Page 69 - JSOM Spring 2026

sample (n=41) displayed moderate convergent validity with all TABLE 2 GBEV and BETS Weapon Category-by-Weapon Category
proxy measures of blast exposure (rho-range=0.595–0.672, Test-Retest Reliability
P<.001 all) (Figure 2). Test-retest reliability (n=13)
Variable ICC (95% CI) SEM
FIGURE 2 Correlation matrix and heatmap between GBEV 3,1
and proxy measures of blast exposure for convergent validity GBEV 0.576 (0.160–0.818) 39,100,000.00
analysis (n=41). Weapon Category 1 0.070 (–0.400 to 0.511) 406,000.00
Weapon Category 2 0.567 (0.148–0.813) 143.00
Weapon Category 3 0.508 (0.066–0.783) 9,940.00
Weapon Category 4 0.766 (0.474–0.906) 2,420.00
Weapon Category 5 0.186 (–0.297 to 0.593) 4,250.00
Weapon Category 1: small and medium arms
Weapon Category 2: large arms, often shoulder-fired, that can be car-
ried on a person
Weapon Category 3: artillery, missile weapon systems, or large arms
carried by vehicle, aircraft, or boat
Weapon Category 4: smaller explosives or grenades
Weapon Category 5: larger explosives or targeted explosives at close
range
GBEV = generalized blast exposure value; ICC = intraclass correlation
coefficient; SEM = standard error of measurement.

moderate convergent validity results agree with prior research
reporting similar correlation coefficients (our study vs cited
Note: Spearman’s rho correlation coefficients are presented within study) between GBEV and number of combat deployments
23
each cell. All correlations with GBEV were significant (P<.001 all). (0.617 vs. 0.53) and GBEV and CES Total (0.672 vs. 0.58).
GBEV = generalized blast exposure value; CES = Combat Exposure The poor-moderate test-retest reliability was likely due to
Scale. multiple factors, including the small sample size, the larger-
than-normal blast exposure experienced by ARSOF members,
Test-Retest Reliability and the self-report nature of the questionnaire.
Thirteen total ARSOF members were analyzed. The mean and
23
median time between post-SFAUC testing time points were For convergent validity, compared to prior research, we in-
49.7 (SD 6.5) and 50.0 (IQR 47.0–56.0) days, respectively. cluded additional proxy measures of blast exposure (such as
GBEV was on the low end of moderate test-retest reliability months deployed, months served in the military, and months
(ICC 0.576 [95% CI 0.160–0.5818]), and each weapon served in ARSOF), which strengthens the validity of GBEV.
3,1
category had an ICC range of 0.070–0.766 (Table 2). All Our sample was also 100.0% ARSOF, which is unique because
3,1
individual items within each weapons category displayed prior research included a more generalizable sample of Ser-
poor-moderate test-retest reliability (Table 3). The exception vicemembers and veterans. ARSOF members have atypical
23
was for item 1 in weapon categories 1-5 and item 2 in weapon deployment schedules compared to conventional forces, which
categories 1–4, which displayed good-excellent test-retest reli- is why they are often excluded from research investigating
ability (Table 3). Removing the three ARSOF members who re- deployment-related variables. 27,28 Therefore, our results show
ported experiencing blast exposures between the post-SFAUC that GBEV has moderate convergent validity for a highly spe-
test-retest reliability time points did not significantly alter the cialized group within the Department of Defense/Department
results (Supplementary Tables S2 and S3). of War. However, this does not mean that our proxy measures
can be used in place of the GBEV as the Spearman’s rho cor-
Exploratory Ad-Hoc Analysis relation coefficients were not perfectly correlated at a value
After seeing the poor test-retest reliability, an exploratory anal- of 1.0 (rho-range=0.595-0.672). GBEV explains additional
ysis was completed to identify if recall disparity worsened with variance due to its more direct measurement of blast exposure
higher GBEV values. To do this, a correlation (Spearman’s rho) compared to the proxy measures.
between the difference and average from both post-SFAUC
time points was conducted. The difference and average GBEV The uniqueness of the ARSOF community is also a potential
values were strongly correlated for the entire test-retest reli- reason for the poor test-retest (n=13) reliability. The origi-
ability sample (n=13, rho=0.907, P<.001) and with the three nal development of the BETS determined a threshold GBEV
ARSOF members who experienced blast between the two value of 200,000 to indicate a point at which an individual
15
post-SFAUC time points removed (n=10, rho=0.867, P=.001) was likely to report intense symptoms; however, no official
(Supplementary Figures S1A and S1B). The strong relationship threshold of blast exposure has been agreed upon via expert
remained after removing two outliers found in Supplementary consensus to determine what constitutes a clinically meaning-
Figures S1A and S1B (n=11, rho=0.845, P=.001) (Supplemen- ful difference. Only seven ARSOF members (17.1% [n=7] of
tary Figure S1C). total/ convergent validity sample [total n=41]) fell below the
15
200,000 GBEV value threshold compared to 57.3% in the
original BETS study, and our ad-hoc analysis showed that as
Discussion
mean GBEV increased, so did the discrepancy in GBEV between
The BETS displayed moderate convergent validity and overall the two post-SFAUC time points. In other words, an ARSOF
poor-moderate test-retest reliability in this sample of ARSOF member’s ability to recall their blast exposure on the BETS
personnel, thereby partially supporting our hypotheses. Our became worse the more blast exposures they experienced. The

BETS Reliability and Validity | 67