Head Injury and Long-term Mortality Risk in Community-Dwelling Adults

Key Points

Question  Is head injury associated with long-term all-cause mortality risk?

Findings  In this cohort study of 13 037 community-dwelling adults followed up between 1987 and 2019, all-cause mortality risk was higher among individuals with head injury vs those without. Mortality risk was also associated with having multiple head injuries or moderate, severe, or penetrating head injury.

Meaning  The findings suggest the importance of public health strategies to prevent head injury and targeted clinical interventions aimed at reducing morbidity and mortality.

Importance  Head injury is associated with significant short-term morbidity and mortality. Research regarding the implications of head injury for long-term survival in community-dwelling adults remains limited.

Objective  To evaluate the association of head injury with long-term all-cause mortality risk among community-dwelling adults, with consideration of head injury frequency and severity.

Design, Setting, and Participants  This cohort study included participants with and without head injury in the Atherosclerosis Risk in Communities (ARIC) study, an ongoing prospective cohort study with follow-up from 1987 through 2019 in 4 US communities in Minnesota, Maryland, North Carolina, and Mississippi. Of 15 792 ARIC participants initially enrolled, 1957 were ineligible due to self-reported head injury at baseline; 103 participants not of Black or White race and Black participants at the Minnesota and Maryland field centers were excluded due to race-site aliasing; and an additional 695 participants with missing head injury date or covariate data were excluded, resulting in 13 037 eligible participants.

Exposures  Head injury frequency and severity, as defined via self-report in response to interview questions and via hospital-based International Classification of Diseases diagnostic codes (with head injury severity defined in the subset of head injury cases identified using these codes). Head injury was analyzed as a time-varying exposure.

Main Outcomes and Measures  All-cause mortality was ascertained via linkage to the National Death Index. Data were analyzed between August 5, 2021, and October 23, 2022.

Results  More than one-half of participants were female (57.7%; 42.3% men), 27.9% were Black (72.1% White), and the median age at baseline was 54 years (IQR, 49-59 years). Median follow-up time was 27.0 years (IQR, 17.6-30.5 years). Head injuries occurred among 2402 participants (18.4%), most of which were classified as mild. The hazard ratio (HR) for all-cause mortality among individuals with head injury was 1.99 (95% CI, 1.88-2.11) compared with those with no head injury, with evidence of a dose-dependent association with head injury frequency (1 head injury: HR, 1.66 [95% CI, 1.56-1.77]; 2 or more head injuries: HR, 2.11 [95% CI, 1.89-2.37]) and severity (mild: HR, 2.16 [95% CI, 2.01-2.31]; moderate, severe, or penetrating: HR, 2.87 [95% CI, 2.55-3.22]). Estimates were similar by sex and race, with attenuated associations among individuals aged 54 years or older at baseline.

Conclusions and Relevance  In this community-based cohort with more than 3 decades of longitudinal follow-up, head injury was associated with decreased long-term survival time in a dose-dependent manner, underscoring the importance of measures aimed at prevention and clinical interventions to reduce morbidity and mortality due to head injury.

Traumatic brain injury (TBI) occurs when external physical forces cause disruption in normal brain function.^1-3 Leading causes of TBI include motor vehicle crashes, unintentional falls, and intentional self-harm.⁴ Based on the most recent data, there were approximately 223 135 TBI-related hospitalizations in 2019 and 64 362 TBI-related deaths in the US in 2020.⁵

Traumatic brain injury is generally understood to consist of 2 distinct phases. Primary injury describes the tissue distortion and destruction that directly result from the initial injury forces, and secondary injury occurs over time and encompasses a wide range of physiologic processes, including blood-brain barrier dysfunction, edema, ischemia, and intracranial hypertension.^6-10 It is well established that mortality in the hours to months following TBI is substantially increased,^11-14 and there is further evidence to suggest that long-term survival is reduced in individuals with TBI compared with the general population.^15-20 Limitations of prior studies include reliance on data exclusively from hospital admissions and study populations derived from a single hospital or geographic region.

As TBI is associated with significant morbidity, disability, and diminished quality of life,^21-24 an understanding of its implications for long-term mortality risk among diverse individuals is essential to inform clinical management. In this study, we leverage data collected over 30 years to examine all-cause and cause-specific mortality risk associated with head injury in a community-based sample of US adults. As prior research suggests that TBI-associated mortality risk may vary meaningfully across population subgroups, we consider the association between head injury and mortality by sex, race, age, and head injury number and severity.

This study assessed data from the Atherosclerosis Risk in Communities (ARIC) study, an ongoing community-based cohort study of 15 792 participants aged 45 to 65 years recruited from the 4 US communities in the northwestern suburbs of Minneapolis, Minnesota; Washington County, Maryland; Forsyth County, North Carolina; and Jackson, Mississippi, in 1987-1989 (>70% participation rate among eligible individuals).²⁵ Participants have attended subsequent in-person visits and are followed semiannually via telephone.²⁶ For the present analysis, follow-up extended from the date of entry into the cohort (1987-1989) until loss to follow-up (4.0% over the study duration), withdrawal from the study (5.4%), death, or administrative censoring on December 31, 2019.²⁷ Of the 15 792 ARIC participants initially enrolled, 1957 were ineligible due to self-reported head injury at baseline. We excluded 103 participants not of Black or White race and Black participants at the Minnesota and Maryland field centers in accordance with ARIC study analysis recommendations due to race-site aliasing, whereby mainly White participants were recruited in Minnesota and Maryland, only Black participants were recruited in Mississippi, and mainly White and Black participants were recruited in North Carolina. A further 373 participants were excluded for missing dates of head injury and 322 with missing covariate data, resulting in 13 037 included participants (Figure 1).

The ARIC study was approved by the institutional review boards of all participating institutions. All participants gave written informed consent to participate at each study visit. Further details about the ARIC study are found in eMethods 1 in Supplement 1. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Head injury with or without loss of consciousness was defined as an absorbing state based on self-report and data from emergency department visits and inpatient hospitalizations. Self-reported head injury was measured from visit 3 onward based on questions about head injury requiring physician or hospital care, loss of consciousness, number of head injuries, and year of head injury (eTable 1 in Supplement 1). The month and date for each self-reported head injury was imputed randomly using the random point method.²⁸ Head injuries requiring hospital care were identified using codes from the International Classification of Diseases, Ninth Revision (ICD-9) and Tenth Revision (ICD-10) based on records from community hospitals and claims data from the Centers for Medicare & Medicaid Services according to the Centers for Disease Control and Prevention case definition for head injury^29-31 (eTable 2 in Supplement 1). Hospitalization records were available from ARIC study surveillance of all community hospitals from January 1, 1987, to December 31, 2019, in Minnesota, Maryland, and North Carolina and to December 31, 2017, in Mississippi. Linked Centers for Medicare & Medicaid Services data for hospitalizations and emergency department visits were available for participants aged 65 years or older enrolled in Medicare fee-for-service Part B from January 1, 1991, through December 31, 2018.

We created a time-varying indicator for head injury, which equaled 0 until an individual self-reported or received an ICD-9 or ICD-10 diagnostic code for head injury and equaled 1 thereafter. We further classified individuals by number of head injuries using a time-varying categorical variable (no head injury, 1 head injury, and ≥2 head injuries). Finally, among the subset of head injury cases identified by ICD-9 or ICD-10 codes, we categorized head injuries by severity based on the Department of Defense head injury severity classification (no head injury; mild head injury; and moderate, severe, or penetrating injury).³²

Vital status and date of death were ascertained from state vital records, death certificates, and the National Death Index. We further classified cause-specific mortality based on diagnostic codes for underlying cause of death into the following categories: unintentional or traumatic; cardiovascular; endocrine, nutritional, or metabolic; gastrointestinal; infectious; kidney disease; neoplasm; neurologic; respiratory; and other.³³

Participant characteristics are shown overall and stratified by head injury status as median and IQR for continuous variables or number and percent for categorical variables. We conducted a descriptive analysis of cause-specific mortality by first calculating the number of overall deaths in each category and then calculating the percentage of deaths by category and head injury status, with associated 95% CIs calculated using the exact binomial method.

Cox proportional hazards regression models with follow-up (ie, time elapsed since enrollment in the ARIC study) as the time scale were used to model all-cause mortality risk as a function of time-varying head injury. We assessed whether the proportional hazards assumption was satisfied based on the complementary log-log plot and using Schoenfeld residuals.^34,35 Our main statistical analysis proceeded in 3 stages. First, we estimated the unadjusted association between head injury and all-cause mortality (model 1). Second, we adjusted for a priori–specified confounding variables, including sex, age, prior military service, race and field center, education, household income, physical activity, and alcohol consumption (model 2). Third, we incorporated adjustment for vascular risk factors (cigarette smoking, hypertension, diabetes, and body mass index) that have been previously associated with head injury and are established risk factors for mortality^36-39 (model 3). A 2-sided P < .05 was set as the threshold of significance. A detailed description of model covariates is found in eMethods 2 in Supplement 1.

As secondary analyses, we evaluated the association between head injury and all-cause mortality in subgroups defined by sex, race, and age (stratified at the median of baseline age for the overall study population [<54 or ≥54 years]). For each subgroup analysis, we conducted formal tests of statistical interaction. We further examined risk of all-cause mortality by number of head injuries. Among the subset of head injuries defined by ICD-9 and ICD-10 data, we evaluated the risk of all-cause mortality by severity. As a sensitivity analysis, we separately considered risk of all-cause mortality associated with self-reported and hospital-diagnosed head injury. Data were analyzed between August 5, 2021, and October 23, 2022. Statistical analyses were performed using R, version 4.0.0 (R Foundation for Statistical Computing).

Of the 13 037 participants included in this study, 2402 (18.4%) had at least 1 head injury, and 512 had 2 or more head injuries. There were 1862 individuals with head injuries identified from hospital diagnosis alone, 299 with head injuries identified via self-report alone, and 241 with head injuries identified from both hospital diagnosis and self-report. Median follow-up time was 27.0 years (IQR, 17.6-30.5 years), and among individuals with head injury, the median time from enrollment to first head injury was 17.0 years (IQR, 7.5-23.0 years). More than one-half of participants were women (57.7%; 43.3% men), 27.9% were Black (72.1% White), and the median age at baseline was 54.0 years (IQR, 49.0-59.0 years). A total of 7353 individuals (56.4%) were recorded as having died during the study period, of whom 1552 (64.6%) were individuals with head injury (Table 1). Among individuals with head injury, the median follow-up time from head injury to death was 4.7 years (IQR, 1.1-10.8 years). Timing of injury to death was similar for older participants (aged ≥54 years: median, 4.6 years; IQR, 1.1-10.7 years) and younger participants (aged <54 years: median, 4.8 years; IQR, 1.1-10.7 years).

We evaluated cause-specific mortality among individuals with and without head injury. Overall, the most common causes of death were neoplasm, cardiovascular disease, and neurologic disorders (Figure 2A). Among individuals with head injury, deaths caused by neurologic disorders and unintentional injury or trauma occurred more frequently as indicated by nonoverlapping 95% CIs (Figure 2B). We further evaluated specific underlying neurologic causes of death among participants with head injury (n = 355) compared with those without head injury (n = 745). Nearly two-thirds of neurologic causes of death were attributed to neurodegenerative disease (62.2%) among individuals with head injury compared with 51.4% among those without head injury. Neurodegenerative diseases, such as Parkinson disease and dementia, composed a greater proportion of overall deaths among individuals with head injury (14.2%) vs those without (6.6%). Among individuals with head injury, a further 30.7% of neurologic causes of death were attributed to cerebrovascular disorders; 1.1% to motor neuron disease; 0.8% to anoxic brain injury; and 5.1% to other, nonspecific neurologic disease.

Cox proportional hazards regression models were used to model all-cause mortality risk as a function of time-varying head injury. The proportional hazards assumption was satisfied based on Schoenfeld residuals and the complementary log-log plot (eFigure 1 in Supplement 1). Kaplan-Meier curves depict that cumulative survival was consistently decreased among individuals with head injury compared with those without head injury (eFigure 2 in Supplement 1). Estimated hazard ratio (HRs) for the association between head injury and mortality for each model specification, as well as for covariates included to control for potential confounding (models 2 and 3), are depicted in Table 2. In unadjusted analysis (model 1), the HR of mortality among individuals with head injury was 2.21 (95% CI, 2.09-2.34) compared with those without head injury. The association was attenuated but remained significant with adjustment for sociodemographic factors (model 2: HR, 1.99; 95% CI, 1.88-2.11) and with adjustment for vascular risk factors in the final model (model 3: HR, 1.92; 95% CI, 1.81-2.03). In subgroup analyses, we found no evidence of a statistical interaction by sex or race, although results were attenuated among individuals aged 54 years or older at baseline (<54 years: HR, 2.26; 95% CI, 2.03-2.51; ≥54 years: HR, 2.00; 95% CI, 1.87-2.14; P for interaction = .03). Model results for subgroup analyses are presented in eTables 3 to 5 in Supplement 1.

We further examined the risk of all-cause mortality by number of head injuries and by injury severity. By number of prior head injuries, the HR for individuals with 1 head injury was 1.66 (95% CI, 1.56-1.77), and the HR for those with 2 or more head injuries was 2.11 (95% CI, 1.89-2.37) compared with no head injury (Figure 3A and C; eTable 6 in Supplement 1). In analyses restricted to hospitalized individuals for whom measures of head injury severity were available, the HR for mild head injury was 2.16 (95% CI, 2.01-2.31), and the HR for moderate, severe, or penetrating head injury was 2.87 (95% CI, 2.55-3.22) compared with no head injury (Figure 3B and D; eTable 7 in the Supplement). In sensitivity analysis, we separately considered head injury identified via self-report vs hospital record. There was no clear evidence of a difference in all-cause mortality among individuals with self-reported head injury vs those with no head injury (HR, 0.98; 95% CI, 0.87-1.10), although we did observe an association among individuals with hospital-recorded head injuries (HR, 2.14; 95% CI, 2.01-2.27) (eTable 8 in Supplement 1).

This cohort study focuses on long-term mortality risk associated with both self-reported and hospital-diagnosed head injuries in a diverse cohort of community-dwelling US adults followed for 30 years. We found an approximately 2-fold increase in all-cause mortality risk associated with head injury. It is important to note that head injuries cannot be and are not randomly assigned; therefore, our findings could reflect that individuals who experienced head injury have increased mortality risk at baseline, that head injury causes an increase in mortality risk, or a combination thereof. Regardless of the underlying mechanism, results of this study underscore the importance of public health measures that prevent head injuries and swift clinical intervention to reduce morbidity and mortality if head injury occurs.

Our findings are consistent with results from several prior studies that showed decreased long-term survival associated with TBI among individuals identified through hospital and rehabilitation facility discharge records.^15-17,40,41 Our results are also consistent with an analysis of data collected from a community-based random sample of Olmsted County, Minnesota, residents with confirmed TBI identified through the Rochester Epidemiology Project (REP). Analysis of the REP cohort revealed increased all-cause mortality risk in individuals with TBI compared with matched population control participants over the 22-year study, although this association did not persist when individuals with head injury were compared with control participants with other injury.¹⁹ Our study naturally extends prior research that includes only those individuals diagnosed via hospital discharge records by including those with self-reported, likely milder, head injuries. Additionally, although analysis of TBI in Olmsted County constitutes a key contribution to the literature by leveraging data gathered from a community-based sample, Olmsted County residents are predominantly White (approximately 90%), and residents receive medical care from 2 group practices.⁴² Our study adds to the literature by evaluating all-cause mortality risk associated with TBI within a geographically diverse community-based sample of Black and White study participants. We estimate that median survival after head injury is 4.7 years (IQR, 1.1-10.8), which is comparable to 6.8 years in the REP population; importantly, the REP study population is significantly younger than the ARIC study population. While our results likely have limited generalizability to survival risk among young adults with head injury, the majority of TBI-related morbidity and mortality is concentrated among older adults. Based on US surveillance data, adults aged 75 years or older account for approximately 32% of TBI-related hospitalizations and 28% of TBI-related deaths.⁵

Another contribution is our analysis of cause-specific mortality. We found that neurologic and unintentional and traumatic causes of death occurred more frequently among individuals with head injury. Neurodegenerative diseases, such as Parkinson disease and dementia, composed a greater proportion of overall deaths among individuals with head injury (14.2%) vs those without (6.6%), and a greater proportion of deaths were due to neurologic causes among individuals with head injury (62.2%) vs those without (51.4%), raising the possibility of reverse causality wherein individuals with neurodegenerative diseases may be more likely to fall and sustain head injuries. Our results, therefore, point to the need for future research that is focused on better elucidating the time-dependent association among neurodegenerative diseases, head injury, and mortality, particularly where detailed longitudinal measures are available. Our finding that deaths due to neurodegenerative disorders comprised nearly two-thirds of deaths due to neurologic causes in head injury is also consistent with a prior analysis of this study population that showed an increased risk of dementia associated with head injury.⁴³

In subgroup analyses, results were similar by race and sex. In a prior analysis of a nationally representative sample of US adults, the prevalence of TBI was higher among men than women.⁴⁴ In our data, head injury occurred more frequently among women than in men, which may partly reflect the female predominance in our study population and greater life expectancy of women. Our findings by sex are inconsistent with some prior research showing an increased mortality risk following TBI in women vs men.^45-47 Few studies have specifically considered racial and ethnic differences in TBI mortality risk. By age, the association between TBI and all-cause mortality was stronger among younger (<54 years at baseline) vs older individuals. Descriptive studies have shown that TBI incidence peaks in young adulthood and among older adults,^4,48 and some research has suggested that mortality risk associated with TBI is higher in middle age.^17,49 Our findings of increased all-cause mortality associated with TBI in older adults may reflect differences in injury severity or mechanism among younger vs older adults.

We further reveal a dose-response pattern in which the strength of association with all-cause mortality increased with the frequency and severity of head injury. In our analysis of head injury frequency, the association with all-cause mortality was greater for individuals with 2 or more head injuries but persisted among individuals with only 1 head injury. This finding suggests that any head injury confers increased mortality risk and that subsequent head injuries may further increase that risk.

In our analysis of severity, we found an association between all-cause mortality and moderate, severe, or penetrating head injuries. These findings are consistent with those from sensitivity analyses in which hospital-diagnosed head injuries were associated with mortality risk, whereas head injuries diagnosed via self-report were not. Additionally, our findings are in keeping with several prior studies that found evidence of increased all-cause mortality risk among individuals with moderate to severe TBI compared with those with mild TBI⁵⁰ or no TBI.^15-17 Importantly, our results reveal that all-cause mortality risk remains increased with mild head injury compared with no head injury, which should motivate ongoing consideration of TBI frequency and severity (including mild TBI) as determinants of subsequent morbidity and mortality.

Our study has several limitations. First, our findings may not be generalizable to other US regions, racial and ethnic subgroups, and age groups not represented in this study population. Although relatively smaller sample sizes from some study sites (ie, Black participants from Forsyth County) may have influenced study results, participants were selected using probability sampling and are therefore representative of each community. Second, head injury was ascertained via hospital or emergency department diagnoses and self-reported head injuries. Definitions of TBI based on ICD-9 or ICD-10 codes are susceptible to misclassification, particularly for mild injuries, with validation studies reporting sensitivities of 55% to 72% and specificity of 80% to 85%, depending on the code definition used.^51,52 We used standard ICD-9 and ICD-10 codes to identify hospital and emergency department diagnoses of head injuries.^29,31 The language of questions used to assess self-reported head injury changed over time but focused on individuals who required medical attention or had an associated loss of consciousness. Therefore, mild head injuries may not have been captured, although the attenuation of the association with mortality observed when using self-report–defined head injury suggests that self-reported cases in the ARIC study are largely milder than the ICD-9 and ICD-10 code–defined ones. However, prior research has suggested that self-report offers reliable assessment of head injury, even among individuals with cognitive impairment.⁵³ Third, study data were available on a daily (not hourly) level, so we could not distinguish between associations of the primary vs secondary phases of head injury with mortality risk. Fourth, we did not have details on other traumatic injuries or TBI mechanisms, which may yield additional insights, particularly among women for whom intimate partner violence is a more frequent cause of TBI.⁵⁴

This study reports an approximately 2-fold increase in long-term mortality risk associated with head injury (identified by both self-report and hospital diagnosis) in a diverse cohort of US adults followed for 30 years. This finding was consistent within subgroups defined by race and sex, with an attenuated strength of association among adults aged 54 years or older at baseline. Analysis of head injury frequency and severity revealed an association with increased mortality risk among individuals with 2 or more head injuries and among those with moderate, severe, or penetrating head injuries. This robust and dose-dependent long-term association underscores the importance of public health measures aimed at preventing head injuries and targeted clinical interventions to reduce morbidity and mortality after head injury.

Accepted for Publication: November 18, 2022.

Published Online: January 23, 2023. doi:10.1001/jamaneurol.2022.5024

Corresponding Author: Holly Elser, MD, PhD, Department of Neurology, Hospital of the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104 (holly.elser@pennmedicine.upenn.edu).

Author Contributions: Drs Elser and Schneider had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Elser, Gottesman, Schneider.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Elser.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Elser, Schneider.

Obtained funding: Coresh, Mosley.

Administrative, technical, or material support: Walter, Coresh.

Supervision: Diaz-Arrastia, Schneider.

Other: Gottesman.

Conflict of Interest Disclosures: Drs Gottesman and Schneider reported being associate editors for the journal Neurology at the American Academy of Neurology outside the submitted work. Dr Coresh reported receiving grants from National Institutes of Health (NIH) during the conduct of the study. Dr Mosley reported receiving grants from NIH during the conduct of the study and outside the submitted work. Dr Schneider reported receiving grant W81XWH-21-1-0590 from the Department of Defense outside the submitted work. No other disclosures were reported.

Funding/Support: The Atherosclerosis Risk in Communities (ARIC) study is supported by contracts 75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, and 75N92022D00005 from the National Heart, Lung, and Blood Institute (NHLBI). The ARIC Neurocognitive Study is supported by grants U01HL096812, U01HL096814, U01HL096899, U01HL096902, and U01HL096917 from the NHLBI, National Institute of Neurological Disorders and Stroke, National Institute on Aging, and the National Institute on Deafness and Other Communication Disorders. Dr Schneider was supported by grant K23NS123340 from the National Institute of Neurological Disorders and Stroke.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2.

Additional Information: The authors thank the staff and participants of the ARIC study for their important contributions.