroups specific to New Mexico and not included in the NAM report. The fact that results were strengthened when some groups were combined indicates the need for continued reflexive analysis, as terminologies are not natural features of populations, but rather ephemeral categories relevant only at particular times, in particular geographies.

Although we recommend using the NAM as a core terminology standard for ethnicity, it is missing a vital subset that many Hispanics in New Mexico use to describe themselves: Hispanic. In addition, other regional terminologies may become apparent when ethnological research is conducted in different states and regions. Therefore, we recommend using the NAM ethnicity standard with the addition of the subgroup of Hispanic, perhaps with additional, region-specific terms.