Epidemiology of head injury and associations with clinical and neuropsychological test scores in older American Indians: data from the Strong Heart Study

Abstract

BACKGROUND

American Indians have highest mortality and hospitalizations from head injury of all US groups, however little is known about prevalence, risk, or outcomes in this population.

METHODS

The Strong Heart Study recruited American Indians representing 11 tribes and communities across 3 regions for two sequential examinations in 2010-2019. Participants were asked to self-report prior head injury, loss of consciousness (LOC), cause, sociodemographics and behaviors (age, sex, education, bilingual, smoking, alcohol use, stroke). Cognitive testing covered executive function, phonemic fluency, processing speed, and memory. Analyses tabulated summaries and multivariate logistic regressions estimated risk associations.

RESULTS

This older cohort of American Indians (Visit 1 N=818, follow-up Visit 2 N=403) was mean age 73 at intake, with mean 6.7 years between exams. At Visit 1, 40% reported prior head injury, majority with LOC; 4-6% reported injury with LOC>20 minutes. Incidence analysis estimated 3.5 cases per 100 person-years. Primary causes were falls, motor vehicles, sports, fight or assault, military (bullet, blast, fragment), and horse-riding incidents. Male sex and prior stroke were independently associated with higher risk, but age, education, bilingual, smoking, and alcohol use were not associated with risk. Those with previous head injury had significantly worse depressive symptoms, quality of life, fatigue, social functioning, pain, general health, and processing speed.

CONCLUSION

These findings suggest very high prevalence, incidence, and risk of head injury in older American Indians, with substantial impacts on quality of life and well-being. Future research should prospectively evaluate risk and prevention opportunities in this population.

INTRODUCTION

Head injuries cause concussion with or without loss of consciousness, directly contribute to functional impairments for 3.2 to 5.3 million U.S. patients. and are a major cause of death and disability.^1-3 American Indians have the highest head injury-related mortality^4,5 and the highest incidence of head-injury-related hospitalizations of all US racial and ethnic groups.⁶ American Indians are also the most likely to sign up for military service, a major risk factor for head injury and traumatic brain injury (TBI),^7,8 carrying a 2-year TBI-related mortality risk that is 20-50% higher among minorities than among non-Hispanic Whites.⁹ Although a large portion of head injury and TBI cases among American Indians is attributed to military service,¹⁰ other substantial risk factors may include motor vehicle accidents, falls such as from horse riding, and domestic incidents, such as fights or assaults.¹¹ However, little is known about differences across racialized minorities in neuropsychological response to head injury and TBI.⁵ Furthermore, because much of the existing data are based on surveillance-based reporting, which can be subject to reporting bias, existing epidemiologic and risk estimates may present an underestimate of true risk.

Given the observed low, population-wide test performance across multiple cognitive domains in American Indian elders,^12,13 identification and prioritization of modifiable risk features contributing to dementia disparities for this population is of critical importance. The Strong Heart Study (SHS), a large population-based cohort representing multiple communities across 3 regions, has established that older American Indians have disproportionate disparities in a majority of key, modifiable dementia risk features identified by the Lancet commission.¹⁴ The 13 life-course features identified in this model include: early life education; mid-life hearing loss, traumatic brain injury, hypertension, heavy alcohol use, and obesity; and late-life smoking, depression, isolation, inactivity, air pollution, and diabetes. In the SHS cohort, findings related to the 13 features from this model suggest a very high modifiable risk of dementia,^13,15-24 although population-based data on hearing loss, air pollution, and traumatic brain injury in American Indians are still missing.

Therefore, this work aims to: establish the self-reported prevalence and incidence of prior head injury in this population of older American Indians; and identify associations with important sociodemographic factors and neuropsychological test scores. This work builds on prior analyses, to establish models of modifiable features related to brain aging and dementia and thus to develop public health programs for primary prevention and risk reduction in this underserved population.

METHODS

Setting.

The parent Strong Heart Study (SHS) recruited 4,549 American Indians residing in 13 communities in the U.S. Northern Plains, Southern Plains, or Southwest, for a baseline examination focused on cardiovascular disease in 1989 to 1991.²⁵ The SHS originally defined eligible participation based on self-identification as American Indian, or individuals who claim ancestry with any of the original peoples of North, Central, and South America, and who maintain affiliation or attachment with any of the 13 partnering tribes and communities. Of note, the term “American Indian” is a contemporary U.S. legal term that was used in the tribal agreements for this study cohort.

From 2010 to 2019, two follow-up examinations collected data on brain aging, with a particular focus on vascular and degenerative features, among 11 of the 13 original SHS Tribes and communities. In 2010-2013, 818 participants were evaluated in Visit 1, representing 86% of original cohort survivors, then aged 64-95 years.²⁶ In 2017-2019, 403 participants were reevaluated in Visit 2, representing 78% of survivors, then aged 70-95 years.¹³ All Institutional Review Boards, Indian Health Service, Research Review Boards, and Tribal Councils reviewed and approved this research, as described previously ^13,26, and all participants signed informed consent. All research manuscripts are submitted to field center tribes and IHS for community approval.

Data collection.

Study procedures for both examination visits have been previously reported in detail.^26,27 In brief, participants completed interview and questionnaire, underwent clinical and neuropsychological assessment, contributed blood and urine samples, and submitted to 1.5T cranial MRI. Self-reported features included age (years), sex (male/female), years of formal education, ability to speak Native language—an indicator of bilingual status in this English-fluent cohort, lifetime smoking (>100 cigarettes), recent alcohol use (any within past year), and prior stroke (based on medical diagnosis).²⁸ Standardized questionnaires and neuropsychological instruments were used to collect data on depressive symptoms, health related quality of life, and multiple cognitive domains, including processing speed, phonemic and semantic fluency, memory, visuospatial skills, executive function, attention, and general or multidomain cognition with screening for impairment. Previous reports have covered psychometric validation of several of these instruments, with work underway for the remaining.^29-33 Supplemental Table 1 provides a full list of instruments collected.

Data availability.

Strong Heart Study data are available for analysis by interested investigators: with SHS permission, per sovereign tribal agreements, as previously reported,^13,26 and as outlined on the SHS website (https://strongheartstudy.org).

Head injury.

Information on prior head injury was also self-reported, based on participants answers to the following questions: (A) “Have you ever had an injury that resulted in loss of consciousness? (Knocked out)” with number of times, age at first occurrence, and age at last occurrence; (B) “Did you have a head injury from any of the following?” with given options including fragment, bullet, blast (rocket, grenade, land mine, explosion), motor vehicle crash, air or water transportation, fall, physical training or sports, assault, horse riding, and other with option to write-in additional causes; and (C) “Did any (head) injury result in any of the following?” with options dazed or seeing stars, symptoms of concussion, loss of memory, LOC <1 minute; LOC 1-20 minutes; LOC >20 minutes”. ³⁴Of note, endorsements to question B and C were not mutually exclusive, and participants were asked to endorse any or all that apply. If participant had positive responses to B, or C, they were defined as having had a prior head injury. Those who responded yes only to question A and not to question B or C were not included in the primary case definition because the specific wording did not preclude other types of (non-head) injuries leading to loss of consciousness. However, sensitivity analyses were done with these respondents also included. New or incident head injuries were defined as those who did not endorse any injury type at Visit 1 but later did at Visit 2. Endorsements of injuries causing symptoms of being dazed, concussive, or having memory loss were combined. Timing of LOC was used to categorize severity of injury with the data that were available, with moderate injury differentiated at LOC >20 minutes, although conventional clinical thresholds distinguish moderate injury at LOC >30 minutes.^35,36 These questions and the assessment of severity were designed to have comparability and consistency with existing cardiovascular cohorts, such as Atherosclerosis Risk in Communities (ARIC) and the Cardiovascular Health Study (CHS).^34,37

Analyses.

Participant characteristics were summarized by characteristics at Visit 1 and Visit 2 using mean and standard deviation (SD) or count (n) and percentage. UpSet plots provided graphical examination of intersections between combinations or sets of participant “yes” responses to each of the 3 head injury questions (A-“Injury LOC”, B-“Cause”, C-“Severity”), separated by Visit. Differences between Visit 1 and Visit 2 in response patterns to any endorsement of head injury (combined questions) were tabulated using count (n) and percentage. Further examination of differences between visits across and within questions and sub-questions was also tabulated. Multivariate logistic regressions evaluated mutually adjusted estimates of risk for endorsement of any head injury (combined questions), both prevalent (Visit 1) and incident (between Visit 1 and Visit 2), from sociodemographic, behavioral, and clinical features. Neuropsychological test scores (see Supplemental Table 1) related to head injury were examined using calculation of mean (SD) and mean comparison tests (P-value), for any prevalent or any incident head injury, as detailed above. Analyses were conducted using STATA v17-18 (College Station, TX).

RESULTS

Participants.

Mean follow-up time between Visit 1 (2010-2013, N=818) and Visit 2 (2017-2019, N=403) was 6.8 years (range 3.8 to 9.3 years). Participants were all older adults (Visit 1 mean age was 73 and Visit 2 mean age was78) and mostly female, with relatively low education (mean 12 years) but high prevalence of self-reported bilingual ability. Many were ever smokers, 16-17% reported using alcohol within the past year, and 7-8% reported prior physician-diagnosis of stroke. Symptoms of depression were common, self-perceived physical health was relatively low, but self-perceived emotional health and social functioning were relatively high. Mean 3MSE scores were 87-89 out of 100 and mean MoCA 18.9 out of 30. Supplement Table 2 provides full tabulation of participant characteristics by Visit.

Missing Data.

There are no missing field center, sex, age, education, bilingual, smoking, or alcohol use data. The CESD score was missing for 22 (2.7%) at Visit 1, 27 (6.7%) at Visit 2; the SF36 score was missing for 25 (3.1%) at Visit 1 and 37 (9.2%) at Visit 2; self-reported stroke was missing for 2 (0.5%) at Visit 2. Neuropsychological (3MSE) test data were missing in 7 (8.6%) at Visit 1 and 1 (0.2%) at Visit 2. Individual missing datapoints were excluded from calculation of counts or tabulation of proportions; regression models excluded individuals with partial data.

Prevalence.

Our primary definition of self-reported prior head injury (affirmative responses to question B--head injury from listed causes, or question C--head injury resulting in symptoms of being dazed, concussion, loss of memory, but not to question A--injury with loss of consciousness of varying times) resulted in 328 (40%) cases at Visit 1 and 146 (36%) at Visit 2 (Table 1). The more inclusive case definition (any affirmative responses to questions A, B, or C) resulted in N=345 (17 additional) cases at Visit 1 and N=154 (8 additional) cases at Visit 2. Correlation among the 3 questions was significant, with pairwise rho 0.5-0.8 (all P<0.0001). The numbers for separate response categories for Visit 1 and 2 respectively were: question A (“Injury LOC”) 23.2% and 18.6%, question B (“Cause”) 33.8% and 30.3%; and question C (“Severity”) 36.4% and 30.2%.

TABLE 1:

Positive endorsements to questions related to injuries causing loss of consciousness , reported causes of head injuries and severity of head injuries at Visit 1 (2010-2013, n=328) and at follow-up Visit 2 (2017-2019, n=146), among American Indians aged 65-95 years

	Visit 1 N=818	Visit 2 N=403
Endorsed yes to any of 3 questions on cause of head injury or head injury severity *	328 (40.1%)	146 (36.2%)
Northern Plains field center	109 (29.1% †)	57 (32.2% †)
Southern Plains field center	175 (50.6% †)	67 (38.1% †)
Southwest field center	44 (44.9% †)	22 (44.0% †)
Endorsed history of injuries causing LOC (A)	190 (23.2%)	75 (18.6%)
Number with >1 injury (range: 1-5)	47	16
Age at first occurrence: mean (SD), range	28 (21), 4-86	43 (25), 8-85
Number of years since last injury: mean (SD), range	32 (23), 0-70	34 (24), 2-64
Endorsed cause(s) of head injury (B) *	277 (33.8%)	122 (30.3%)
Fall	114	55
Motor vehicle accident	99	41
Sports or physical training	43	8
Domestic incident or assault	26	25
Fragment, bullet, or blast	13	14
Horse riding	8	15
Air or water transportation	2	0
Other, unknown	2	2
Endorsed questions on severity of head injury (C) *	298 (36.4%)	122 (30.2%)
Dazed, symptoms of concussion, loss of memory	194	59
Loss of consciousness <1 min or 1-20 min	105	47
Loss of consciousness >20 min	53	16

^*Severity and cause categories are not mutually exclusive, as many participants reported multiple head injury events with multiple precipitating causes or degrees of severity. Additional question on injuries causing loss of consciousness not included in primary case definition; however, sensitivity analyses with this question also included resulted in 17 additional cases at Visit 1 and 8 additional cases at Visit 2.

^†Calculated prevalence for each field center based on participation within each field center, and not by overall participation (Northern Plains Visit 1 n=374, Visit 2 n=177; Southern Plains Visit 1 n=346, Visit 2 n=176; Southwest Visit 1 n=98, Visit 2 n=50)

In graphical examination of unique combinations of endorsements to questions A (“Injury LOC”), B (“Cause”), or C (“Severity”), the largest proportion, said “yes” to all 3 questions (Supplement Figure), or “yes” to questions B (“Cause”) and C (“Severity”); non respondents answered “yes” only to question A (“Injury LOC”), due to the case definition. However, few responded to question A alone, with only 17 (Visit 1) and 8 (Visit 2) that would be recategorized if the broader case definition were used. The number of those reporting in Question A (LOC) to having more than 1 head injury was 47 (5.7% of Visit 1 participants) and 16 (4.0% of Visit 2 participants), with range up to 5 injuries (Table 1). Age at first occurrence ranged from 4 to 86 years old, with vide variance. Many had not had a head injury for decades, with mean number of years since last injury estimated between 32-34 years (range: 0-70 years).

Comparing regional variations in prevalence of endorsement to any of the questions (A, B, or C), the Northern Plains field center reported 29.1%, Southern Plains 50.6%, and Southwest 44.9% at Visit 1 and 32.2%, 38.1%, and 44.0%, respectively, at Visit 2 (Table 1). Kruskal-Wallis test for population differences by field center in proportion of positive endorsements to any of these three questions (A, B, or C) was significant at Visit 1 (P<0.001), but not at Visit 2 (P=0.247).

Cause and Severity.

The primary reported in B causes of head injuries were falls, motor vehicle accidents, and sports or physical training; however, fights or assaults, fragments/ bullets/ blasts, and horse riding were also reported (Table 1). In terms of C (“Severity”), those reporting prior injury with LOC>20 minutes, suggesting TBI of moderate severity, numbered 53, representing 17.% of 298 injuries or 6.5% of the cohort (Visit 1) and 16, representing 13.1% of 122 injuries or 4.0% the cohort (Visit 2). The majority of participants who responded to question C reported at least some degree of LOC (53% Visit 1, 52% Visit 2), or being dazed, concussed, or loss of memory (65% Visit 1, 48% Visit 2).

Incidence.

Among those who did not report any head injury at Visit 1 (n=490), 55 (11.2%) reported a head injury at Visit 2, most likely representing incident cases (Table 2). Incidence rate analysis of these 55 new cases over 1566.9 person-years between Visit 1 and Visit 2 suggests incidence rate for any head injury of 3.5 per 100 person-years (95% confidence interval: 2.6-4.5).

TABLE 2:

Change in self-report of history of any prior head injury between Visit 1 (2010-2013, n=818) and Visit 2 (2017-2019, n=403) among American Indian participants of the Strong Heart Study, aged 65-95 years

	No head injury reported at Visit 1 N=490	Any prior head injury reported at Visit 1 N=328
Lost to follow-up Visit 1 to Visit 2 (N=416)	251 (51.2%) *	165 (50.3%) *
No head injury reported at Visit 2 (N=256)	184 (37.6%)	72 (22.0%)
Any prior head injury reported at Visit 2 (N=146)	55 (11.2%) †	91 (27.7%) †

Percentages calculated based on columnar tabulations.

^*P-values calculated using test of equality in proportions: P=0.8574

^†Incidence rate analysis of 55 new cases over 1566.9 person-years accumulated between visits: 3.5 new cases per 100 PY (95% CI: 2.6-4.5)

Reporting differences over time.

Although a substantial number of participants were lost to follow-up, groups who did and did not report head injury at Visit 1 were not significantly different in likelihood to return for the follow-up visit (P>0.8, Table 2). However, of the 328 participants who did report any head injury at Visit 1, 72 (22.0%) did not later report any such injury at Visit 2, perhaps reflecting inconsistency in case reporting over time. In a sensitivity analysis of whether these 72 were substantively different in reported severity, compared with others who reported injuries at Visit 1 and who contributed data at Visit 2, these individuals were more likely to report being dazed (48% vs 16%, P<0.001), or having LOC<1 minute (25% vs 8%, P<0.001) but not more likely to report concussion, loss of memory, longer duration of LOC, and not to have differences in number of injuries (P=0.09), age at first injury (P=0.18), or year since last injury (P=0.16).

Risk factors.

Evaluations of sociodemographic, behavioral, and clinical features with prevalent head injury (Model A) suggest that male sex (Odds Ratio, OR: 1.8, 95% CI: 1.3-2.5, P<0.001) and prior stroke (OR: 2.6, 95% CI: 1.5-4.8, P=0.002) were both strongly and independently related to higher risk, whereas age, education, language use, smoking, and alcohol use were not associated (Table 3). In examinations of risk factors for incident head injury (Model B), male sex and prior stroke were both positively associated, although only stroke was statistically significant (P=0.039).

TABLE 3:

Mutually adjusted logistic regressions for risk features of reported prior head injury at Visit 1 (Model A), or for new head injury over approximately 7 years of follow-up between Visit 1 and Visit 2 (Model B), among American Indians aged 65-95 (2010-2019)

	Model A: Prior head injury before Visit 1		Model B: New head injury between Visit 1-Visit 2
	OR (95%CI)	P-value	OR (95%CI)	P-value
Male sex	1.83 (1.32, 2.53)	<0.001	1.97 (0.81, 4.79)	0.133
Age, years	0.98 (0.95, 1.01)	0.135	1.03 (0.96, 1.10)	0.457
Education, years	1.03 (0.98, 1.08)	0.227	0.95 (0.83, 1.10)	0.523
Bilingual	0.77 (0.57, 1.03)	0.082	0.86 (0.40, 1.86)	0.705
Ever smoking	1.20 (0.90, 1.59)	0.225	0.89 (0.35, 2.27)	0.813
Alcohol use within year	1.10 (0.70, 1.73)	0.677	0.95 (0.37, 2.42)	0.914
Prior stroke	2.63 (1.45, 4.78)	0.002	6.16 (1.10, 34.59)	0.039

Cases (Model A) were reported head injury at Visit 1. Cases (Model B) were new reported head injury at Visit 2, among those with no reported head injury at Visit 1. Time varying features (age, smoking, alcohol use, stroke) were modeled based on relevant examination visit: Visit 1 (Model A) or Visit 2 (Model B).

Neuropsychological instruments.

Neuropsychological test scores associated with prevalent head injury (Table 4) included more depressive symptoms (P-value=0.002); worse quality of life, fatigue, well-being, social functioning, pain, and general health (P-values <0.01). Also associated with prior head injury was worse processing speed (P-value=0.038) at Visit 2, and better phonemic fluency (P-value=0.04), although only statistically significant at Visit 1. Comparisons of other neuropsychological domains suggest that executive function was also worse among those with head injury (P-values=0.06), although these associations were not statistically significant.

TABLE 4:

Neuropsychological test scores comparing those with endorsement of prior head injury at Visit 1, or for new head injury over mean 7 years of follow-up between Visit 1 and Visit 2, among American Indians aged 65-95 (2010-2019)

	Visit 1			Visit 1 to Visit 2
	No head injury	Any head injury		No new head injury	Any new head injury
	N=490	N=328	P-value	N=184	N=55	P-value
CESD	10.4 (8.3)	12.3 (9.1)	0.002	12.4 (8.9)	12.6 (7.1)	0.88
SF36 overall score	57.7 (18.5)	54.5 (19.2)	0.017	54.9 (18.0)	49.9 (11.5)	0.069
SF36 physical functioning subscore	54.9 (28.8)	52.4 (29.3)	0.24	52.9 (28.2)	49.0 (24.9)	0.38
SF36 limitations due to physical health subscore	29.9 (41.1)	25.5 (38.8)	0.13	24.5 (37.7)	13.4 (27.3)	0.047
SF36 limitations due to emotional health subscore	46.1 (46.6)	41.6 (45.9)	0.17	30.4 (41.7)	24.2 (39.3)	0.33
SF36 energy/fatigue subscore	60.2 (14.5)	57.6 (14.8)	0.013	59.0 (13.9)	57.4 (13.9)	0.44
SF36 emotional well being subscore	70.9 (13.2)	68.6 (13.8)	0.013	70.9 (13.4)	70.8 (10.1)	0.94
SF36 social functioning subscore	79.4 (22.1)	74.5 (24.5)	0.003	78.3 (22.4)	75.2 (21.3)	0.37
SF36 pain subscore	68.1 (25.1)	60.2 (24.7)	<0.001	65.0 (24.2)	58.1 (22.3)	0.061
SF36 general health subscore	64.3 (18.9)	60.6 (19.5)	0.007	62.9 (18.9)	61.9 (17.4)	0.75
3MSE	88.3 (9.9)	88.9 (10.3)	0.42	86.8 (9.6)	86.9 (8.3)	0.97
WAIS DSST	44.5 (14.8)	45.2 (14.6)	0.56	42.7 (13.8)	38.3 (13.2)	0.038
COWA	23.9 (11.0)	25.6 (11.8)	0.040	23.4 (11.1)	25.0 (10.3)	0.33
CVLT-II, Short Delay Free Recall	6.0 (2.1)	5.9 (2.0)	0.52	5.5 (2.1)	5.5 (2.0)	0.95
CVLT-II, Long Delay Free Recall	5.5 (2.3)	5.4 (2.1)	0.58	4.8 (2.4)	4.7 (2.2)	0.74
MoCA	--	--	--	18.9 (4.4)	18.3 (4.0)	0.39
Craft, immediate recall	--	--	--	9.8 (4.5)	10.5 (4.0)	0.30
Craft, delay recall	--	--	--	8.4 (4.4)	9.0 (4.2)	0.43
Benson, copy	--	--	--	15.5 (1.6)	15.6 (1.5)	0.71
Benson, recall	--	--	--	8.8 (3.8)	8.4 (3.6)	0.54
Animal	--	--	--	13.5 (4.6)	13.8 (4.4)	0.64
Vegetable	--	--	--	9.5 (3.4)	9.9 (2.6)	0.46
Number, forward	--	--	--	5.6 (1.4)	5.8 (1.3)	0.37
Number, backward	--	--	--	3.5 (1.2)	3.5 (1.1)	0.91
TMT, A (seconds)	--	--	--	62.5 (30.2)	71.4 (32.4)	0.064
TMT, B (seconds)	--	--	--	159.2 (74.0)	178.6 (74.0)	0.10
MINT	--	--	--	27.1 (3.3)	27.0 (3.4)	0.90

Values given as mean (SD). Center for Epidemiologic Studies Depression Scale, CESD; Short Form 36, SF36; Modified Mini Mental Status Examination, 3MSE; Wechsler Adult Intelligence Scale 4th Edition digit symbol coding , WAIS DSST; Controlled Oral Word Association F,A,S, COWA; California Verbal Learning Test II edition Short Form, CVLT-II; Montreal Cognitive Assessment, MoCA; Craft Story (paraphrase), Craft; Benson Complex Figure, Benson; Animal, Vegetable naming tests, Animal, vegetable; Number Span Test, Number; Trail Making Test time to complete (seconds), TMT; Multilingual Naming Test, MINT.

DISCUSSION

This study provides population-based estimates among older American Indian adults of head injury prevalence (36-40%) and incidence (11% over mean 7 years, or 3.5 per 100 person-years). Approximately one-third of this population-based, aging cohort (30-36%), reported concussive symptoms and/or loss of consciousness, likely representing those with at least mild TBI, according to common grading criteria;^38,39 13-17% of these (4-7% of cohort population) reported symptoms consistent with TBI of moderate severity (LOC>20 minutes).

These findings of head injury incidence (3.5/100 PY) are higher, compared with epidemiologic estimates of head injury in White Europeans (2.2/100 PY),⁴⁰ and of TBI in the North American general population (1.3/100 PY)—although it is noteworthy that this analysis measured head injury and not TBI, per se.⁴¹ These population-based estimates of prevalent head injury (36-40%) are also higher, compared with prior studies on head injury prevalence in UK general population (18%) ,⁴² in similarly-aged (older adult) residents of eight middle-income countries in Asia and Central/South America (0.3-15%),⁴³ and in White Australians (6%).⁴⁴ Collectively, our data suggest that American Indian adults may have especially high accumulated risk of head injury, although differences in methodology to define head injury and TBI across studies, across regions, and over time may make result in bias for some cross-comparisons.

This study identified falls and motor vehicle accidents to account for most cases. However, injuries sustained from causes suggestive, but not exclusive, of military service (fragments/ bullets/ blasts), from fights or assaults, and from horse riding comprised a substantial portion as well. In the general population, the most common causes of head-injury-related deaths include self-harm, falls, and motor vehicle accidents; among persons aged > 65 years, falls are the most common cause.³⁸ Thus, our study identified causes of head injury that may be of particular importance to American Indian elders and their communities, suggesting that population-specific solutions for prevention and treatment may be needed. Furthermore, future studies should apply qualitative and community-based participatory research methodologies to capture the range of experience and perceptions related to this important public health problem.

Differences in response patterns among questions regarding different facets of head injury (injuries causing LOC, head injury types and causes, and head injury severity) suggest that careful design of data collection instruments are needed, especially among community-dwelling populations that may have differing cultural perspectives, socioeconomic backgrounds, or especially high risk. Many factors can influence accuracy of case identification, including recall bias for injuries sustained many years ago or injuries causing substantive change to memory; social desirability bias, especially affecting injuries sustained because of events that may reflect sensitive societal disparities; question-order bias, potentially resulting in respondent confusion; and others. In particular, if head injury and TBI are risk factors for development of dementia, and dementia patients are less likely to recall prior head injuries, then differential bias is likely to result in Type II measurement error, with bias towards the null. Furthermore, assessment of severity should be aligned with current clinical standards. Differences in sensitivity of our three questions on head injury, and the losses in positive endorsements between Visit 1 and Visit 2 suggest that future research in this and other populations may address such methodologic limitations with prospective data collection designed for this purpose.

Detected differences by field center were especially noteworthy, and warrant further consideration. In particular, whether reporting or recall bias may account for any of the observed differences by region; whether the observed differences by region are associated with different causes, risk factors, or outcomes; and whether such differences may result in long-term differences in related outcomes are important questions for further study.

Furthermore, although statistical testing may have been underpowered especially for the incident analysis between Visit 1-Visit 2, patterns of risk and feature associations were consistent overall, with males and those with prior stroke representing high risk groups. Development of public health prevention programs targeting males may be especially likely to result in effective primary prevention. Future research should examine whether confusion in response of vascular or traumatic brain injury could clarify the associations with stroke, or whether those with stroke also represent a high-risk group that may benefit from public health or clinical intervention.

Finally, future research should continue to examine additional risk and neuropsychological test scores, especially with prospective data collection tailored to key risk and protective factors that may represent modifiable targets for public health intervention. Given the limited potential for remote head injuries to cause continued changes to later-life cognition, these analyses were highly subject to potential for Type 2 error, emphasizing the potential importance of even the limited associations detected. Furthermore, although we did detect a few associations in the direction expected (significant negative association for processing speed; non-significant negative association for executive function), we also detected an association in the opposite direction from expected (positive, or protective significant association for phonemic fluency). Such a finding may represent bias, such as selection or recall bias, or it may represent an unexpected finding. Future research should build upon these findings with more direct, prospective data collection of both acute head injury with these and other cognitive outcomes.

In this study, those with self-reported prior head injury also had more depressive symptoms and worse health related quality of life. These findings were consistent with well-established associations with depression symptoms,⁴⁵ as well as with more recent reports on quality of life in male prisoners with reported lifetime history of head injury.⁴⁶ In contrast, prior studies have identified clear associations of head injury or traumatic brain injury with processing speed, phonemic fluency, and executive function, especially among veterans.⁴⁷ Future research with more precise case definitions may be able to assess specific neuropsychological profiles resulting from head injury in this population. Also, of note, prior head injury has been associated with subsequent maladaptive behaviors, such as substance use.^48,49 Our risk association analyses were cross-sectional, and so future work should build upon these findings with longitudinal data collections aimed at temporal quantification of risk behaviors.

Overall, our study has many strengths, including a consistent and rigorous data collection protocol in a population-based cohort of people underrepresented in epidemiological studies. Our data contribute new information on prevalence and incidence, as well as some novel insights in risk and consequences related to head injury. However, our study also has limitations. First, history of head injury was self-reported, which can limit its accuracy. Also, this report was focused on self-reported history of any prior head injury, instead of prospective collection of acute injury, which could further reduce accuracy of reporting. In addition, although questions A and B may represent those with non-head or superficial injuries, the concordance with question C is high, suggesting the majority of responses represent traumatic type head injuries. Although our neuropsychological test battery was extensive, it was designed to evaluate cognitive changes related to vascular and Alzheimer’s dementias, and not traumatic brain injury. It is possible that some participants who reported prior head injury had vascular or other events, which could also result in reporting bias. We did not collect age, date, recency, or number of prior head injuries, inhibiting ability to accurately evaluate severity of individual case histories. Furthermore, these analyses did not have access to continuous surveillance data, but rather only two-phase follow-up data, which can limit the accuracy and precision of estimates. Future studies would benefit from direct, high-quality, prospective data collection of acute head injuries over a specific timeframe with evaluation of detailed neuropsychological outcomes targeted to traumatic brain injury in this high-risk population. However, until such a study is able to be conducted, these data from the SHS represent the best available on an important topic to brain health and cognitive aging.

Second, if participants with head injury are less likely to survive or are less likely to remember a prior health event, our analyses may suffer from missing data or misclassification. This study had minimal missing data, and we found little evidence of selective survival based on self-reported history of head injury, among those who survived from Visit 1, ages 65-95 to Visit 2, aged 70-95. However, those who had prior head injury with especially severe effect may not have survived to participate in this study at all, and thus evaluation of younger groups at youth, young adult, and midlife is needed.

Also, although those with head injury may not consistently report such injuries later on, whether this is a reflection of initial case misclassification or later case recall bias is yet unclear. Future research should focus on detailed, objective, prospective data collection with instruments designed to collect information related to different degrees of head injury and TBI with alignment to ICD codes, in order to allow more powerful analytic assessment and to minimize error. Further, studies using measurable biomarkers related to acute injury—such as plasma neurofilament light chain, a marker of general neural injury and remodeling⁵⁰ —may supplement self-reported data in distinguishing individuals at highest risk. Our study-- limited to American Indians--did not allow direct comparisons of prevalence or incidence with other populations, and future works should enable cross-population investigations with high-quality data collection protocols.

Despite these limitations, our analyses present important descriptive and epidemiologic insight into head injury in a unique population. Although these data are flawed, they do provide critical insights that high-quality research focused on head injury and TBI are critically needed among American Indian and perhaps other Indigenous peoples, so that clinical care standards, public health prevention programming, and public policy may be tailored to specific need and context of the population. Future studies should further examine cause specific differences; probe more deeply into community, environmental, and cultural contexts; and use measures of recency or severity to improve understanding of cause and consequence in TBI among American Indian peoples.