Page 66 - JSOM Summer 2019

and happiness, which cannot mitigate psychopathological Attention to precision, or the lack thereof, affected that and
processes of PTSD and suicidality. Indeed, Vie et al. affirm other studies of resilience measurement in the military. Vyas
42
that the theoretical foundations of the GAT’s resilience instru- et al. did not discuss any subscale elements of the RSES nor
16
mentation are “framed by positive psychology . . . the GAT did they explain how it measures resilience. Therefore, their
assesses positive emotions, meaning, and personal attributes assertions that resilience shares an inverse relationship with
(i.e. optimism) that contribute to a full life.” 42 PTSD symptom prevalence warrant additional study. In other
instances, investigators chose the CD-RISC to measure re-
46
Resilience-building programs in the military have millions, if silience, but we recall here that the CD-RISC is designed for
not billions, of dollars of dedicated funding and are often a therapeutic, clinical use to quantify patients’ use of coping
lodestar, useful to Secretaries of Defense and other government mechanisms through cognitive activity. Using the CD-RISC
77
leaders, in particular when they are asked to answer about to estimate resilience before and after MRT could compromise
Servicemembers’ PTSD and suicide. Comprehensive soldier results’ validity, internally and externally. When used to quan-
72
fitness was declared a success on the basis of one US Army tify resilience outside of therapeutic milieus, CD-RISC preci-
73
study in which Lester et al. cited statistically significant re- sion is questionable, especially in military sample populations
66
66
sults of MRT (effectively the precursor for CSF) in more than not undergoing active, consistent cognitive therapies and/or
10,000 Soldiers studied, but they acknowledge marginal effect psychotherapy.
sizes (e.g., 1.31% increases in emotional fitness). Lester et al.
66
used omnibus tests (e.g., regression, analyses of covariance) Several studies captured divergent elements of resilience, and
rather than using advanced multivariate methods to determine one mixed military participants with civilians despite those
45
mediating and moderating effects of covariates and confound- populations having vastly different stressors and extant sup-
ers on MRT effectiveness and Soldiers’ resilience. With such a port structures. In contrast to the GAT, in which spiritual fit-
powerful sample size, statistical significance can be detected ness is an essential element of resilience, De La Rosa et al.
45
even when clinical significance may be marginal. asserted that spirituality does not factor into resilience. In their
last sample (sample 4), they tested an abbreviated version of
Studies led by other investigators found that resilience declined the RSES, in which participant numbers were approximately
in Servicemembers after their mandatory participation in resil- 1,380 fewer than in sample 1, and heterogenized their sample
ience-building initiatives 40,46 which align with the Vaughan et population to include DoD civilians. It was this last sample
al. external evaluation that found the OSCAR program was from which they were able to show significant associations of
58
ineffectual, or at least inconclusive, in improving Servicemem- the four-item RSES to other instruments. In the abbreviated
45
bers’ mental health. Vyas et al., authors affiliated with the version, no measures of spirituality or specific social supports
16
US Navy’s OSCAR resilience program, estimated from retro- are included in the four questions, which were responded to on
spective chart reviews that building resilience would save hun- a Likert scale (e.g., “During life’s most stressful events, I tend
dreds of millions in health care costs. Vyas et al. substituted to: find a way to carry on, know I will bounce back, learn im-
16
incidence of PTSD and depression as “functions of resilience” portant and useful life lessons, practice ways to handle it better
without visible justification. They also used adjusted odds next time”). Though likely appealing in utility and ease of
45
ratios (AORs) extensively, methods accompanied by notable administration, this reductionist approach to Servicemembers’
limitations and restrictions outlined in the very manuscript on resilience appears flawed in its approach to holistic assessment.
odds ratios and likelihood estimations those authors followed
to execute their analyses. 74 Multiple quantitative studies of military resilience focused on
the psychometric properties of the instruments of measure-
First, logistic regression with AORs are used for dichotomous ment, to the detriment of reporting actual numerical levels of
outcome variables, such as lived/died, treatment success/fail- resilience. Similarities were found in empirical and/or theoreti-
ure, and so forth —not for outcome variables sought by Vyas cal literature in which resilience was deemed a process or an in-
75
et al. Second, there was no report of which statistical program born personality trait, though contributory constructs of what
16
was used for analysis, nor was it noted if binomial or ordinal constitutes resilience differed significantly. Other evidence con-
logistic regression was used. Third, their statistical reporting tradicted those findings in asserting that resilience can be devel-
16
aligned with analyses of variance and bivariate correlation but oped and augmented by using protective factors. Similarities in
not regression used to predict phenomena. For example, the protective factor elements were found, though methodological
study reported F and p values, but not other measures, such rigor was inconsistent in quantitative and qualitative studies
as proportional odds, full likelihood-ratio testing, compar- examined. Overall, the literature lacked consensus on resil-
ing a fitted model to a model with variant parameters, χ re- ience, particularly in its measurement within military popula-
2
sults, goodness-of-fit testing, final model versus intercept-only tions at risk; most studies proposed that resilience is augmented
model, linearity to logit of dependent variables (e.g., Box- by protective resources, but significant divergence was evident
Tidwell procedure), any Bonferroni corrections, standardized in pinpointing which factors were specifically germane.
residuals, Nagelkerke R for variance, sensitivity, specificity,
2
or positive/negative predictive values. Cost calculations were Qualitative studies of military resilience were few and mostly
75
similarly unclear: The Tanielian and Jaycox 2008 report on tangential in examining Servicemembers’ resilience, with the
traumatic brain injury and PTSD was used by Vyas et al. to exception of studying trans-vets and Muslim Servicemem-
76
60
project costs through multiple micromodels and scenarios of bers directly. Though labeled mixed-methods, the Scott et
64
care. It is not clear how Vyas et al. determined parameters al. study does not appear to have followed accepted tenets in
16
16
59
for determining cost savings with point-increases on the RSES executing mixed-methods methodologies. Omission of quanti-
when (1) stress responses are not identical to resilience and (2) tative results prevented reflexivity and triangulation with their
those investigators did not specify directly which of Tanielian qualitative results and stymies future attempts at replication
59
and Jaycox’s many robust cost models they used. by other researchers.
64 | JSOM Volume 19, Edition 2 / Summer 2019