The present research leveraged data from the Naval Health Research Center’s Career History Archival Medical and Personnel System (CHAMPS) database (41). This database maintains a longitudinal record of service members’ pay-affecting career records and medical diagnoses recorded during Tricare-reimbursable medical encounters in military- and civilian-based care settings. Service members were eligible for inclusion if they initially enlisted in the U.S. military between 1 October 2005 and 30 September 2014 and served for at least one year in the Army, Air Force, Marine Corps, or Navy. They were followed until one of the following occurred: they changed service branches, they became warrant officers or officers, their occupation changed that resulted in a risk categorization shift, they were discharged from military service for at least 30 days, or the study window end date of 31 December 2015. Eligible participants were classified as being at low, moderate, or high occupational risk of repetitive LLB exposure based on their military occupational specialty (categorized using Department of Defense-wide naming conventions) as described in Belding et al. (40). High-risk occupations included general armor and amphibious; artillery and gunnery; aviation ordnance; general combat engineering; general combat operations control; explosive ordnance disposal/Underwater Demolition Team; expeditionary medical services; general infantry; infantry, gun crews, seamanship specialists; military training instructor; missile artillery operating crew; rocket artillery; and special forces. Moderate-risk jobs included ammunition repair, artillery repair, counterintelligence, general armament maintenance, general law enforcement, independent duty hospital services, operational intelligence, security guards, and tracked vehicles. The remaining 160 occupations were categorized as low-risk.

Diagnoses of interest included TBI, conditions commonly comorbid with concussion, behavioral health conditions, and blast-related conditions (see Table 1). TBI diagnoses were examined together, as well as separately by severity (i.e., mild, moderate, severe, penetrating), in accordance with case criteria established by the Armed Forces Health Surveillance Branch Division (42). Conditions commonly comorbid with concussion and behavioral health conditions were identified using case criteria from a recent RAND report on concussion in the military (19). Diagnoses commonly comorbid with concussion included alteration in mental status, cognitive problems, communication disorders, dizziness/vertigo, gait and coordination problems, headache, hearing problems, non-headache pain, skin sensation disturbances, sleep disorders/symptoms, smell and taste disturbances, syncope/collapse, and vision problems. Behavioral health conditions included adjustment disorders, anxiety disorders, acute stress disorders, alcohol abuse/dependence, attention-deficit/hyperactivity disorder (ADD/ADHD), bipolar disorder, delirium/dementia, depression, drug abuse/dependence, personality disorders, and PTSD. Blast-related conditions were identified consistent with Belding et al., and included postconcussive syndrome, tinnitus, fatigue, and migraines (40). For each medical diagnosis or category of interest, the date of earliest diagnosis on record and whether the encounter occurred in inpatient or outpatient settings were identified. Specific ICD-9 codes are available in the supplement to Belding et al. (40).

When service members separate from military service, the CHAMPS database retains the official reason for separation. A total of 137 reasons were recorded which were categorized into one of the following mutually exclusive groups: administrative separation (e.g., “court-martial,” “failure to meet minimum qualifications for retention”), medical separation (e.g., “disability, severance pay,” “death, cause not specified”), early release (e.g., “early release, in the national interest”), end of active service (e.g., “expiration of term of service”), other (e.g., “conscientious objector,” “record correction”), reenlistment (e.g., “immediate reenlistment”), and retirement (e.g., “retirement, 20 to 30 years of service”). Those who were not separated were coded as “not discharged.”

Descriptive statistics were used to characterize the sample. Whether service members were medically or administratively separated, respectively, was regressed on flags for each diagnosis of interest, occupational risk of LLB, and their interaction while adjusting for branch of service (referent: Air Force) and sex (referent: males) in separate Cox proportional hazard models. Due to the number of comparisons and large sample size, a strict threshold of p ≤ 0.001 was used to determine statistical significance for omnibus tests. When significant interactions emerged, we decomposed them by rerunning the Cox proportional hazard models stratifying by occupational risk and compared 95% confidence intervals for overlap. Non-overlapping confidence intervals were interpreted as being significantly different. We opted to utilize 95% confidence intervals (rather than alternatives) for ease of comparison of our findings with those reported in other papers. Although this means the significance threshold for omnibus effects differs from post-hoc analyses, we felt that this decision was justified by the fact that we had specific directional a priori hypotheses and different thresholds for significance in decomposing interactions is inherent in calculations that adjust for family-wise error rates.

Across the board, occupational risk of LLB was associated with greater risk of being administratively separated from service (ORs ranged from 1.03–1.06; see Table 2). In contrast, medical diagnoses were associated with greater risk, no difference in risk, or less risk depending on the condition. For example, any TBI, mild TBI, and moderate TBI were all associated with significantly less risk of being administratively separated, whereas severe and penetrating TBI were not associated with differences in administrative separation at the p < 0.001 level. Among the conditions commonly comorbid with concussion, all were associated with decreased risk of being administratively separated, except for altered mental status and syncope and collapse, which were both associated with significantly greater risk. With the exception of delirium/dementia, all of the behavioral health conditions were associated with significantly greater risk of being administratively separated. One particularly notable finding is that those with (vs. without) drug abuse/dependence conditions were more than 5 times more likely to be administratively separated. A similar finding was noted with personality disorders such that those with (vs. without) the disorder were more than 4 times more likely to be administratively separated. Each of the blast-associated conditions were associated with significantly less likelihood of being administratively separated.

Of most importance, significant interactions emerged for several conditions including any TBI, mild TBI, moderate TBI, cognitive problems, communication problems, headache, hearing problems, any behavioral health problem, adjustment disorders, alcohol abuse/dependence, delirium/dementia, PTSD, postconcussive syndrome, and migraine (see Table 3). These significant interactions were decomposed by comparing 95% confidence intervals as a function of occupational risk. With only one exception (any behavioral health condition), the diagnosis of interest was associated with significantly less likelihood of being administratively separated from service among those in high-risk occupations relative to lower risk occupations. However, we note that planned comparisons were not always statistically significantly different at each level. Stated differently, while odds ratios were always lower among those in high-risk occupations than moderate or low-risk occupations, it was not always significantly lower than both. It is also worth noting that odds ratios for adjustment disorders and alcohol abuse/dependence were still greater than 1.0, whereas the others were all lower than 1.0.

Across all analyses, those working in high-risk occupations were significantly more likely to be medically separated from service than their lower risk peers (ORs ranged between 1.05 to 1.19; see Table 4). Additionally, those diagnosed with the conditions of interest were significantly more likely to be medically separated from service than those without the conditions of interest (ORs ranged between 1.11–7.07). The six diagnoses of interest with the largest odds of medical separation were: severe TBI (OR = 7.07), penetrating TBI (OR = 5.26), gait and coordination problems (OR = 4.68), skin sensation disturbances (OR = 3.50), bipolar disorder (OR = 3.47), and communication problems (OR = 3.05).

Significant interactions of occupational risk and diagnosis were observed for all diagnoses except severe TBI, penetrating TBI, gait and coordination problems, syncope and collapse, bipolar disorder, and drug abuse/dependence. This suggests that the association between diagnoses for these conditions and medical separation was not significantly greater for those working in high-risk occupations than lower-risk occupations.

However, significant interactions emerged for all of the other diagnoses of interest (see Table 5). When decomposed as a function of occupational risk, the data suggest that each diagnosis of interest (with the exception of tinnitus, which had overlapping confidence intervals) had a stronger association with medical separation among those working in high-risk occupations than their lower risk counterparts. For example, the association between mild TBI and medical separation was significantly greater for those in high-risk occupations (OR = 1.91) than those in moderate (1.31) and low-risk (OR = 1.38) occupations. Additionally, those with (vs. without) postconcussive syndrome were 3.21 times more likely to be medically separated when they worked in a high-risk occupation compared to 2.19 or 2.06 times more likely when they worked in a low- or moderate-risk occupation, respectively. As discussed with the findings for administrative separation previously, we note that planned comparisons were not always statistically significantly different at each level.

This research shows that having been diagnosed with any TBI, concussion, and moderate TBI were significantly associated with greater risk of being medically separated from service. They were also significantly less likely to be administratively separated from service, presumably because medical reasons for separation took precedence over administrative reasons, as the two determinations are mutually exclusive. Severe and penetrating TBIs were also significantly associated with being medically, but not administratively, separated from service. Conditions that are often comorbid with concussion in military populations were associated with greater risk of medical separation and were generally associated with less risk of being administratively separated, which is consistent with changes in military policy over time that aim to ensure that service members are eligible for post-service benefits when appropriate. However, diagnoses of altered mental status and syncope and collapse were associated with greater risk of administrative separation. Although surprising, it is possible that positive drug screens were obtained at the time SMs received care for altered mental status or syncope and collapse, which would likely result in administrative separation from service over medical separation from service.

Additionally, all of the behavioral health conditions examined were associated with increased risk of being both administratively and medically separated from service, which appeared consistent with military policy. For example, those with diagnoses of drug abuse/dependence were more than 5 times more likely to be administratively separated from service and only 1.66 times more likely to be medically separated from service. Conversely, whereas those with (vs. without) PTSD were 2.78 times more likely to be medically separated from service; PTSD was only associated with a 1.03 times greater risk of being administratively separated from service, which is an effect size too small to be practically meaningful. Finally, the four blast-associated conditions examined were each associated with greater risk of medical separation and reduced risk of administrative separation. For example, those diagnosed with postconcussive syndrome or migraines were 2.61 or 1.97 times more likely to be medically separated from service, respectively, than those without these diagnoses on record.

The present research is the first to demonstrate involuntary military job loss outcomes associated with TBI, mental health conditions, conditions associated with blast exposure using population-level data that includes both inpatient and outpatient medical records. Whereas previous research demonstrated adverse career outcomes among those hospitalized with TBI, the present research suggests that such outcomes are not limited to injuries severe enough to require hospitalization but may also occur with concussions diagnosed in outpatient settings. Although these specific analyses were conducted on a military sample, these findings have important implications for both military and civilian workplace environments. If TBI inhibits an employee’s ability to perform adequately in the workplace, the associated costs may be much larger than previously estimated when considered in the context of additional costs to employers (e.g., to recruit and train new staff), to the employee (e.g., lost earning potential and insurance coverage), and to society at large (e.g., long-term taxpayer funded healthcare) (43).

In addition to the findings described above, the present research examined whether those who work in occupations marked by increased risk of repetitive LLB exposure are significantly more likely to be medically or administratively separated from service. Generally speaking, when adjusting for the diagnoses of interest, occupational risk was associated with increased risk of medical and administrative separation from service, though the effect sizes were relatively small in nature. However, the association between diagnoses of 14 specific conditions (e.g., concussion, postconcussive syndrome, cognitive problems) and administrative separation was significantly lower among those working in high-risk occupations compared to their lower risk counterparts. However, service members in high-risk occupations were significantly more likely to be medically separated from service when diagnosed with 26 of the 33 conditions of interest. Taken together, these findings suggest that there may be an association between occupational risk of LLB and administrative and medical separations from service. It is particularly noteworthy that service members diagnosed with concussion, conditions associated with concussion, mental health conditions, and blast-associated conditions are significantly more likely to be medically separated from service when they work in occupations categorized as high (vs. lower) risk of repetitive LLB exposure.

Previous research suggests that service members with repetitive exposure to LLB are more likely to report a variety of subclinical symptoms including headaches, memory lapses and decrements in working memory, difficulty concentrating, and more (44–46). Although a substantial body of research is currently underway to assess performance outcomes (47), the present findings provide preliminary evidence that such performance outcomes could extend as far as an inability to perform one’s occupational duties and ultimately subsequent involuntary military job loss when the service member has sustained a TBI or associated diagnosis. These findings also underscore the importance of understanding the full range of outcomes associated with LLB exposure and subsequent medical and non-medical care required. For example, it is possible that additional resources may be required to care for Veterans with (vs. without) chronic LLB exposure, particularly in light of potential neurodegenerative effects as they age (48, 49). Because these findings suggest that chronic exposure to LLB may be associated with involuntary military job loss, it is also plausible that Veteran homelessness may be an additional outcome worthy of investigation (50). However, we suggest that additional research would be necessary in light of the relatively small effect sizes identified herein.