Correspondence of the Boston Assessment of Traumatic Brain Injury-Lifetime and In-Theater Department of Defense Medical Records

Background: Since 2006, efforts have been made to increase the identication of traumatic brain injuries (TBIs) in post-9/11 military personnel. The BAT-L is the rst validated instrument to diagnose TBIs throughout the lifespan in post-9/11 Veterans. The objective is to investigate the correspondence of the Boston Assessment of TBI-Lifetime (BAT-L) diagnostic prevalence and injury severity of traumatic brain injury with in-theater medical records from Department of Defense (DoD). Methods: A convenience sample of 153 Veterans deployed in 2011 enrolled in the TRACTS longitudinal cohort study was examined. Retrospective review of DoD online medical records to determine diagnostic prevalence and injury severity for all head injury cases during deployment were compared with diagnostic prevalence and injury severity from the BAT-L clinical interview using Chi-square analyses. Results: There was moderate correspondence for TBI diagnosis between the BAT-L and DoD records (κ = 0.42). Sensitivity was 72.7% and specicity was 82.8%. Comparison of injury severity also had moderate correspondence (κ = 0.41). Missing TBI diagnostic data from DoD records was frequent; 43% percent of TBIs reported on the BAT-L did not have any documentation of mTBI assessment or diagnosis in DoD records while 83% did not have in-theater documentation. Conclusions: Diagnosis of TBI via the BAT-L retrospective interview was both sensitive and specic when compared to DoD medical records. However, diagnostic correspondence was only moderate. This lack of diagnostic agreement was related to multiple factors including lack of documentation of injury, differences in assessment tools and goals, and other combat-related motivational factors associated with failure to report injuries while deployed. Several policies were implemented to address underreporting and under-documentation of TBI, yet challenges remain. Findings suggest changes at both individual-level (e.g. service members) and system-level (e.g. DoD/military branches) are needed to adequately diagnose and document all TBI during deployment.

diagnostic agreement was related to multiple factors including lack of documentation of injury, differences in assessment tools and goals, and other combat-related motivational factors associated with failure to report injuries while deployed. Several policies were implemented to address underreporting and under-documentation of TBI, yet challenges remain. Findings suggest changes at both individuallevel (e.g. service members) and system-level (e.g. DoD/military branches) are needed to adequately diagnose and document all TBI during deployment.

Background
Mild traumatic brain injury (mTBI) is prevalent among post-9/11 Veterans, comprising of an estimated 82% of total TBIs worldwide occurring from 2000 to 2018. 1 Approximately 20% of US troops experienced a probable TBI (mild, moderate, or severe). Of these, an estimated 57% were not clinically evaluated. 2 Underreporting and under-documentation of injury is likely driven by a variety of factors including unawareness of sustaining a mTBI, lack of knowledge regarding concussion evaluation, and/or military TBI screening policies that have evolved over the con icts. 3 Further, some Veterans may have poor recollection of the head injury itself as a result of altered mental status (AMS) or posttraumatic amnesia (PTA) at the time of injury. 4 Mild TBIs are often not visible, not readily identi ed, and as a result often overlooked. Perceived stigma, a "tough it out" military mentality, con dentiality issues, commitment to duty/mission, and challenges with navigating the military healthcare system all contribute to underreporting. [5][6][7] In 2006, the military services and Department of Defense (DoD) instituted a number of policies initiating screening measures to address underreporting and under-documentation of TBI among deployed service members. 8 In 2010, the DoD further mandated in-theater "event driven" TBI protocols for those who were within 50 meters of any blast explosion, were in a vehicle associated with a blast event, collision/rollover, or sustained a direct blow to the head, 9,10 In addition to screening measures and in-theater TBI protocols, the Post-Deployment Health Assessment (PDHA) was mandated to be administered to all being demobilized following deployment. The PDHA included a series of TBI screening questions, all of which must be endorsed to prompt a follow-up evaluation. 11 Challenges in TBI screening and assessment Retrospective TBI identi cation methods are prone to high false positive and false negative error rates.
Many screening methods consist of closed-ended questions that overestimate prevalence rates, as they are designed to capture all potential head injuries that cause residual symptoms. 12 Most screening measures do not reliably assess severity of injury or duration of symptoms, 13 are not designed to disentangle acute TBI symptoms (AMS, PTA, LOC) from symptoms due to other etiologies (e.g., psychological sequelae associated with trauma, postconcussive symptoms), and fail to evaluate for multiple concussions. 13 Another signi cant challenge in TBI identi cation methods is reliance on self-report of historical events.
Experts have argued that TBI can only be accurately diagnosed at the time of injury or within the rst 24 hours. 4 However, immediate TBI assessment, particularly in active combat settings is often not possible.
Self-report using a validated TBI assessment instrument is currently considered the criterion standard for evaluating historical TBI. 13,15 Iverson and colleagues suggested that retrospective diagnostic challenges could be remediated with the development of validated, structured clinical interviews. 4,12 The Boston Assessment of Traumatic Brain Injury (BAT-L) addressed several of Iverson's suggestions. 14 The BAT-L is a well-validated, semi-structured, comprehensive clinical interview designed to assess TBI etiology speci c to post-9/11 Veterans across the lifespan including exposure to blast munitions, TBIs sustained during military service (blast and blunt etiology), and civilian TBIs before and after military service. 13 The BAT-L uses an open-ended, forensic approach to establish a timeline for each TBI event and approximate duration of acute TBI symptoms (AMS, PTA, LOC). The BAT-L demonstrated strong correspondence with the Ohio State University TBI Identi cation Method (OSU-TBI-ID), a valid and reliable assessment tool used to determine lifetime history of TBI (Cohen κ = 0.89; Kendall τ -b = 0.95) and has strong inter-rater reliability (κ s > 0.80). 13,16 The long-term consequences of mTBI remain unclear. [17][18][19][20] However, research suggests that mTBI may have a synergistic effect on functional 21,22 and mental health outcomes. 23 Identi cation of mTBI is essential to target treatment and understand long-term outcomes in military service members. 24 This study investigated the correspondance of retrospective TBI diagnosis on the BAT-L with in-theater (e.g., occurring during deployment) medical record documentation in close proximity date of injury. Speci cally, we compared incidence and severity of injury of each positive TBI diagnosis during military deployment from the BAT-L with DoD medical records. We further compared concordant (true positives; true negatives) with discordant (false positives; false negatives) groups to determine if there were any differences in demographics or clinical characteristics that may have contributed to discrepancy in TBI diagnosis. Lastly, using narrative injury accounts from the BAT-L, we explored reasons for diagnostic discrepancies.

Participants
Participants were Veterans (age 18-75) enrolled in the TRACTS longitudinal cohort study 25 , which assesses the biological, neurobiological, psychological, and cognitive functioning of post-9/11 Veterans. Veterans were recruited by a recruitment specialist who attended military events involving US Air National Guard, Marine and Marine Reserves, and Army and Army Reserve Units. Exclusionary criteria were individuals with a history of neurological illnesses, seizure disorders, psychotic disorders, bipolar disorders, or active suicidal/homicidal ideations. This study has been approved by VA Boston by Institutional Review Boards for human participants' protection.
The Defense and Veteran Brain Injury Center, which is responsible for tracking military TBI data attained from Armed Forces Health Surveillance Branch and Theater Medical Data Store, reported the highest total worldwide TBI diagnoses occurred in 2011 with more than 30,000 diagnoses. 26 To target participants with greater likelihood of TBI assessment and documentation in-theater, we included participants from the larger TRACTS sample 25 who were positive for military TBI and deployed during the 2011 calendar year (n=163). We excluded study participants who deployed in other years, those who failed the Medical Symptom Validity Test (n=7), which could interfere with validity of self-report, and those who could not be identi ed in the Joint Legacy Viewer (JLV) clinical application that provided access to DoD online medical records (n=3). The nal analytic sample included 153 individuals (149 men/4 women) with TBI occurring during military service as a result of blast (e.g., improvised explosive device) and/or blunt impact exposures (e.g., vehicular accident, combatives, fall). Three Veterans sustained two TBIs within their 2011 deployments, resulting in 156 injury events available for comparison ( Figure 1).
Procedures BAT-L. Information collected from the BAT-L included diagnosis of TBI (positive or negative) and severity of TBI (mild, moderate, or severe) following ACRM and DoD criteria. For details regarding TBI diagnostic criteria and interview procedures see. 13 DoD online medical records. Joint Legacy Viewer was used to access DoD records from deployment. Each Veteran's medical chart was ltered from 9/12/2001 to 3/24/2020 for review. All notes during this time period that clearly related to any head injury (blast or blunt etiology regardless of severity) and notes with no or ambiguous titles (e.g., consultation report, emergency visits) were thoroughly reviewed by a clinical psychologist. Any documented head injury occurring within a deployment that overlapped any period within the 2011 calendar year was identi ed. In-theater notes closest in proximity to date of head injury were considered the primary data source. If they were not available, other notes (e.g., PDHA) were examined. 27 In order to examine the predictive validity of the BAT-L as compared to DoD records, each speci c head injury case was matched based on the details of narratives available in both records. Speci cally, narratives often provided information such as location, date, method of injury, and/or sequence of events that were used to match the head injury reported in the DoD records to the BAT-L.
Information from DoD was used to determine both TBI diagnosis and TBI severity. Diagnostic consensus for DoD documentation was held with at least two doctorate-level psychologists who were blinded to the BAT-L data and TBI was diagnosed if there was any information in the records that indicated AMS, PTA, or LOC had occurred. Severity was classi ed as mild, moderate, or severe based on clear documentation of the duration of LOC, AMS, and PTA.

Statistical Analyses
Chi-square analyses were used to compare diagnostic agreement (Yes/No TBI diagnosis; TBI severity) between the BAT-L and DoD records. Sensitivity, speci city, kappa, and Kendall's tau-b were calculated for TBI diagnosis and severity. Demographic information and other clinical characteristics were examined (Table 1). These characteristics were compared between concordant TBI cases ("true positives" = positive TBI diagnosis on BAT-L and DoD records; "true negatives" = negative TBI diagnosis on BAT-L and DoD records) and discordant TBI cases ("false positives" = negative TBI diagnosis on the BAT-L, but positive in DoD records; "false negatives" = positive TBI diagnosis on the BAT-L, but negative in DoD records) using chi square, Fisher exact test (expected cell counts < 5), and t-test analyses. All p-values reported were corrected for multiple comparisons using the false discovery rate method 28 and refer to two-tailed tests (Table 1). All analyses were conducted using SPSS 26.0.0.1 software (IBM Corporation, Armonk, NY).

Results
Mean participant age was 31.85 years old (range 20-64, SD = 9.06). Participants were predominantly white (78%) and male (97%; Table 1). Of the 156 injury events, 19% had in-theater DoD documentation, 43% had post-deployment notes only, and 38% did not have any notes in the DoD records.
Overall, individuals in concordant and discordant groups were similar in regard to clinical and demographic variables ( Table 1)

Correspondence between BATL and DoD
There was moderate correspondence for TBI diagnosis between the BAT-L and DoD records (κ = 0.42; Kendall τ-b = 0.45; Table 2). Sensitivity was 72.7% and speci city was 82.8%. Comparison of TBI severity between the BAT-L and DoD records was also moderate overall (κ = 0.41; Kendall τ-b = 0.46; Table 2). Sensitivity was 66.7% for mild TBI (n = 14) but 100% for moderate TBI (n = 1). There were no severe TBI cases; Speci city for TBI severity was 82.8%. Six Veterans were positive for TBI based on the DoD records, but were not diagnosed with TBI on the BAT-L (Table 3). Based on event narratives from BAT-L and DoD, almost all ( ve of six false positives cases) were con rmed to be the same head injury event on both the BAT-L and DoD records; one case was unable to be matched due to lack of information from DoD records. Diagnostic discrepancies were primarily due to difference in approach in TBI evaluation. The BAT-L used a forensically guided semistructured interview focused on duration of acute TBI symtoms, while in-theater documentation consisted of brief screening.  Veteran reported vague symptoms (e.g., confusion) that lasted for a "few seconds" were not considered su cient to warrant positive TBI diagnosis.

Psychological vs. AMS symptomology
Veteran's disrupted thinking after a blast explosion was determined to be more likely attributable to a psychological response rather than acute AMS from TBI.

PCS vs. AMS symptomology
Veteran reported PCS symptoms only (e.g., headaches) that were misdiagnosed as AMS in DoD record.

Confounding factors 1 Unable to match head injury
One Veteran failed to report a TBI on the BAT-L that was documented in his DoD records.
False Negatives Twenty-three Veterans were negative for TBI based on DoD records, but were diagnosed with TBI on the BAT-L. Only four of the false negative cases had in-theater documentation of injury, while 19 cases had either no DoD records (n = 10) or only post-deployment screening notes available (n = 9; Table 4). Veteran reported undergoing a concussion evaluation, but was subsequently sent back out to continue mission.

Discussion
In-theater documentation of head injury was unavailable in DoD records for the large majority of cases (81%). Among Veterans who reported injuries that met diagnostic criteria for TBI on the BAT-L but were negative for TBI on DoD records (false negatives), 43% of cases did not have any DoD documentation of injury (in-theater or post-deployment notes). Importantly, when DoD documentation was available, injury events could be con rmed in the large majority of cases based on corresponding narratives, supporting the predictive validity of the BAT-L retrospective TBI diagnosis. These ndings indicate that although DoD and military services made substantial efforts to increase TBI screening, documentation of injury assessment was rare. We cannot determine if the lack of available records is due to lack of reporting, lack of documentation, or more likely, a combination of these factors.
The BAT-L yielded a diagnostic sensitivity of 73% and speci city of 83% for TBI with available DoD records. Further analyses of TBI severity ratings indicated that sensitivity was 67% for mild TBI and 100% for moderate TBI. Sensitivity and speci city are inversely related; an increase in sensitivity will result in a decrease in speci city and vice versa. 29 Due to this tradeoff, TBI assessment tools should be selected based on clinical utility. For example, initial screener tools, which have a primary goal of identifying all possible TBI, bene t from having higher sensitivity rates. After screening, individuals should receive a more comprehensive assessment to determine de nitive diagnosis. Diagnostic agreement was lower for mTBIs, but was excellent for injuries of greater severity (moderate TBI).
The moderate correspondence observed for TBI diagnosis between the BAT-L clinical interview and DoD records was likely a result of multiple factors. Among the false negatives, six Veterans reported that they did not report injury or seek any care for their head injuries on the BAT-L. There are many potential reasons a Veteran may not report injury during deployment including lack of awareness they sustained a concussion, perceived stigma of TBI, a "battlemind" mindset related to military culture, con dentiality issues that may impact career advancement, and/or could delay returning home from deployment. 3,5,14 Among those who were evaluated for TBI in-theater and had DoD documentation of the injury, there were four Veterans who reported (BAT-L) that they were mid-mission when their TBI event occurred and screened negative on in-theater TBI medic assessment. It is possible that these Veterans did not endorse symptoms because they did not want to disrupt their mission and/or wanted to return to the eld. 4 Another nine Veterans reported they were examined by a medic when queried speci cally about in-theater medical attention during the BAT-L clinical interview. However, there was no documentation of any other injuries (such as a fractured shoulder) that likely took precedence over mTBI assessment. Additionally, TBI assessments in battle eld conditions are understandably challenging. During combat/missions, the attending medic often provides triaged care, quickly evaluating a number of individuals, tending to critical needs rst, or potentially participating in combat before he/she is able to attend to injuries. Documentation of mTBI under these circumstances may not always be feasible.
For Veterans who were evaluated and in-theater DoD documentation was available, there were a number of disagreements between the BAT-L and DoD records. The primary cause of the discrepancies was the differing nature of the TBI assessment tools. In-theater assessments necessarily involve briefer screening, usually limited to a symptom checklist format, designed to be more sensitive than speci c, and prone to false positives errors to catch all potential injuries. Furthermore, the PDHA, which was admistered postdeployment, contained a single yes/no question to assess AMS. Service members were asked only if they were "dazed and confused or seeing stars." Without additional information, endorsement of this single item may have also contributed to a high false positive rate, especially in the context of a combat situation (e.g., chaos of battle, stress, acute traumatic reaction). 14 Comparatively, the BAT-L clinical interview adopts a more conservative, forensic approach including a detailed narrative and timeline for each injury event. Semi-structured diagnostic interviews administered by a quali ed clinician are considered the criterion standard for diagnosing lifetime history of TBI. 4,15 Combat settings pose unique diagnostic challenges given common co-occurring physical and psychological sequelae. Therefore, the BAT-L probes for eyewitness accounts, severity of other injuries, and other factors that may be misinterpreted as an acute symptoms of TBI. Examiners query the Veteran's psychological state at the time of the head injury to provide context to assess whether confusion may be better explained by shock/acute trauma reaction. Lastly, the BAT-L guides the examiner to disentangle acute AMS symptoms from lingering PCS (e.g., generally slowed thinking versus acute disruption of thinking at the time of injury), which are often inappropriately lumped together in briefer screening assessments as demonstrated in at least one diagnostic discrepancy in this study. Thus, the BAT-L, by design, was more conservative in rendering a positive TBI diagnosis.

Study Limitations
Although TRACTS Veteran self-selected for research participation, they are representative of US Veterans in terms of demographics, and other factors. 25 Additionally, DoD documentation was limited to records available in the JLV system, which included both computer records and scanned, handwritten eld notes.
Paper records within DoD that have not been scanned to the JLV system may possibly provide additional information.

Conclusions
Mild TBI is a common injury among deployed post-9/11 Veterans and is critical to clinical conceptualization and treatment. Particularly when mTBI occurs in combination with other psychiatric factors (e.g., PTSD, depression), it has been shown to worsen outcomes and complicate treatement as soliders reintegrate back into the civilian life. 21,22 Thus, it is crucial to accurately diagnose mTBI in order to aid clinical decision making and treatment. These ndings suggest that TBI, and mTBI in particular, are vastly under-documented in military DoD records. This nding is very relevant to Veteran health as well as to their potential compensation for injury. Both system-level and individual-level solutions are needed to address injury reporting among military personnel. Before deploying to a military zone, service members should be provided with education regarding TBI risk, acute symptom identi cation, and protocols to support evaluation and documentation, as this may increase the likelihood they will report a head injury when it occurs. Along with the implementation of DoD policies, quali ed clinicians should be thoroughly trained in TBI protocols and understand the selected assessment tool's strengths and limitations. Intheater assessment should include mandatory post-mission assessment using an open-ended format with queries for actue TBI symtoms and queries to rule out confounding etiologic factors that may impact symptom presentation as close to date of injury as possible. Finally, a combination of both retrospective self-report and review of in-theater medical records should be used to determine diagnoses in future con icts. Consent for publication: Not applicable.

Abbreviations
Availablity of data and materials: The data that support the ndings of this study are available upon reasonable request from the corresponding author, CBF. The data are not publicly available due to restrictions: 1) a subset of participants did not consent to share their data in a data repository and 2) the dataset contains information that could compromise the privacy of research participants.
Competing interests: The authors declare that they have no competing interests. Authors' contributions: SK and BB reviewed participant charts and were responsible coding information.
SK, AC, and JRF analyzed and interpreted the patient data. SK, CF, and AK were involved in consensus of diagnoses, and major contributors in writing the manuscript. All authors read and approved the nal manuscript