The independent and combined impact of moral injury and moral distress on post-traumatic stress disorder symptoms among healthcare workers during the COVID-19 pandemic

ABSTRACT

Background: Healthcare workers (HCWs) across the globe have reported symptoms of Post-Traumatic Stress Disorder (PTSD) during the COVID-19 pandemic. Moral Injury (MI) has been associated with PTSD in military populations, but is not well studied in healthcare contexts. Moral Distress (MD), a related concept, may enhance understandings of MI and its relation to PTSD among HCWs. This study examined the independent and combined impact of MI and MD on PTSD symptoms in Canadian HCWs during the pandemic.

Methods: HCWs participated in an online survey between February and December 2021, with questions regarding sociodemographics, mental health and trauma history (e.g. MI, MD, PTSD, dissociation, depression, anxiety, stress, childhood adversity). Structural equation modelling was used to analyze the independent and combined impact of MI and MD on PTSD symptoms (including dissociation) among the sample when controlling for sex, age, depression, anxiety, stress, and childhood adversity.

Results: A structural equation model independently regressing both MI and MD onto PTSD accounted for 74.4% of the variance in PTSD symptoms. Here, MI was strongly and significantly associated with PTSD symptoms (β = .412, p < .0001) to a higher degree than MD (β = .187, p < .0001), after controlling for age, sex, depression, anxiety, stress and childhood adversity. A model regressing a combined MD and MI construct onto PTSD predicted approximately 87% of the variance in PTSD symptoms (r² = .87, p < .0001), with MD/MI strongly and significantly associated with PTSD (β = .813, p < .0001), after controlling for age, sex, depression, anxiety, stress, and childhood adversity.

Conclusion: Our results support a relation between MI and PTSD among HCWs and suggest that a combined MD and MI construct is most strongly associated with PTSD symptoms. Further research is needed better understand the mechanisms through which MD/MI are associated with PTSD.

HIGHLIGHTS

• MI and MD were each independently associated with PTSD symptoms (including dissociation), when controlling for sex, age, childhood adversity, depression, anxiety and stress.

• Combining both MI and MD constructs into a single latent variable accounted for the greatest proportion of variance explained in PTSD symptoms among HCWs during the COVID-19 pandemic.

• Results suggest that expanding the construct of MI to include team and systemic organisational MD may be appropriate in the healthcare context.

Antecedentes: Trabajadores sanitarios (HCWs, en sus siglas en inglés) de todo el mundo han manifestado síntomas de trastorno de estrés postraumático (TEPT) durante la pandemia de COVID-19. El daño moral (MI, en sus siglas en inglés) se ha asociado a poblaciones militares, pero no se ha estudiado bien en contextos sanitarios. La angustia moral (MD, en sus siglas en inglés), un concepto relacionado, puede mejorar la comprensión del MI y su relación con el TEPT entre los HCWs. Este estudio examinó el impacto independiente y combinado de la IM y la DM sobre los síntomas de TEPT en los HCWs canadienses durante la pandemia.

Métodos: Los HCWs participaron en una encuesta en línea entre febrero y diciembre de 2021, con preguntas sociodemografícas, salud mental y antecedentes de trauma (por ejemplo, MI, MD, TEPT, disociación, depresión, ansiedad, estrés, adversidad infantil). Se utilizó un modelo de ecuaciones estructurales para analizar el impacto independiente y combinado del MI y la MD sobre los síntomas de TEPT (incluida la disociación) entre la muestra cuando se controló el sexo, la edad, la depresión, la ansiedad, el estrés y la adversidad infantil.

Resultados: Un modelo de ecuación estructural que regresionaba de forma independiente tanto el MI como la MD sobre el TEPT explicaba el 74,4% de la varianza en los síntomas del TEPT. Aquí, el MI se asoció fuerte y significativamente con los síntomas de TEPT (β = 0,412, p < 0,0001) en mayor grado que la MD (β = 0,187, p < 0,0001), después de controlar la edad, el sexo, la depresión, la ansiedad, el estrés y la adversidad infantil. Un modelo de regresión de un constructo combinado de MD e MI sobre el TEPT predijo aproximadamente el 87% de la varianza en los síntomas de TEPT (r² = 0,87, p < 0,0001), con MD/MI fuerte y significativamente asociado con el TEPT (β = 0,813, p < 0,0001), después de controlar la edad, el sexo, la depresión, la ansiedad, el estrés y la adversidad infantil.

Conclusión: Nuestros resultados respaldan la relación entre el MI y el TEPT entre los HCWs y sugieren que un constructo combinado de MD y MI se asocia más estrechamente con los síntomas de TEPT. Se necesitan más investigaciones para comprender mejor los mecanismos a través de los cuales MD/MI se asocian con el TEPT.

KEYWORDS:

• Post-traumatic stress disorder
• moral injury
• moral distress
• healthcare workers
• COVID-19

PALABRAS CLAVE:

• Trastorno de estrés postraumático
• daño moral
• angustia moral
• personal sanitario
• COVID-19

1. Introduction

Healthcare workers (HCWs) have reported elevated symptoms of Post-Traumatic Stress Disorder (PTSD) throughout the COVID-19 pandemic (Andhavarapu et al., 2022; Benfante et al., 2020; D’ettorre et al., 2021; Ouyang et al., 2022). Increased and repeated exposure to diverse stressors, such as patient suffering and death, visitor violence, and elevated risk of contracting and spreading the virus to loved ones, may in part account for heightened endorsement of PTSD symptoms among HCWs during the pandemic period (Andhavarapu et al., 2022; D’ettorre et al., 2021). PTSD is conceptualised as a response to a potentially psychologically traumatic event (PPTE; i.e. exposure to actual or threatened death, injury or sexual violence), including intrusion symptoms, avoidance behaviours, negative changes in cognitions and mood and alterations in reactivity (American Psychiatric Association, 2013; CIPSRT Glossary of Terms, 2019). In addition to PPTEs, HCWs have also faced ethical and moral dilemmas during the pandemic period, including potentially morally injurious events (PMIEs) (Litz, Stein et al., 2009), thus heightening risk of developing moral injury (MI) (Billings et al., 2021; Litz, Stein et al., 2009; Riedel et al., 2022; Ritchie et al., 2023; Xue et al., 2022). PMIEs are described as situations in which an individual: (i) personally violates their moral values by either acting (commission) or failing to act (omission); or (ii) experiences a violation of their morals by the actions or inactions of others, including betrayal by a trusted authority (Litz, Stein et al., 2009; Shay, 2014). For example, throughout the pandemic period, HCWs reported having to enforce no visitor policies, witnessing patients dying alone, working with inadequate personal protective equipment (PPE), strained working relationships with healthcare leaders and a lack of perceived support from the government as PMIEs (Brophy et al., 2021; Hegarty et al., 2022; Riedel et al., 2022; Ritchie et al., 2023 Xue et al., 2022). Indeed, almost 50% of Romanian physicians (N = 114) reported exposure to a PMIE during the initial months of the COVID-19 pandemic (Maftei & Holman, 2021). Not everyone who experiences a PMIE will go on to develop MI, however, among those impacted, MI can involve profound feelings of guilt, shame, anger and betrayal, a changed identity, social withdrawal, and spiritual/existentialcrisis (A. O. Bryan et al., 2014; Currier, Farnsworth et al. 2018; Currier, Holland et al., 2015; Jinkerson, 2016; Levi-Belz et al., 2020; Litz, Plouffe et al., 2022; Litz, Stein et al., 2009; Mantri et al., 2020). PMIE exposures during the pandemic have been associated with MI, alongside elevated symptoms of other negative mental health outcomes including anxiety, depression, burnout, PTSD, and suicidal ideation among HCWs (Amsalem et al., 2021; D’ettorre et al., 2021; Hines et al., 2021; Litam & Balkin, 2021; Marvaldi et al., 2021; Riedel et al., 2022; Shaukat et al., 2020; Zerach & Levi-Belz, 2021). Although MI is not currently a formalised mental health diagnosis, it may impact critical domains of functioning, including the psychological, emotional, behavioural, social, and spiritual/religious well-being of those who face PMIEs (Jinkerson, 2016; Litz, Stein et al., 2009).

In contrast to the high visibility of MI in military research, this concept remains relatively under explored in the healthcare context. It has been suggested that whereas MI in a military veteran context may relate to direct offence and harm toward human beings, among HCWs, MI may instead derive from ‘being unable to provide high-quality care and healing in the context of healthcare’ (Talbot & Dean, 2018). As demonstrated in military contexts, it appears probable that PMIE-exposure and MI would be associated with elevated endorsement of PTSD symptoms among HCWs. It has been queried whether MI and PTSD represent related, albeit distinct traumatic stress responses (C. J. Bryan et al., 2018; D'Alessandro et al., 2022; Litz, Stein et al., 2009). It is proposed that the types of stressors which may give rise to PTSD and MI, namely PPTEs and PMIEs, respectively, maintain unique features (American Psychiatric Association, 2013; CIPSRT Glossary of Terms, 2019; Litz, Stein et al., 2009). For example, although a criterion A index trauma (i.e. PPTE) associated with the development of PTSD (American Psychiatric Association, 2013) may involve a personal or witnessed violation of moral values (as in PMIEs), these moral conditions are not necessary to meet definitional criteria for the presence of a PPTE (CIPSRT Glossary of Terms, 2019). Conversely, a PMIE may involve actual or threatened death, sexual violence or injury (as in PPTEs), however, this feature is not necessary to determine the presence of a PMIE (Barnes et al., 2019; Litz, Stein et al., 2009). Notably, however, increased exposure to PMIEs and endorsement of MI are each independently and concurrently associated with an increase in symptoms of PTSD (Currier, Farnsworth et al. 2018; Currier, Holland et al., 2015; Easterbrook et al., 2023; Koenig et al., 2018; Nash et al., 2013). Interestingly, whereas Litz, Stein et al. (2009) suggested that the long-term trajectory of MI may include features consistent with PTSD, such as avoidance behaviours, intrusive memories and self-sabotaging behaviours (Litz, Stein et al., 2009), C. J. Bryan et al. (2018) focus instead on the distinct symptom profiles of MI and PTSD, where MI is thought most strongly associated with guilt, shame, anhedonia and alienation, and PTSD with memory loss, startle reflex and intrusive memories. Importantly, depression was associated with both presentations of MI and PTSD Bryan and colleagues’ work (C. J. Bryan et al., 2018).

A nascent theory proposes that the function of overlapping symptoms (e.g. anhedonia, avoidance) observed in MI and PTSD may differ, such that avoidance behaviour in the context of PTSD may relate most strongly to fear and safety concerns, but to shame and self-condemnation in the context of MI (Currier, Farnsworth et al. 2018). A study by Maguen et al. (2022) examined the unique contributions of MI, PTSD and depression symptoms on outcomes among military veterans. Here, as compared to MI related to betrayal or witnessing PMIEs, MI related to perpetration-based PMIEs was associated with worse outcomes in terms of work, health and relationship functioning, even after controlling for PTSD and depression (Maguen et al., 2022). Finally, no studies to date have examined the relation between MI and dissociation, despite findings from a recent meta-analysis and systemic review suggesting that almost 40% of individuals with PTSD may display symptoms consistent with its dissociative subtype (White et al., 2022). The dissociative subtype of PTSD is characterised by detachment from the body and environment (Spiegel, 2012) and has previously been linked to greater clinical severity and functional impairment (Boyd et al., 2018; Park et al., 2021; Stein et al., 2013). Capturing symptoms of dissociation in the context of PTSD may then be relevant to understanding the relation between MI and PTSD among HCWs.

As noted, prior to the COVID-19 pandemic, MI and PMIEs received scant attention in the healthcare arena. Instead, moral distress (MD), a term used to describe the psychological impact of being prevented by external constraints from acting in line with knowledge of an ethical course of action, was more frequently identified as a concern among HCWs (Epstein et al., 2019; Epstein & Hamric, 2009). MD has been associated with a range of deleterious psychological outcomes, including rage, anxiety, sadness, helplessness, powerlessness and loss of self-worth (Grimell & Nilsson, 2020). MD has been described at the patient, team and system levels of healthcare (Epstein et al., 2019). At the patient level, HCWs may experience distress when providing care that is perceived as futile or when a patient's agency is disrespected in deference to the patient's family (Epstein et al., 2019). At the team level, HCWs may experience morally distressing events, such as being disrespected by colleagues or working in teams with poor communication (Epstein et al., 2019). Furthermore, at the system level, moral distress may arise when working with unsafe staffing levels or inadequate levels of resource supply (Epstein et al., 2019).

At present, it remains unclear how MI and MD represent distinct or potentially overlapping concepts (Grimell & Nilsson, 2020). Although both MI and MD place emphasis on events that violate an individual's sense of right and wrong, consequently yielding various moral emotions (e.g. anger, guilt, shame), MD tends to be discussed as a short term response to moral challenges and MI as a long-term outcome of repeated exposure to moral challenges or to a single, extraordinarily difficult moral challenge (Grimell & Nilsson, 2020). Litz and Kerig (2019) posited that MD and MI differ with respect to precipitating stressors and the degree of impairment endured. In their model, there exists a continuum of types of moral stressors in the environment, including moral stressors and PMIEs, which vary in degree of frequency of occurrence, such that PMIEs are rare events when compared to moral stressors (Litz & Kerig, 2019). From these types of stressors, the outcomes MD and MI, respectively, may present as responses to such stressors (Litz & Kerig, 2019). Notably, the degree of psychological, social and spiritual harm and impact is thought to increase across outcomes, where MI represents the greatest degree of impairment across outcomes (Litz & Kerig, 2019). However, empirical evidence to support any conceptual distinctions between MI and MD remains weak, possibly hindered by the lack of adequate measurement tools to capture independent facets of MD and MI (Epstein et al., 2019; Litz et al., 2022; Nash et al., 2013; Plouffe et al., 2021).

Grimell and Nilsson (2020) argued for a unified MD/MI construct in the military literature that is based on the limitations present in each theory alone. Specifically, they argue that theories of each construct fail to address not only the broad range of moral challenges (e.g. in person life as well as work life), but also the complexity of the potential outcomes following exposure (Grimell & Nilsson, 2020). Accordingly, they suggest neither theory may be adequate to ‘capture the breadth and depth of moral stress in both time and space of military duty’ (Grimell & Nilsson, 2020, p. 385). In line with this theory, we posit that consideration of MD and MI as a unified construct may lend greater accuracy in capturing HCWs experiences than each construct independently. This view follows from the hypothesis that a unified construct permits a more complete range of moral challenges and health-related outcomes to be considered. As such, to better understand the association between MI and PTSD within the healthcare context, here we account for both MD and MI as potential contributors to PTSD symptoms among HCWs during the pandemic. Accordingly, the purpose of the present study was to explore the independent and combined impact of MD and MI on PTSD symptoms among a sample of Canadian HCWs who were employed in a healthcare setting during the COVID-19 pandemic. Notably, our conceptualisation of PTSD includes dissociative symptoms to reflect the heterogeneity of PTSD more accurately (White et al., 2022). We hypothesised that MD and MI would independently predict PTSD symptoms, with MI doing so to a greater degree than MD, and the combined impact being greater than either alone. Exploration of the association between MI and PTSD remains important in the healthcare arena as the risk for MI and PTSD among HCWs will not cease with the end of the pandemic period, particularly given the nature of healthcare work which inherently involves exposure to widespread stressors (Maftei & Holman, 2021).

2. Methods

2.1. Participants

The present study used a subset of data from a broader investigation of HCWs’ and public safety personnel's experiences during the COVID-19 pandemic (Hamilton Integrated Research Ethics Board #12667). Data were collected between February and December 2021. Canadian HCWs were recruited through social media and email advertisements from willing healthcare organisations across Canada. Participants were at least 18 years of age, resided in Canada, spoke and read English, and had contributed to patient care during the COVID-19 pandemic. The survey was housed on Research Electronic Data Capture (REDCap) software (Harris, Taylor, Minor et al., 2019; Harris, Taylor, Thielke et al., 2009). Electronic informed consent was received before participant survey access.

In the present study, 719 participants consented to participate. Of those, 613 participants completed demographics questions pertinent to the present analysis and at least one other scale in full; the other 106 participants (i.e. those who only completed demographics) were excluded. Missing data were imputed using multiple imputation methods (see Supplementary Materials and Patel, Easterbrook et al., 2023). This yielded a final sample size of n = 613 HCWs from across Canada.

2.1.1. Measures

The survey was composed of a battery of questionnaires (Table 1). The subset of data used in the present investigation included: a demographic form (e.g. age, sex assigned at birth, province of residence, occupation), the Moral Injury Outcome Scale (MIOS) (Litz, Plouffe et al., 2022; Yeterian et al., 2019), Measure of Moral Distress – Healthcare Professional (MMD-HP) (Epstein et al., 2019), Post-Traumatic Checklist for DSM-5 (PCL-5) (Weathers et al., 2013), Multiscale Dissociation Inventory (MDI) (Briere et al., 2005), Depression Anxiety Stress Scale-21 (DASS-21) (Lovibond & Lovibond, 1995), and the Adverse Childhood Experiences Scale (ACEs) (Merrick et al., 2017). The DASS-21 subscales of Depression, Anxiety, and Stress, along with the ACES score, were used as covariates in our structural equation model due to their associations with MI (Battaglia et al., 2019; Protopopescu et al., 2021; Roth et al., 2022) and PTSD (Frewen et al., 2019; Price & van Stolk-Cooke, 2015; Yehuda et al., 2001). The MIOS version used in the present study was received through personal communication with the scale's authors prior to its publication. As such, the MIOS used here does not reflect the most recent version of the scale as presented in Litz, Plouffe et al. (2022), but is similar to the final version apart from a slight change in wording of one item and a limited assessment of functional impacts.

Table 1. List of measures and variables used for structural equation modeling in study.

Scale	Construct	Scoring and interpretation
Moral Injury Outcomes Scale	Assesses endorsement of MI as multidimensional construct impacting the following domains: self-perception, moral thinking, social impacts, self-harming/sabotaging, impairing moral emotions, beliefs about purpose and meaning.	Participants answer preliminary questions regarding PMIE exposure, including PTSD screener items and an open-field textbox to describe the event. With the worst PMIE experienced in mind, participants read 14 statements and rate their degree of agreement on 5-point scale ranging from 0 (Strongly Disagree) to 4 (Strongly Agree). These 14 items can be organised into two subscales by summing the appropriate items: shame-related outcomes (1, 3, 7, 8, 12, 13, 14), trust-violation related outcomes (11, 2, 4, 5, 6, 9, 10). An additional question rated on a 7-point scale (0, Not at all; 6, Extremely) invites participants to rate their self-reported degree of functional impairment.
Measure of Moral Distress – Healthcare Professional	Measures the frequency of exposure to morally distressing events and degree of distress associated across patient, team and system levels of healthcare.	Participants read 27 statements of experiences they may have endured in their healthcare profession and rated each statement twice on a 5-point scale for both frequency and level of distress, where 0 represented Never/None and 4 represented Very Frequently/Distressing. A total score was calculated by summing the products of the frequency and distress ratings for each statement. The subscales were calculated by summing the products of the frequency and distress ratings for the following items: patient (items 1,2,3,5,8,10), team (items 6,9,11,12,13,14,15,20,21,24,25,26,27), system (items 4,7,16,17,18,19,22,23).
Post-Traumatic Checklist for DSM-5	Assesses the presence and severity of PTSD symptoms across four clusters: (i) intrusions, (ii) avoidance, (iii) negative alterations in cognitions and mood, (iv) reactivity.	Participants rated their degree of agreement with 20 items on a 5-point scale ranging from 0 to 4 as experienced over the past month. A total score was calculated by summing the 20 items. Total scores equal to or greater than 33 were considered above the cutoff indicative of likely PTSD. The subscale scores were calculated as follows: intrusions (sum of items 1–5), avoidance (sum of items 6 and 7), negative alterations in cognitions and mood (sum of items 8–14), reactivity (sum of items 15–20).
Multiscale Dissociation Inventory	Assess for the presence of dissociative symptomology across 6 subscales: disengagement, depersonalization, derealization, emotional constriction, memory disturbance, identity dissociation.	Participants rated their degree of agreement with 30 statements regarding their experience over the past month on a 5-point scale ranging from 1 (Never) to 5 (Very Often). The subscales used in this analysis were calculated by summing the following items: Depersonalization (items 2,8,14,20,26), Derealization (items 3,9,15,21,27).
Depression Anxiety Stress Scale 21	Assesses for the severity of symptoms of depression (e.g. dysphoria, hopelessness, devaluation of life, self-deprecation, inertia), anxiety (e.g. autonomic arousal, skeletal muscle effects, situational anxiety, subjective anxious affect) and stress (e.g. difficulty breathing, nervous arousal, being easily upset/agitated, irritable/over-reactive and impatient).	Participants rated their level of agreement with 21 statements on a 4-point scale ranging from 0 (Did not apply to me at all) to 3 (Applied to me very much or most of the time) as applicable over the past week. The subscales were calculated by summing the following items and then multiplying the sum by 2: depression (items 3,5,10,13,16,17,21), anxiety (items 2,4,7,9,15,19,20), stress (items 1,6, 8,11,12,14,18).
Adverse Childhood Experiences Scale	Assesses endorsement of adversity in childhood.	Participants answered 10 yes/no questions about their experience before the age of 18. Total scores were calculated by summing the ratings for the 10 items, where ‘yes’ received a score of 1 and ‘no’ received a score of 0.

2.1.2. Data analytic strategy

2.1.2.1. Descriptives and structural equation modeling

Prior to modelling, descriptive statistics were run on all variables of interest in the study. Structural equation modelling was implemented in MPlus Version 8.4 (Muthén & Muthén, 1998–2012) using maximum likelihood estimation to assess the association between MI and PTSD symptoms. Before modelling, the Depersonalization and Derealization subscales from the MDI were square-root transformed based on benchmarks of skewness >2 and kurtosis >7 (Curran et al., 1996). There were no significant outliers within the sample (Zs > 3.29). Two-tailed tests with a statistical significance of α < .05 were used. Following established conventions (Hu & Bentler, 1999; Schreiber et al., 2006), an excellent-fitting model has a comparative fit index (CFI) and Tucker Lewis Index (TLI) ≥ .95, standardised root mean squared residual (SRMR) of ≤.08, and a root mean square error of approximation (RMSEA) ≤ .06.

Three latent variables were created for the first model: (1) MI, (2), MD, and (3) PTSD. The MI latent variable was created from all Likert-type items on the MIOS: (1) Trust subscale; (2) Shame subscale; (3) and functional impairment item. The initial dichotomous variables of the MIOS were not included as they do not contribute to a total score for the MIOS, nor reflect specifically the construct of MI as defined in the literature (e.g. changes in self-perception, beliefs about meaning and purpose, moral thinking, social impacts and self-harming/sabotaging behaviours) (Litz, Plouffe et al., 2022). Rather, the dichotomous initial questions on the MIOS serve to characterise the PMIE, which differs from moral injury as an outcome. The MD latent variable was created from the following subscales on the MMD-HP: (1) Team subscale; (2) Patient subscale; and (3) System subscale. The latent variable for PTSD was created from the following subscales: (1) Intrusions subscale of PCL-5; (2) Avoidance subscale of PCL-5; (3) Negative Alterations in Cognition and Mood subscale of PCL-5; (4) Reactivity subscale of PCL-5; (5) Depersonalization subscale of MDI; and (6) Derealization subscale of MDI. These scales were used to depict the symptom presentation of PTSD analytically (Patel, Easterbrook et al., 2023; Patel, Holshausen et al., 2021; Patel, O’Connor et al., 2022) and includes dissociative symptoms of PTSD, which align with the dissociative subtype. The first structural equation model evaluated the independent impact of MI and MD on PTSD symptoms. Within the second model, we created a combined MI and MD latent variable from the following subscales: (1) Trust subscale from MIOS; (2) Shame subscale from MIOS; (3) and functional impairment item from MIOS; (4) Team subscale from MMD-HP; and (5) System subscale from MMD-HP. The Patient subscale from MMD-HP was initially included in the latent variable but dropped as the factor loading weight was extremely low at a value of <.2 (Dash & Paul, 2021). In the second structural equation model, an exploratory analysis regressing the combined MI and MD latent variable onto PTSD symptoms was constructed.

3. Results

3.1. Participant characteristics

The sample consisted of 613 HCWs with a median age of 30–39 years old, and 88.9% identified as female. Approximately 26.1% of the sample met criteria for a possible diagnosis of PTSD according to the commonly used PCL-5 cut-off score of 33 (Weathers et al., 2013). Additionally, 49.6%, 49.9% and 40.1% of the sample reported moderate to extremely severe depression, anxiety, and stress symptoms, respectively, based on established cut-offs (Lovibond & Lovibond, 1995). Three-quarters (75.5%) of the sample endorsed at least one adverse childhood experience (Table 2). Bivariate correlations between all variables of interest are presented in Table 3.

Table 2. Participant characteristics for HCWs sample.

	Percentage
Demographics
Age (years)
20–29	20.1
30–39	34.4
40–49	26.1
50–59	15.2
60–69	3.8
70–79	.5
Sex
Female	88.9
Male	11.1
Province/Territory
British Columbia	12.1
Alberta	12.7
Saskatchewan	3.4
Manitoba	3.1
Ontario	56.8
Quebec	1.5
New Brunswick	2.3
Nova Scotia	5.7
Prince Edward Island	.5
Newfoundland and Labrador	.8
Northwest Territories	.3
Nunavut	.2
Yukon	.5
Occupation
Respiratory Therapist	41.0
Registered Nurse, Practical Nurse	27.8
Medical Physician	4.1
Personal Support Worker	2.1
Occupational Therapist	6.7
Social Worker	2.0
Physiotherapist	1.0
Other (e.g. Paramedic, Medical Office Assistant, Dental Hygienist)	15.3
Trauma-Related Variables	M (SD)	Cronbach's α
Childhood Adversity	2.77 (2.38)	–
Moral Injury	21.33 (8.86)	–
Trust	12.86 (4.79)	.73
Shame	8.47 (5.42)	.85
Functional Impairment	2.69 (1.59)	–
Moral Distress	137.04 (77.35)
Patient	33.22 (23.38)	Freq: .85, Dist: .85
Team	56.93 (38.69)	Freq: .88, Distr: .90
System	46.89 (28.89)	Freq: .85, Distr: .86
Probable PTSD Diagnosis	26.1
PTSD Severity	26.15 (15.90)	–
Intrusions	6.32 (4.35)	.88
Avoidance	2.78 (2.13)	.81
Negative Alterations in Cognition and Mood	9.41 (6.24)	.90
Reactivity	7.64 (4.93)	.84
Trauma-Related Dissociation		–
Depersonalization	7.85 (3.06)	.80
Derealization	8.00 (2.97)	.81
Psychiatric Symptoms
Depressive Symptoms	13.63 (9.01)	.90
Anxiety Symptoms	10.02 (7.40)	.80
Stress Symptoms	17.66 (8.10)	.84

Note: All values presented for demographics and probable PTSD diagnosis are percentages. Means and standard deviations presented for all other variables. Childhood Adversity = sum score on ACES; MI = sum score on MIOS; Trust = subscale on MIOS; Shame = subscale on MIOS; Functional Impairment = item score on MIOS; Moral Distress = sum score on MMD-HP; Patient = subscale on MMD-HP; Team = subscale on MMD-HP; System = subscale on MMD-HP; Probable PTSD Diagnosis = percentage of individuals above 33+ cutoff score on PCL-5; PTSD Severity = sum score on PCL-5; Intrusions = cluster B subscale on PCL-5; Avoidance = cluster C subscale on PCL-5; Negative Alterations in Cognition and Mood (NACM) = cluster D subscale on PCL-5; Reactivity = cluster E subscale on PCL-5; Depersonalization  = subscale on MDI; Derealization = subscale on MDI; Depressive Symptoms = depression subscale on DASS-21; Anxiety Symptoms = anxiety subscale on DASS-21; Stress Symptoms = stress subscale on DASS-21.

Table 3. Bivariate correlations between variables of interest.

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1. MIOS_Total	–
2. MIOS_Shame	.885**	–
3. MIOS_Trust	.849**	.506**	–
4. MIOS_16	.603**	.524**	.524**	–
5. MMDHP_Total	.329**	.236**	.342**	.347**	–
6. MMDHP_Patient	.177**	.117**	.195**	.203**	.732**	–
7. MMDHP_Team	.343**	.243**	.361**	.340**	.937**	.575**	–
8. MMDHP_System	.283**	.216**	.280**	.318**	.856**	.415**	.724**	–
9. PCL5	.524**	.445**	.467**	.527**	.490**	.269**	.513**	.417**	–
10. PCL_INT	.414**	.331**	.391**	.430**	.477**	.269**	.508**	.390**	.891**	–
11. PCL_AVO	.415**	.365**	.355**	.406**	.417**	.231**	.444**	.342**	.806**	.744**	–
12. PCL_NACM	.543**	.472**	.470**	.526**	.445**	.228**	.461**	.398**	.932**	.740**	.683**	–
13. PCL_REA	.458**	.386**	.412**	.476**	.414**	.240**	.429**	.347**	.908**	.733**	.646**	.792**	–
14.Depression	.536**	.492**	.436**	.395**	.307**	.192**	.302**	.269**	.620**	.484**	.425**	.652**	.564**	–
15. Anxiety	.408**	.382**	.323**	.317**	.314**	.208**	.321**	.251**	.604**	.556**	.457**	.540**	.576**	.651**	–
16. Stress	.477**	.397**	.434**	.410**	.342**	.236**	.321**	.305**	.601**	.507**	.389**	.577**	.592**	.709**	.669**	–
17. DPER	.358**	.354**	.261**	.299**	.296**	.211**	.286**	.247**	.542**	.422**	.418**	.534**	.517**	.496**	.498**	.431**	–
18. DREAL	.375**	.360**	.286**	.355**	.314**	.203**	.309**	.269**	.583**	.480**	.430**	.563**	.557**	.466**	.526**	.432**	.776**	–
19. ACES	.149**	.152**	.104*	.179**	.102*	−.064	.145**	.127**	.303**	.229**	.246**	.313**	.273**	.150**	.179**	.134**	.161**	.226**	–
20. Age	−.044	−.080*	.020	−.033	.018	−.052	.027	.020	.036	.026	.029	.016	.054	−.007	−.087*	−.058	−.090*	−.096*	.019	–
21. Sex	−.001	−.005	.015	.010	.105**	.053	.099*	.093*	.058	.094*	.090*	.02	.035	.026	.039	.032	.053	.043	.013	−.014	–

All values presented are Pearson's r correlations except for age and sex, which are Spearman's rho correlations.

** Correlation is significant at the .01 level (2-tailed).

* Correlation is significant at the .05 level (2-tailed).

3.2. Independent impacts of MI and MD on PTSD

In the first structural equation model, both MI and MD were separately regressed onto PTSD to examine the independent impact of each factor on PTSD (Figure 1). This model accounted for 74.4% of the variance in PTSD (r² = .744, p < .0001). MI strongly and significantly predicted PTSD severity (β = .412, p < .0001) even after controlling for age, sex, depressive symptoms, anxiety symptoms, stress, and childhood adversity. MD also significantly predicted PTSD but at a lower strength of association (β = .187, p < .0001) after controlling for the same covariates. There were significant effects of age (β = .058, p < .05), childhood adversity (β = .145, p < .0001), depressive symptoms (β = .189, p < .0001), and anxiety symptoms (β = .237, p < .0001). Sex and stress did not significantly predict PTSD (p's > .05).

Figure 1. MI and Moral Distress independently and significantly predict PTSD symptoms among HCWs during COVID-19.

Note: Structural equation model depicting associations between MI and PTSD severity among HCWs. All values are standardised. Standard errors for residuals and covariances in parentheses. MI  =  MI latent factor, FI = Functional Impairment Item on MIOS, PTSD = PTSD latent factor, INT = ntrusions, AVO = Avoidance, NACM = Negative Alterations in Cognition and Mood, REA = Reactivity, DPER = Depersonalization, DREAL = Derealization. Covariates included depressive symptoms, anxiety symptoms, stress, and childhood adversity. Model fit was excellent (X²(111, N = 613) = 457.24, p < .0001, TLI = .922, CFI = .941, RMSEA = .071 [95% CI = .065–.078], SRMR = .043).

3.3. Combined impact of MI and MD on PTSD

An exploratory structural equation model was run to assess the combined effect of MI and MD (in a single latent variable) on PTSD (Figure 2). The model accounted for 87% of the variance in PTSD (r² = .87, p < .0001). The combined MI and MD latent variable strongly and significantly predicted PTSD (β = .813, p < .0001) after controlling for age, sex, depressive symptoms, anxiety symptoms, stress, and childhood adversity. There were significant effects of age (β = .074, p < .01), childhood adversity (β = .091, p < .05), and anxiety symptoms (β = .160, p < .01) on PTSD severity, but sex, depressive symptoms and stress did not significantly predict PTSD (ps > .05).

Figure 2. A latent variable combining MI and MD significantly predicts PTSD symptoms among HCWs during COVID-19.

Note: Structural equation model depicting associations between MI and PTSD severity among HCWs. All values are standardised. Standard errors for residuals and covariances in parentheses. MI = MI latent factor, FI = Functional Impairment Item on MIOS, PTSD = PTSD latent factor, INT = Intrusions, AVO = Avoidance, NACM = Negative Alterations in Cognition and Mood, REA = Reactivity, DPER = Depersonalization, DREAL = Derealization. Covariates included depressive symptoms, anxiety symptoms, stress, and childhood adversity. Model fit was excellent (X²(102, N = 613) = 440.58, p < .00001, TLI = .926, CFI = .943, RMSEA = .074 [95% CI = .067–.081], SRMR = .056).

4. Discussion

The present study explored the independent and combined impact of MI and MD on PTSD symptoms among Canadian HCWs during the COVID-19 pandemic. In support of our hypotheses, our results demonstrated that MI more strongly predicted PTSD symptoms than MD when MI and MD were considered independently. In the combined model, the latent variable MD/MI (defined by trust violation, shame, functional impairment, team- and system-level distress) accounted for a greater proportion of variance in PTSD symptoms (including dissociation) than either MI or MD alone, even when controlling for factors previously associated with PTSD and MI (e.g. age, sex, depression, anxiety, childhood adversity).

The exploratory model in the present study demonstrated that a combined MD/MI construct predicted a high proportion (87%) of variance in PTSD symptoms. The use of latent variable modelling to create latent constructs for MD/MI and PTSD symptoms builds upon prior literature on PTSD in the field (Patel, Holshausen et al., 2021; Patel, O’Connor et al., 2022). Notably, the proportion of variance accounted for in PTSD symptoms depended on all variables included within the model, including the covariates (e.g. age, sex, depression, anxiety, stress, childhood adversity), and not simply MI and MD. Although several covariates did not significantly predict PTSD symptoms, they nonetheless accounted for some of the overall variance in PTSD symptoms. Here, the model's beta estimate for MD/MI predicting PTSD symptoms was .813, indicating that for every one-unit change in MD/MI, PTSD symptoms increase by approximately .813 units given the positive direction of the beta coefficient. This highlights the strong magnitude of relation between MD/MI and PTSD symptoms among this sample of HCWs during the COVID-19 pandemic.

The conceptualisation of MI within the healthcare arena has proven difficult, as there exists limited empirical evidence on how this concept fits with other commonly used terms, including MD (Čartolovni et al., 2021). Assumptions that MD may compound over time to result in MI (Čartolovni et al., 2021; Epstein & Hamric, 2009; Riedel et al., 2022), and theoretical models that MD and MI may differ in severity (Litz & Kerig, 2019), suggest continuity between these concepts. Drawing on Grimell and Nilsson’s (2020) theory that a combined MD/MI construct may lend greater conceptualisation to the field, our results show that by accounting for team- and system-related MD in combination with shame, trust-violation and functional impairment related to MI, a substantial proportion of the variance in PTSD symptoms was explained – more than either MD or MI could explain alone. Team- and system-related MD may involve conflicts within units, among colleagues, within the organisation, or within the healthcare system (Epstein et al., 2019). Indeed, many HCWs have described interpersonal and organisational issues that caused them distress during the COVID-19 pandemic, including working with colleagues who lacked experience in critical care, and feeling unsupported by their organisation when working with inadequate staffing levels, resources and personal protective equipment (Billings et al., 2021; Riedel et al., 2022; Ritchie et al., 2023; Xue et al., 2022). Here, our findings are consistent with Dean et al.’s (2019) conceptualisation of MI in healthcare, which emphasised MI as arising from being constrained from acting in line with the interests of a patient due to broken systems that emphasise profit over patient care (Dean et al., 2019). Similarly, our findings, offer a more nuanced perspective of MI in healthcare, suggesting that also accounting for institutional, systemic constraints may better fit the conceptualisation of MI among HCWs and its relation to PTSD. Critically, however, this study and the field remain limited by the tools used to assess MD and MI (Maguen & Griffin, 2022). Indeed, there is yet to be agreement on appropriate measurement tools for these constructs and, indeed, the MMD-HP used in the study to assess MD may be criticised for its multiplicative scoring structure, which conflates exposure to distressing events and MD as an outcome. Adequate measurement tools for MD and MI are urgently needed to better capture the everyday experiences of HCWs (Maguen & Griffin, 2022).

With increased endorsement of PTSD symptoms from HCWs throughout the COVID-19 pandemic (Andhavarapu et al., 2022; Benfante et al., 2020; D’ettorre et al., 2021; Ouyang et al., 2022), adequate prevention and intervention strategies are urgently needed to support this vital workforce. Our findings demonstrate the further necessity of considering MD/MI in understanding PTSD symptoms among this population. Here, it is important to acknowledge that lived experience of events in which morals are violated, along with the types of response elicited from the event, will not always fit neatly into a conceptual category, such as MD or MI (Rosen et al., 2022). Despite this lack of ‘fit’, consideration of morally distressing and morally injurious experiences remains critical to the development of targeted interventions to support HCWs. For example, assuming a continuum between MD and MI, Rosen et al. (2022) highlighted the unique interventions required in response to MD and MI, respectively, such that moral distress may be inherent to the healthcare field as tension between one's personal morals and the requirements of healthcare roles may conflict (Rosen et al., 2022). By contrast, they suggest MI may be best addressed by organisational change as ‘the problem is not individual weakness but rather systems and contexts in need of reform’ (Rosen et al., 2022). Importantly, the team and system subscales of MD that loaded well into our latent MD/MI construct were related to organisational and systemic issues that also require rectification at the structural level. Here, additional research is required to develop and test targeted interventions for MD/MI in the context of PTSD among HCWs, with this need extending well beyond the pandemic period. In addition, extending this association study and exploring the mechanisms through which MD/MI relate to PTSD may prove useful. Finally, disentangling the relation between MD/MI and PTSD symptoms as predictive rather than associative will inform the identification and development of therapeutic targets for intervention based on mitigating MD/MI to prevent or alleviate PTSD symptoms among HCWs following exposure to PMIEs and PPTEs.

4.1. Limitations

The novel findings of this study must be interpreted within the context of methodological limitations. Although a broad recruitment strategy was employed, most of the sample included in the analysis were female respiratory therapists and nurses who resided in Ontario, Canada. The sample is not representative of all Canadian or global HCWs. Secondly, the present study did not account for previous mental health history, rendering it unclear to what extent MD/MI and PTSD symptoms reported by the sample were unique due to the COVID-19 context or from previous HCW experiences. Additionally, among the combined MD/MI latent variable, the Patient subscale from the MMD-HP was dropped due to low factor loading. When considering the combined impact of MI and MD on PTSD symptom severity, it must be noted that MD related to patient care was not included. Finally, burnout is an additional syndrome commonly reported among HCWs even prior to the COVID-19 pandemic. Here, some authors theorise that MI may be a precursor to burnout, such that a continuum may exist ranging from MD to MI, to burnout, where burnout represents a state ‘beyond feeling;’ detachment, depersonalisation and absence of distress (Rosen et al., 2022). The present study did not consider burnout as a clinical syndrome which may further aid in explaining elevated symptoms of PTSD symptoms among HCWs during the COVID-19 pandemic. Further research is needed to enhance conceptual clarity among the constructs MD, MI and burnout and to better understand how these presentations may be related to PTSD symptom endorsement.

5. Conclusions

HCWs have been exposed to diverse stressors working on frontlines during the COVID-19 pandemic, placing them at heightened risk for adverse mental health outcomes, including MD, MI and PTSD. As the relation between these traumatic responses remains poorly understood in the healthcare context, the present study investigated the independent and combined impact of MI and MD on PTSD, finding that heightened endorsement of combined MD/MI was significantly associated with increased endorsement of PTSD symptoms. Further research is needed to understand the mechanisms through which MD/MI are associated with PTSD.

Acknowledgements

We thank all Canadian HCWs who served at the forefront of the COVID-19 pandemic, and we are grateful to those who participated in this study. The authors recognise and acknowledge that this work was conducted on the traditional territories of the Mississauga and Haudenosaunee nations, and within the lands protected by the Dish with One Spoon wampum belt agreement.

Data Availability Statement

The data used in this study come from the McKinnon Trauma and Recovery Research Unit at McMaster University. All interested researchers may apply for access to these data through online application subject to review by the Data Access Committee, ethics approval, and signing of a data sharing agreement. Data are provided only once a data sharing agreement is in place between McMaster University (the custodian of the data) and the researchers’ institution. For more information about data access please contact https://www.thetraumaandrecoverylab.com/contact.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research was supported by a contract to McMaster University from the federally funded Atlas Institute for Veterans and Families, a generous donation from Homewood Health Inc. to Homewood Research Institute, a grant from the Canadian Institutes of Health Research (grant number MVP-171647) to MCM and RL, and a grant to MCM from the Public Health Agency of Canada. MCM is supported as the Homewood Chair in Mental Health and Trauma at McMaster University. RL is supported as the Harris-Woodman Chair in Psyche and Soma at Western University of Canada. A.M.D’A-L is supported by the CIHR Canada Doctoral Graduate Scholarship (#493412), the Horne Family Memorial Fellowship in Post-Traumatic Stress Injury and Recovery Fellowship, and the Research Institute of St. Joe's Hamilton Studentship Award.