Abstract
As the construct of moral injury has gained increased conceptual and empirical attention among military personnel and veterans, preliminary attempts to operationalize and measure the construct have emerged. One such measure is the Moral Injury Event Scale (MIES). The aim of the current study was to further evaluate the MIES’s psychometric properties in two military samples: a clinical sample of Air Force personnel and a nonclinical sample of Army National Guard personnel. Exploratory and confirmatory factor analyses across both samples supported a three-factor solution: transgressions by others, transgressions by self, and betrayal. Transgressions-Others was most strongly associated with posttraumatic stress; Transgressions-Self was most strongly associated with hopelessness, pessimism, and anger; and Betrayal was most strongly associated with posttraumatic stress and anger. Results support the construct validity of the MIES, although areas for improvement are indicated and discussed.
Military personnel often experience morally ambiguous situations that require them to make decisions and/or act in ways that can lead to inner turmoil and personal conflict. For instance, combatants may be required to use violence or aggression toward others, such as killing, under conditions in which threat levels and hostile intentions are not clear or easily determined (e.g., Is the approaching vehicle filled with explosives that could kill you and others, or is it just a father driving to pick up his child from school?). Although violence and killing are prescribed in war, in most other contexts these very same actions may be considered illegal, unethical, and/or immoral (Drescher et al., 2011; Litz et al., 2009). For example, research indicates that approximately one in three soldiers and marines deployed to Iraq have killed an enemy combatant (Hoge et al., 2004), suggesting that moral injury may be relatively common for U.S. military personnel and veterans. Although moral injury is most often associated with violence and aggression within the context of combat, military personnel can also experience inner turmoil secondary to nonviolent events, such as exposure to dead bodies or human remains, reported by 65% of Iraq and Afghanistan veterans (Hoge et al., 2004), and/or seeing wounded civilians and being unable to assist, reported by 60% of Iraq and Afghanistan veterans (Hoge et al., 2004). The potential conflicts between these experiences and a service member’s moral standards can lead to lasting emotional distress and inner turmoil for military personnel, a situation that has been termed moral injury (Litz et al., 2009; Shay, 2014).
Morally injurious experiences include those in which the individual perpetuates, fails to prevent, bears witness to, or learns about acts that transgress deeply held moral beliefs and experiences (Litz et al., 2009). According to Litz et al., moral injury requires an act of perceived transgression that violates or contradicts the individual’s code of conduct or rules about “right” versus “wrong” and can occur as a consequence of one’s own actions or of someone else’s actions. This can contribute to inner conflict and emotional distress but only if the individual is aware of a discrepancy between the experience and his or her own moral code, which in turn leads to negative self-attributions, guilt, and shame. These cognitive–affective states are hypothesized to lead to social withdrawal and the emergence of trauma-related symptoms including self-condemnation, difficulties with forgiveness, self-handicapping, and demoralization. The emergence of guilt, shame, and anxiety is facilitated by idiographic factors such as shame proneness and/or neuroticism. In contrast, factors such as self-forgiveness and supportive networks who offer understanding and forgiveness can block the pathway from perceived transgression to symptoms of posttraumatic stress disorder (PTSD), such as intrusive memories and nightmares, social withdrawal, and inability to experience positive emotions.
The experience of moral dilemma is not limited to the use of aggression and violence, however, and is not limited to those military personnel assigned to traditional combat roles. For instance, military medical personnel are often required to prioritize the care of some wounded individuals over others, which may be experienced as “choosing who lives and who dies.” Furthermore, military leaders and commanders are routinely required to make decisions about missions that can lead to the injury and/or death of their subordinates. Military personnel also passively experience or witness morally ambiguous or troubling events in the line of duty, such as confronting intense human suffering and witnessing the results of violence and atrocity, whether such events are associated with combat or not. Research among noncombatant military personnel (e.g., medical, logistics, and support), for example, suggests very high rates of exposure to these events while deployed (Peterson, Wong, Haynes, Bush, & Schillerstrom, 2010). Such experiences can be experienced as traumatic, resulting in psychological and behavioral responses that are very similar in nature to PTSD.
Over the past few decades, PTSD has predominantly been conceptualized as a fear-based anxiety disorder, but more recent conceptual and empirical work suggests that the requirement for a fear-related response may not be necessary for an accurate diagnosis of PTSD (Friedman et al., 2011). Among military personnel and veterans, exposure to stressors and experiences such as atrocities, loss of close friends, and killing is associated with PTSD even though the experiences do not necessarily entail life-threatening situations (Currier & Holland, 2012; King, King, Gudanowski, & Vreven, 1995; Maguen et al., 2010). Furthermore, many military personnel and veterans with PTSD do not report fear and anxiety as their primary features of psychological distress and impairment but rather describe guilt and shame, negative changes in ethical attitudes and behavior, changes in spirituality, difficulties with forgiveness, and reduced ability to trust others as being particularly problematic (Drescher et al., 2011; Vargas, Hanson, Kraus, Drescher, & Foy, 2013).
Difficulties in trust may be especially likely subsequent to traumas in which the individual perceives that he or she has been betrayed by others. Perceived betrayal has been shown to uniquely contribute to the emergence of PTSD apart from the level of physical violence and/or perceived dangerousness of the situation. According to Freyd, DePrince, and Gleaves (2007), betrayal trauma occurs “when the people or institutions on which a person depends for survival violate that person in a significant way” (p. 297) and is assumed to be more strongly associated with stressors characterized by violations of trust (e.g., interpersonal violence, sexual assault) as compared to other stressors that are unrelated to trust (e.g., natural disasters, sports injuries, or accidents). Supporting this perspective is preliminary research indicating that betrayal is associated with overall PTSD symptom severity and is a better predictor of reexperiencing, avoidance, and numbing symptoms than the perceived level of threat (Kelley, Weathers, Mason, & Pruneau, 2012). These findings converge with conceptual work hypothesizing that moral injury should be associated with the reexperiencing, avoidance, and numbing symptoms of PTSD but may not necessarily be associated with physiological arousal (Litz et al., 2009; Shay, 2014). As applied to military personnel and veterans, betrayal trauma may be a more salient feature of the trauma reaction among those who have experienced events such as military sexual trauma, in which a service member is sexually assaulted by another service member, or “green on blue” events, in which a presumed ally (e.g., Iraqi Police, Afghan National Army, local villagers) sabotages or attacks U.S. military personnel.
Despite the frequency with which guilt, shame, self-deprecation, and difficulties with trust are endorsed among trauma survivors, they have not traditionally been included as diagnostic criteria for PTSD (Drescher et al., 2011), although the accumulating evidence supporting this broader perspective of trauma has led to recent changes in the psychiatric classification of PTSD, most notably in the movement of the diagnosis from the anxiety disorders section to the trauma- and stressor-related disorders section of the Diagnostic and Statistical Manual of Mental Disorders (5th ed., American Psychiatric Association, 2013). In addition, persistent cognitive–affective states including guilt, shame, and self-deprecation were added to the diagnostic criteria for PTSD in a new, separate symptom cluster referred to as “negative alterations in cognitions and mood.”
Although the notion of moral injury certainly is not new, it is only recently that attempts have been made to operationalize and measure it as a distinct psychological construct. Given the relative early stage of conceptual development and empirical investigation of moral injury, however, it is not yet fully known how moral injury is related to PTSD. In an attempt to operationalize and better understand the construct of moral injury, Nash et al. (2013) recently developed the Moral Injury Event Scale (MIES) in order to measure exposure to potentially morally injurious events. Out of an original 11 items generated by a team of subject matter experts, preliminary psychometric properties from two large samples of U.S. Marines in infantry units were reported, with results of exploratory (n = 533) and confirmatory factor analyses (n = 506) yielding a nine-item measure with two factors that demonstrated good fit: perceived transgressions and perceived betrayals. Perceived transgressions included six items assessing internal conflict due to the actions of oneself and of others (both omission and commission), whereas perceived betrayals included three items assessing internal conflict due to perceived duplicity or deceit by military leaders, fellow service members, and individuals external to the military. Both scales showed good internal consistency (αs ≥ .82) and temporal stability over the course of 3 months. Factor correlations with measures of psychological distress (e.g., depression, PTSD symptoms, anxiety), social support, and unit cohesion ranged from small to moderate, suggesting that MIES items were measuring a construct that was relatively distinct from other indicators of psychopathology.
Results of these preliminary psychometric evaluations were therefore positive. Nonetheless, Nash et al. (2013) noted that because their sample included only male marines assigned to infantry units, further evaluation of the MIES was needed in samples characterized by (a) both men and women, (b) service members from other branches of service, and (c) service members from professions other than combat arms. To this end, the current study addresses these three limitations by conducting a psychometric evaluation of the MIES in a clinical sample of Air Force personnel and a sample of U.S. Army National Guard personnel. Furthermore, existing research focused on the measurement of moral injury in the military has primarily focused on combat-related experiences. Military personnel also experience morally troubling experiences outside the context of combat, such as perceived betrayal by peers or military leaders, interpersonal violence (e.g., domestic violence, sexual assault), and exposure to human remains and devastation in the aftermath of natural disasters. No studies have yet examined MIES responses with respect to these different potential sources of moral injury, however. In light of these knowledge gaps, in the present study, we sought to answer the following questions:
What is the factor structure of the MIES across diverse military samples?
What are the descriptive statistics of MIES scores across diverse military samples?
How do MIES scores correlate with other indicators of psychopathology?
How do MIES scores correlate with different types of life stressors?
Method
Participants
Sample 1
Participants included 151 active duty Air Force personnel (63.8% male, 36.2% female) ranging in age from 20 to 54 years (M = 34.12, SD = 8.41) who were seeking outpatient mental health treatment at two military treatment facilities in the Western and Southern United States. Racial distribution was 66.9% Caucasian, 20.5% African American, 2.0% American Indian, 1.3% Asian, 1.3% Pacific Islander, 6.0% other, and 2.0% unreported; 9.3% additionally reported Hispanic or Latino ethnicity. Rank distribution was 22.3% junior enlisted (E1-E4), 41.2% noncommissioned officer (E5-E6), 16.2% senior noncommissioned officer (E7-E9), and 22.0% officer. Just over half (57.3%) of participants had deployed at least once to Iraq or Afghanistan, and 25.2% endorsed being exposed to “direct combat.”
Sample 2
Participants included 935 U.S. military personnel (82.3% male, 17.7% female) ranging in age from 17 to 61 years (M = 27.05, SD = 8.11). The sample was drawn largely from the Army National Guard (84.0%), with 34.6% participating during demobilization from Afghanistan. Racial distribution was 57.4% Caucasian, 24.3% African American, 4.1% Hispanic/Latino, 1.8% Asian or Pacific Islander, 1.0% American Indian, 3.6% other, and 7.8% unreported. Over two thirds (68.9%) of participants had deployed at least once to Iraq or Afghanistan. Only 8.7% of the sample had job duties classified as “noncombatant” (e.g., medical personnel). Rank was not assessed in Sample 2.
Procedures
Sample 1
Patients receiving outpatient mental health care were invited to participate in the current study by clinic staff, who provided them with an information sheet and briefly summarized the study’s purpose and procedures. Interested patients signed an informed consent document and were given a survey packet that they completed anonymously in the clinic waiting room. Participants returned completed survey packets to a clinic staff member, who then placed the surveys in a centralized box located in a secured area of the clinic accessible only to approved staff members. Data were shipped to the National Center for Veterans Studies at the University of Utah for entry into an electronic database and subsequent analysis. Approval for the current study was obtained from the institutional review board located at Wright-Patterson Air Force Base.
Sample 2
Soldiers participated in groups of up to 25, reporting to a large classroom setting in a Joint Forces Training Center in the Southern United States. Each soldier was seated at a laptop computer provided by the research team, connected to a nongovernment wireless network also provided by the research team in an effort to reduce concerns regarding confidentiality among military participants. After study procedures were described, informed consent was obtained from soldiers interested in participating. The protocol consisted of a series of self-report questionnaires, followed by completion of an implicit association task. Future assessments will take place remotely at 6-, 12-, and 18-month follow-ups. Each soldier was compensated with $20. Due to Department of Defense regulations in place at the time of data collection, soldiers who were classified as active duty were ineligible for payment. Data were missing from 253 of the 935 participants (26.5%) due to military training requirements, which interrupted data collection for many participants. There were no significant demographic differences between those participants with complete data and those participants with missing data, however. Little’s test for missingness further indicated that data were missing completely at random, χ2(231) = 239.27, p = .340. Missing data were therefore handled using maximum likelihood estimation. Approval for the current study was obtained from the institutional review board located at the University of Southern Mississippi and the Human Research Protection Office of the U.S. Army Medical Research and Materiel Command.
Instruments
Sample 1
Moral Injury Event Scale
The MIES (Nash et al., 2013) is a self-report scale comprised of nine statements that ask about exposure to perceived transgressions committed by the respondent and/or others, and perceived betrayals by other military and nonmilitary individuals. Respondents indicate how much they agree with each statement on a scale ranging from 1 (strongly agree) to 6 (strongly disagree), with higher scores indicating greater moral injury. As noted previously, the MIES has demonstrated good preliminary factor structure and reliability (αs = .82-.89) and demonstrates only small to moderate correlations with other indicators of psychopathology, indicating that it is a relatively distinct construct.
Posttraumatic Stress Disorder Checklist–Military Version (PCL-M)
The PCL-M (Weathers, Litz, Herman, Huska, & Keane, 1993) is a self-report measure used to assess the severity of PTSD symptoms. The PCL-M directs respondents to indicate the severity with which each symptom of PTSD has been experienced within the past 30 days. The scale has demonstrated excellent reliability and diagnostic utility for PTSD (Weathers et al., 1993). Cronbach’s alpha for the PCL-M in the current sample was .97.
Patient Health Questionnaire–9 (PHQ-9)
The PHQ-9 (Kroenke, Spitzer, & Williams, 2001) is a self-report measure used to assess the severity of depression symptoms. The PHQ-9 directs respondents to indicate the frequency with which each symptom of major depressive disorder within the past 2 weeks on a scale ranging from 0 (not at all) to 3 (nearly every day). The scale has demonstrated excellent reliability and diagnostic utility for major depressive disorder (Kroenke et al., 2001). Cronbach’s alpha for the PHQ-9 in the current sample was .92.
Future Dispositions Inventory (FDI)
The FDI (Osman et al., 2010) is a self-report measure used to assess the severity of two dimensions of future-oriented thinking: negative focus (i.e., hopelessness and pessimism) and positive focus (i.e., optimism and hope). The negative focus subscale consists of eight items (e.g., “I worry that things will never go well for me no matter what I do,” I doubt whether things will ever get better for me in life,” “I fear that I will run into more difficulties in the years ahead”) that respondents rate on a 5-point Likert-type scale ranging from 1 (not at all true) to 5 (extremely true). The positive focus subscale consists of eight items (e.g., “I expect things to turn out better for me in life,” “I expect to enjoy the results or outcomes of all my hard work in life,” “I expect to be happier and more content with my life”) with the same scoring scheme as the negative focus subscale. Both scales are reliable (>.83); correlate strongly in the opposite directions with measures of hopelessness, adaptive coping, and psychological symptoms; and can differentiate between suicidal and nonsuicidal groups (Osman et al., 2010). Cronbach’s alpha for the FDI within the current sample was .93 for both scales.
Personal Feelings Questionnaire–2 (PFQ-2)
The PFQ-2 (Harder, Rockart, & Cutler, 1993) is a self-report measure used to assess the severity of guilt and shame. The PFQ-2 directs respondents to indicate the frequency with which they have experienced 22 different emotional or cognitive states on a scale ranging from 0 (never) to 4 (continuously or almost continuously). Ten items (e.g., embarrassment, feeling ridiculous, feeling stupid) load onto the shame subscale, six items (e.g., mild guilt, worry about hurting or injuring someone, regret) load onto the guilt subscale, and six items are used as filler items. Both subscales have good reliability and correlate well with other measures of guilt and shame (Harder et al., 1993). Cronbach’s alphas for the PFQ-2 guilt and shame subscales in the current sample were .85 and .86, respectively.
Sample 2
Moral Injury Event Scale
The MIES was described above for Sample 1.
Posttraumatic Stress Disorder Checklist–Military Version
The PCL-M was described above for Sample 1. Cronbach’s alpha for the PCL-M in the current sample was .93.
Patient Health Questionnaire–9
The PHQ-9 was described above for Sample 1. Cronbach’s alpha for the PHQ-9 in the current sample was .89.
Generalized Anxiety Disorder 7-Item Scale (GAD-7)
The GAD-7 (Spitzer, Kroenke, Williams, & Lowe, 2006) is a self-report measure used to assess the severity of the symptoms of generalized anxiety disorder defined in the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev., American Psychiatric Association, 2000). Participants rate the frequency with which they have experienced each symptom during the past 2 weeks on a scale ranging from 0 (not at all) to 3 (nearly every day). The scale has established reliability and validity (Spitzer et al., 2006). Cronbach’s alpha for the GAD-7 in the current sample was .91.
State-Trait Anger Expression Inventory–2 (STAXI-2)
The STAXI-2 (Spielberger, 1999) is a self-report measure used to assess the experience, expression, and control of anger. For the current study, only the state anger subscale was used, which focuses on the intensity of anger as an emotional state during the past 2 weeks. Participants rate the intensity of 15 items on a 4-point scale, with higher scores indicating more intense anger. The scale has very good reliability and validity (Spielberger, 1999). Cronbach’s alpha for the STAXI-2 state anger scale in the current sample was .96.
Beck Hopelessness Scale (BHS)
The BHS (Beck, Weissman, Lester, & Trexler, 1974) is a 20-item self-report measure that assesses past week negative expectancies about one’s self and future. Items are scored in a true/false format, with a total score ranging from 0 to 20. Extensive research has reported on the reliability and validity of the scale (e.g., Steed, 2001). All studies funded by the Military Suicide Research Consortium include three items from the BHS as part of a pool of common data elements. Responses on these items were scored as 0 (false) or 1 (true), with one item reverse scored (“I happen to be particularly lucky, and I expect to get more of the good things in life than the average person”) Internal consistency in this sample was .53, which is lower than typical for the BHS, most likely due to the low number of items.
Deployment Risk and Resilience Inventory (DRRI)
The DRRI (King, King, & Vogt, 2003) is a series of self-report scales that assess deployment-related risk and protective factors that can contribute to military personnel and veterans’ psychological and physical health. For the present study, the Prior Stressors, Combat Experiences, and Aftermath of Battle scales were used. The Prior Stressors scale assesses exposure to potentially traumatic events prior to deployment, such as physical and sexual assault, crime, and highly stressful life events. The Combat Experiences scale assesses exposure to common warfare-related experiences such as being fired on, killing, and witnessing death and injury. The Aftermath of Battle scale assesses exposure to the consequences of combat such as human remains, caring for wounded individuals, and devastation. Each scale lists a series of events and provide a yes/no response format. Affirmative responses on each scale can be summed to provide an overall metric of exposure within each domain, with higher scores indicating greater exposure. Internal consistency estimates for the three scales range exceed .75 in both Gulf War I and Iraq/Afghanistan veterans (Vogt, Proctor, King, King, & Vasterling, 2008). In the current sample, internal consistency estimates for all three scales exceeded .80.
Results
Factor Analyses for the Moral Injury Events Scale
We first conducted a confirmatory factor analysis (CFA) using Sample 1 to examine the proposed two-factor latent structure of the MIES previously reported by Nash et al. (2013). The CFA was performed with the Mplus 6.12 software (Muthén & Muthén, 1998-2011) using maximum likelihood estimation. Adequacy of model fit was assessed using the standardized root mean square residual (SRMR) <.08 supplemented with the comparative fit index (CFI) >.90 (Hu & Bentler, 1999). Although the Tucker–Lewis Index (TLI) and root mean square error of approximation (RMSEA) were calculated and are reported, they were not used as fit indices in Sample 1 because these scores have been found to overreject correct models in samples with less than 200 participants, while the combined SRMR and CFI have much lower error rates in smaller samples (Chen, Curran, Bollen, Kirby, & Paxton, 2008; Hu & Bentler, 1999). The variance of each factor was fixed to 1.0 for identification purposes, and error variances were not allowed to be correlated in any of the analyses. Results of the CFA indicated that although item–factor loadings were all greater than .350 (see Table 1), overall fit estimates for the Sample 1 data were poor: SRMR = .125, CFI = .779.
Factor Loadings From Factor Analyses in Sample 1 and Sample 2.
Note. CFI = comparative fit index; EFA = exploratory factor analysis.
Values in bold indicate the factor on which the item loaded.
We therefore conducted an exploratory factor analysis (EFA) using Sample 1 to identify the MIES’s latent factor structure. The Kaiser–Meyer–Olkin measure of sampling adequacy (.795) and Bartlett’s test of sphericity (p < .001) indicated the matrix was suitable for factor analysis. The Mplus 6.12 software was used for the EFA because the program allows for the specification of a series of successive factor solutions. Maximum likelihood estimation was used since it performs better than other estimation methods even under conditions of nonnormal distribution (Olsson, Foss, Troye, & Howell, 2000); when analyses were repeated using robust maximum likelihood estimation, results did not change. The geomin (oblique) rotation procedure was used because moderate correlations were expected among the derived factor solutions. Adequacy of model fit was assessed using the same criteria listed previously for the CFA. Because we intended to compare the relative fit for several competing models, the Akaike information criterion (AIC) and Bayesian information criterion (BIC) values were additionally used as indicators of model fit, with smaller values indicating better fit. Parallel analysis was conducted using the procedures recommended by Glorfeld (1995), with results suggesting a three-factor solution was optimal.
Table 2 shows the fit estimates for the one-, two-, and three-factor solutions after the geomin rotation. As can be seen, the three-factor solution attained adequate model fit to the sample data (SRMR = .024, CFI = .922) and demonstrated better fit relative to the one- and two-factor solutions, although the fit was not optimal by more rigorous contemporary standards (e.g., CFI < .95), most likely due to the small sample size. Factor 1 correlated moderately with Factor 2 (r = .43) and with Factor 3 (r = .57). Factor 2 correlated low with Factor 3 (r = .35). Consistent with the content of the items, Factor 1 was labeled “Transgressions-Others,” Factor 2 was labeled “Transgressions-Self,” and Factor 3 was labeled “Betrayal.” Results of the item–factor compositions are presented in Table 1. Each item loaded on a single factor higher than the .320 level recommended by Tabachnick and Fidell (2001). Overall, the three-factor solution explained 75.6% of the observed total variance.
Fit Statistics for an Exploratory Factor Analysis (Sample 1) and a Confirmatory Factor Analysis (Sample 2).
Note. CFI = comparative fit index; EFA = exploratory factor analysis; TLI = Tucker–Lewis Index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; AIC = Akaike information criterion; BIC = Bayesian information criterion.
We next conducted a CFA using Sample 2 to replicate the three-factor solution. The CFA procedures described above for Sample 1 were repeated. Results of the CFA indicated that the three-factor solution yielded item–factor loadings exceeding .759 (Table 1) and excellent fit to the data (see Table 2): SRMR = .056, CFI = .962. We compared these fit statistics to the fit statistics derived from the two-factor model proposed by Nash et al. (2013) and the fit statistics from a one-factor solution. As can be seen in Table 2, the three-factor solution demonstrated the best overall fit to the data.
Internal Consistency Estimates
Internal consistency estimates (Cronbach’s alphas) for each of the three scales were good across Sample 1 and Sample 2: .79 and .79 for Transgressions-Others, .96 and .94 for Transgressions-Self, and .83 and .89 for Betrayal. Item intercorrelations for the three factors are displayed in Table 3. As can be seen, item intercorrelations ranged from .60 to .65 (M = .63) for Transgressions-Others, from .78 to .91 (M = .81) for Transgressions-Self, and from .46 to .94 (M = .68) for Betrayal.
MIES Item Intercorrelations Across Two Military Samples.
Note. MIES = Moral Injury Event Scale.
Descriptive Statistics for MIES Scores
Because of the unequal number of items retained on each factor, mean scores for each scale were calculated. Sample descriptive statistics for the MIES scale scores are presented in Table 4. Scores were significantly higher on all three scales in Sample 2 as compared to Sample 1: Transgressions-Others, U = 4.268, p < .001; Transgressions-Self, U = 13.366, p < .001; and Betrayal, U = 9.029, p < .001. In terms of skew, in Sample 1 the Transgressions-Others and Betrayal scores were approximately normally distributed (skew = −0.37 and 0.06, respectively) but the Transgressions-Self scores were positively skewed (0.61). On all three scales, a score of 1 (the lowest possible score) was the most frequent score: 18.8% of Transgressions-Others scores, 41.1% of Transgressions-Self scores, and 22.5% of Betrayal scores). In contrast, Sample 2 showed substantial negative skew on all three MIES scales (Transgressions-Others, −0.49; Transgressions-Self, −1.17; Betrayal = −0.82). On all three scales, a score of 6 (the highest possible score) was the most frequent score: 38.8% of Transgressions-Other scores, 54.3% of Transgressions-Self scores, and 47.8% of Betrayal scores had a value of 6.
Scale Internal Consistency Estimates and Descriptive Statistics for Men and Women.
Mann–Whitney U test.
For both samples, there were no differences in means scores between men and women for any of the scales, although a nonsignificant trend was found between men and women in Sample 2 on the Betrayal scale (Mann–Whitney U = −1.820, p = .069), with men scoring somewhat higher than women (see Table 3).
Associations of the MIES Scores With Indicators of Psychological Distress
To explore the associations of the three identified MIES scale scores with other psychological constructs, we used multiple linear regression. Due to the considerable negative skew in Sample 2, a log transformation was conducted prior to data analyses for all three MIES scales, which reduced the severity of skew (Transgressions-Others, −0.01; Transgressions-Self, 0.70; Betrayal, 0.36). Each MIES scale was entered as a separate criterion variable. For Sample 1, the following indicators of psychological distress were entered as predictor variables: posttraumatic stress, depression, hopelessness, pessimism, optimism, guilt, and shame. For Sample 2, the following indicators of psychological distress were entered: posttraumatic stress, depression, hopelessness, anxiety, and state anger. Results are summarized in Table 5.
Summary of Results From the Bivariate and Multiple Regression Analyses Predicting MIES Scale Scores.
Note. MIES = Moral Injury Event Scale.
Log-transformed.
p < .05. **p < .01.
The coefficients of determination and zero-order correlations of MIES scores with predictor variables were generally small across both samples, suggesting that only a small proportion of the variance in MIES scores were accounted for by other indicators of psychological distress (R2s < .11). A statistically significant positive association of posttraumatic stress symptoms with Transgressions-Others was observed in both samples (Sample 1: β = .353, p = .004; Sample 2: β = .160, p = .002). Transgressions-Others also had a statistically significant negative association with depression (β = .388, p = .004) and a positive association with pessimism (β = .236, p = .033) in Sample 1. In terms of Transgressions-Self scores, a significant positive association with pessimism existed in Sample 1 (β = .227, p = .043) and a significant positive association with hopelessness existed in Sample 2 (β = .156, p < .001). Transgressions-Self also showed a significant positive association with anger in Sample 2 (β = .118, p = .014). In terms of Betrayal scores, a significant positive association with posttraumatic stress existed in both samples (Sample 1: β = .244, p = .050; Sample 2: β = .137, p = .006). Betrayal scores also had a significant positive association with anger in Sample 2 (β = .160, p = .001).
To further understand how MIES scores were related to PTSD symptoms, follow-up analyses were conducted to examine the associations of MIES scores with reexperiencing, avoidance, numbing, and hyperarousal symptoms. Results are summarized in Table 6. In Sample 1, Transgressions-Others scores were positively associated with the reexperiencing (β = .198, p = .032) and avoidance (β = .183, p = .047) clusters, and Betrayal scores were positively associated with the reexperiencing cluster (β = .164, p = .048). In Sample 2, Transgressions-Others scores were positively associated with the reexperiencing (β = .119, p = .007), numbing (β = .152, p = .005), and hyperarousal (β = .120, p = .021) clusters, and Betrayal scores were positively associated with the reexperiencing (β = .109, p = .011) and numbing (β = .182, p = .001) clusters. Transgressions-Self scores were positively associated with the numbing cluster in both samples (βs > .101, ps < .049).
Standardized Regression Coefficients (i.e., Beta Weights) Testing the Association of MIES Scores With PCL Symptom Cluster Scores.
Note. MIES = Moral Injury Event Scale; PCL = Posttraumatic Stress Disorder Checklist.
p < .05. **p < .01. ***p < .001.
Associations of the MIES Scales With Trauma and Stressor Exposure
Pearson correlations were calculated to examine the associations of the MIES scores with deployment-related and non–deployment-related stressors among participants in Sample 2 (similar data were not available for Sample 1). The log-transformed MIES scores were correlated with the DRRI’s Predeployment Stressors (M = 2.30, SD = 2.66), Combat Experiences (M = 2.83, SD = 3.47), and Aftermath of Battle (M = 2.46, SD = 3.68) scores. More than two thirds of the sample (69.7%) endorsed at least one item from the Predeployment Stressors scale, 72.9% endorsed at least one item from the Combat Experiences scale, and 55.7% endorsed at least one item from the Aftermath of Battle scale, suggesting that a considerable majority of participants had experienced at least one life event that is consistent with a Criterion A event for PTSD. Mean scores for the three DRRI scales were comparable to other military samples (Heron, Bryan, Dougherty, & Chapman, 2013; King et al., 2003). Transgressions-Others scores had significant positive associations with all three DRRI scale scores: Predeployment Stressors (r = .200, p < .001), Combat Experiences (r = .093, p = .012), and Aftermath of Battle (r = .189, p = .001). Transgressions-Self scores had a significant positive association with Combat Experiences only (r = .125, p = .028), and Betrayal scores had a significant positive association with Predeployment Stressors only (r = .217, p < .001).
Exploratory follow-up correlation analyses were subsequently conducted to determine if MIES scores were correlated with specific deployment-related experiences and life stressors in conceptually meaningful ways. Results are summarized in Table 7 along with the proportions of participants reporting each experience. In general, Transgressions-Others and Betrayal scores were positively correlated with noncombat events in which participants were the target or recipient of harm by others (e.g., being robbed, emotionally mistreated, or abused/assaulted). Transgressions-Others scores also tended to be positively correlated with combat-related events characterized by being the target or recipient of harm (e.g., being attacked or coming under fire). Transgressions-Self scores, in contrast, tended to be positively correlated with events in which participants were the agent of harm toward others (e.g., firefights, killing enemy combatants).
Proportion of Soldiers in Sample 2 Reporting Exposure to Various Potentially Traumatic Events and Correlation Coefficients of MIES Scale Scores With Each Event.
Note. MIES = Moral Injury Event Scale; DRRI = Deployment Risk and Resilience Inventory.
Bold values indicate correlations of MIES scales with DRRI full-scale scores
Discussion
The specific aims of the current study were to further evaluate the psychometric properties of the MIES, which was conceptualized as a multidimensional measure of moral injury, in two diverse military samples: a clinical psychiatric outpatient sample of Air Force personnel and a nonclinical sample drawn largely from the Army National Guard. In both samples, results of our exploratory and confirmatory factor analyses did not support the two-factor solution previously identified by Nash et al. (2013) but rather supported a three-factor solution that identified two separate dimensions of perceived transgressions—transgressions committed by others (Items 1 and 2) and transgressions committed by the self (Items 3-6)—as well as a factor consistent with perceived betrayals by others (Items 7-9). Our results therefore replicate Nash et al.’s (2013) previous findings of a distinction between transgressions and betrayal but suggest that differentiating between the agent of perceived transgressions (i.e., self vs. other) may also hold important conceptual and empirical value. When considering the results of Nash et al.’s (2013) previous EFA and CFA, it is notable that although a single perceived transgressions factor was identified, Items 1 and 2 of this scale yielded relatively smaller item–factor loadings on the perceived transgressions scale than Items 3 through 6, similar to the results of our CFA in Sample 1, suggesting their analyses may also have been influenced by a differentiation in the agent of transgressions.
Two important limitations of the three-factor solution identified in the current study warrant discussion prior to further consideration of the current study’s implications. First, the similarity in wording for the MIES’s items could give rise to what Cattell (1978) has referred to as a “bloated specific,” which can lead to high internal consistency estimates at the cost of narrow construct measurement. According to Kline (1979), item intercorrelations that exceed .70 may suggest that a scale is too narrow or specific, which can limit scale validity. With respect to the MIES, the current results suggest that this issue may be especially relevant to the Transgressions-Self factor, which showed item intercorrelations ranging from .78 to .90. Future work to improve the MIES and the measurement of moral injury more generally might therefore focus on the development and testing of items that assess a broader range of potentially morally injurious experiences. For example, the number of items that assess the witnessing of perceived transgressions committed by others could be increased. Increasing the number of items for the MIES would address the second major limitation of the best fitting solution identified in this study: the two-item Transgressions-Others factor. The two items that comprise the Transgressions-Others factor may therefore reflect a high correlation coefficient more so than a distinct factor.
Scores on each MIES scale satisfied the criterion for adequate internal consistency reliability, with all reliability estimates being ≥.79 in both samples. In the current study, Army National Guard personnel (Sample 2) scored significantly higher on all three MIES scales than Air Force psychiatric outpatients (Sample 1). MIES scores for Army National Guard personnel were negatively skewed, whereas scores for Air Force personnel showed an approximately normal distribution or positive skew, suggesting Army personnel were much more likely to report distress regarding the actions of others and themselves, and were much more likely to have felt betrayed by others. This could be due to differences between the two samples in terms of exposure to potentially morally injurious events. For instance, 25% of participants in Sample 1 reported direct combat exposure as compared to nearly three quarters of Sample 2. Because we did not assess for combat intensity and trauma exposure in both samples, however, this conclusion should be considered with caution until additional studies can adequately assess and control for such factors. An alternative explanation may be related to the timing of data collection relative to participants’ deployment cycle. In Sample 1, data were collected from personnel receiving routine treatment at an outpatient mental health clinic, whereas in Sample 2, over one third of the participants had just returned from Afghanistan. Although morally injurious events are not restricted to deployment or combat, deployments are nonetheless associated with increased likelihood for such exposures. As a result, there may have been a disproportionately larger number of participants in Sample 2 with very recent exposure to morally injurious events, which could mean these events were more salient when participants were responding to the scale.
Yet another possibility may be related to differences in social context associated with active duty versus National Guard personnel. Specifically, as full-time military personnel, active duty service members are typically surrounded by the military community across multiple domains of life (e.g., employment, community services, family life, neighborhoods). National Guard personnel, in contrast, are much more likely to be geographically separated from their military units and their military peers, have much less access to military-focused support services, and are much more likely to live in communities with a relatively low density of military personnel and veterans. National Guard personnel may therefore have fewer opportunities to openly share their experiences with others, which could lead to greater internal distress. Additional research is needed to clarify any differences that might exist between active versus National Guard personnel and to examine how social support and context might influence the experience of psychological distress following exposure to morally injurious experiences. Despite the differences in the response distributions between the two diverse military samples used in this study, the consistency of the three-factor solution identified across both samples lends confidence to our findings.
Consistencies across both samples were also observed in terms of the correlations of MIES scores with other measures of psychological distress. Transgressions-Others scores were most strongly associated with posttraumatic stress symptoms, for instance, suggesting some overlap with traditional conceptualizations of PTSD. The relationship of Transgressions-Others scores with psychological distress aligns with previous research among Vietnam veterans and Iraq/Afghanistan veterans, for whom exposure to atrocities committed by others (e.g., seeing dead children or mutilated bodies) is more strongly associated with posttraumatic stress symptoms than exposure to more general combat-related experiences (Armstrong, Bryan, Stephenson, Bryan, & Morrow, 2014; Vargas et al., 2013). This may also suggest that the possible overlap of moral injury with PTSD is related to distress and emotional turmoil resulting from exposure to perceived immoral acts of others, which might include events such as witnessing injury, death, and atrocity committed by enemy combatants or even peers.
Transgressions-Self scores were most strongly associated with hopelessness and pessimism across both samples, which may suggest that inner distress associated with one’s own actions and decisions may be associated with negative expectations and/or a bleak outlook for the future. In Sample 2, Transgressions-Self scores were also correlated with anger intensity, which may suggest self-directed anger or hostility, although this conclusion is made cautiously since we did not assess for the target of participants’ anger. Unexpectedly, Transgressions-Self scores were not correlated with guilt or shame, suggesting that scores on this scale were not related to feelings of regret or embarrassment. The relationship of Transgressions-Self scores with hopelessness, pessimism, and anger, all of which are well-established risk factors for suicide (Bryan & Rudd, 2006), may explain why these scores are more strongly associated with suicide ideation and suicide attempts among military personnel than other MIES scores (Bryan, Bryan, Morrow, Etienne, & Ray-Sannerud, 2014). Longitudinal studies are needed to further examine the relation of Transgressions-Self scores with suicide-related outcomes among military personnel and veterans. Betrayal scores were also positively correlated with posttraumatic stress symptoms and anger. With respect to posttraumatic stress, this finding is consistent with the positive correlation of Transgressions-Others scores with posttraumatic stress symptoms in that both Betrayal and Transgressions-Others scores assess emotional distress related to others’ actions and decisions. Perhaps not surprisingly, perceived betrayal was also significantly correlated with intensity of anger, suggesting that military personnel who feel betrayed by others tend to experience more severe anger.
Overall, the fairly small zero-order correlations between the MIES scores and other indicators of psychological distress in both samples are generally consistent with the findings of Nash et al. (2013), who similarly found small to moderate correlations of MIES scores with other indicators of emotional distress, although it is important to note that Nash et al. aggregated MIES subscale scores into a single, global “moral injury” score. Direct comparisons of their results with the present findings are therefore not possible. Nonetheless, our results suggest that Transgressions-Others and Betrayal overlapped to some degree with several symptom clusters of PTSD, most notably the reexperiencing, avoidance, and hyperarousal symptom clusters. In contrast, Transgressions-Others and Betrayal scores had relatively smaller correlations with the hyperarousal symptom cluster, a pattern that has been hypothesized by Litz et al. (2009) and Shay (2014). Transgressions-Self scores, in contrast, were more strongly correlated with emotional numbing, which comprises a foreshortened sense of future and an inability to experience positive emotions in life. The correlation of Transgressions-Self scores with emotional numbing therefore aligns with our previously discussed findings regarding the scale’s association with hopelessness and pessimism.
The pattern of correlations among MIES scores and different types of potential traumas also lends some support to the factor structure of the MIES. Specifically, Transgressions-Self and Betrayal scores, which assessed emotional distress regarding the actions of others, were most strongly correlated with life stressors in which the respondent was the target or recipient of violence, hostility, or aggression of some kind. Transgressions-Self scores, by comparison, were much less frequently correlated with life stressors. Of note, Transgressions-Others scores were correlated with deployment-specific events as well as non–deployment-related events, but Betrayal scores had relatively few correlations with deployment-specific events. Conceptually, this converges with the notion of betrayal trauma, which entails a violation of trust by people or institutions on which a person depends for survival (Freyd et al., 2007). Because military personnel do not depend on enemy forces for survival, being attacked or harmed by them would not necessarily be considered a violation of trust, although it may nonetheless contribute to inner conflict related to the witnessing of violence and aggression more generally.
Although we have presented useful psychometric information regarding the MIES in two diverse military samples, it is important to highlight a number of limitations of the current study. First, all data were collected using self-report methodology, which might introduce response bias. Future studies that use validated diagnostic interviews would be important for further establishing the MIES’s construct validity as well as its ability to differentiate between military personnel with and without a range of clinical outcomes such as PTSD, major depressive disorder, and substance use disorders. This is particularly relevant in light of recent evidence suggesting that self-report symptom checklists like the PCL and the Beck Depression Inventory may be assessing generalized distress in addition to specific features of the diagnosis (Arbisi et al., 2012). Second, because the current methods were only cross-sectional in nature, we are unable to determine how the MIES is related to psychopathology among military personnel over time. Prospective studies are needed to determine how moral injury emerges with respect to other indicators of psychological distress over time. Despite these limitations, results from this study suggest that the MIES is multidimensional, has good internal consistency, and is associated with distinct psychological constructs in conceptually meaningful ways. Future research with the MIES should further develop the breadth of item content, assess the responsiveness of scale scores to clinical intervention, and determine its ability to predict the future emergence of psychopathology such as PTSD, depression, and suicide risk.
Footnotes
Authors’ Note
The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the U.S. Government, the Department of Defense, the Department of the Air Force, the Department of the Army, or the Military Suicide Research Consortium.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was in part supported by the Military Suicide Research Consortium, funded through the Office of the Assistant Secretary of Defense for Health Affairs under Award No. W81XWH-10-2-0181.