Hostname: page-component-78c5997874-ndw9j Total loading time: 0 Render date: 2024-11-05T00:48:29.829Z Has data issue: false hasContentIssue false

Health trajectories of international humanitarian aid workers: growth mixture modelling findings from a prospective cohort study

Published online by Cambridge University Press:  17 May 2023

Kaz de Jong*
Affiliation:
Médecins Sans Frontières, Amsterdam, The Netherlands
Saara. E. Martinmäki
Affiliation:
Centre of Excellence Impact, ARQ Centre of Expertise for the Impact of Disasters and Crises, Diemen, The Netherlands
Hans te Brake
Affiliation:
Centre of Excellence Impact, ARQ Centre of Expertise for the Impact of Disasters and Crises, Diemen, The Netherlands
Ivan Komproe
Affiliation:
Research and Development Department, Utrecht University, Utrecht, The Netherlands; and HealthNet TPO, Amsterdam, The Netherlands
Rolf J. Kleber
Affiliation:
Department of Clinical Psychology, Utrecht University, Utrecht, The Netherlands; and ARQ National Psychotrauma Centre, Diemen, The Netherlands
Joris F. G. Haagen
Affiliation:
ARQ Centre of Expertise for the Impact of Disasters and Crises, Diemen, The Netherlands
*
Correspondence: Kaz de Jong. Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Background

Most staff stay healthy during humanitarian work, although some worsen. Mean scores on health indicators may be masking individual participants struggling with health issues.

Aims

To investigate different field assignment-related health trajectories among international humanitarian aid workers (iHAWs) and explore the mechanisms used to stay healthy.

Method

Growth mixture modelling analyses for five health indicators using pre-/post-assignment and follow-up data.

Results

Among 609 iHAWs three trajectories (profiles) were found for emotional exhaustion, work engagement, anxiety and depression. For post-traumatic stress disorder (PTSD) symptoms, four trajectories were identified. The ‘healthy/normative’ trajectory had the largest sample size for all health indicators (73–86%). A stable (moderate) ‘ill health’ trajectory was identified for all health indicators (7–17%), except anxiety. An ‘improving’ trajectory was found for PTSD and anxiety symptoms (5–14%). A minority of staff (4–15%) worsened on all health indicators. Deterioration continued for PTSD, depressive symptoms and work engagement 2 months post-assignment. A strong sense of coherence was associated with higher odds of belonging to the ‘healthy’ trajectory. Female biological sex was associated with higher odds of belonging to the ‘worsening’ depression and anxiety trajectories. Extended duration of field assignment was related to higher odds of belonging to the ‘worsening’ depressive symptoms trajectory.

Conclusions

Most iHAWs stayed healthy during their assignment; a stable ‘ill health’ trajectory was identified for most health indicators. Sense of coherence is an important mechanism for understanding the health of all iHAWs in the different health trajectories, including the ‘healthy’ profile. These findings give new possibilities to develop activities to prevent worsening health and help strengthen iHAWs’ ability to remain healthy under stress.

Type
Paper
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2023. Published by Cambridge University Press on behalf of the Royal College of Psychiatrists

A recent study among international humanitarian aid workers (iHAWs) of a large international humanitarian emergency organisation demonstrated that most staff, despite a highly stressful work environment, stayed healthy during and after humanitarian aid assignments.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 Aid workers consist of different groups, such as international staff, professional consultants and locally contracted national staff.Reference Stoddard, Harvey, Czwarno and Breckenridge2 The findings also showed a large range among the health scores. In particular, in healthy populations mean scores may be masking individual participants struggling with health issues. It warrants further exploration to ensure appropriate differentiation between subgroups of individuals with different response patterns on the study variables. Latent class growth analysis (LCGA) enables presentation and understanding of this heterogeneity. Findings from other populations exposed to extreme events report different pre-, post- and follow-up deployment trajectories.Reference Galatzer-Levy, Huang and Bonanno3Reference Van der Wal, Gorter, Reijnen, Geuze and Vermetten5 These trajectories show various expressions: chronic ill health (low pre- and post-assignment health scores relative to their peers), worsening health (high pre-, low post-assignment health scores), healthy (high pre- and post-assignment health scores) and improving (moderate pre-assignment, high post-assignment health scores).

Identifying and understanding the different health trajectories and their predictors is important for the development of both theoretical and practical knowledge. It may open new avenues on how to prevent, mitigate and treat iHAWs’ health problems. Different mechanisms and variables may be involved. For example, the theory of salutogenesis focuses on the individual's capacity to manage, comprehend and give meaning to the perceived (dis)stress and demands and factors that support human health and well-being, rather than on factors that cause disease. The related concept of sense of coherence postulates how individuals manage, comprehend and search for meaning in (extreme) stress and stay healthy. A high sense of coherence may protect iHAWs against the negative psychological impact of potentially traumatic stress.Reference Veronese and Pepe6 Antonovsky considered sense of coherence to be a unidimensional construct built on three key components: comprehensibility (ability to clarify, structure stressors), manageability (awareness and confidence to manage stressors successfully) and meaningfulness (willingness and motivation to manage stressors). He considered the sense of coherence to be a trait-like disposition, stable over time.Reference Antonovsky7 In the transactional stress model,Reference Lazarus and Folkman8 sense of coherence is a personal resource, an underlying mechanism that determines the stress or coping response. Other mechanisms iHAWs may use to stay healthy while enduring high levels of stress are coping self-efficacy (one's belief in one's ability to succeed in highly demanding situations) and several types and resources of social support. High levels of both are associated with good health while handling stress, emotions and demanding aid work.Reference van der Velden, van Loon, Benight and Eckhardt9,Reference Eriksson, Bjorck, Larson, Walling, Trice and Fawcett10

In the present study we aim to demonstrate different assignment-related health trajectories in iHAWs. Statistical modelling of health change trajectories is a unique approach to gain new insights into the health of iHAWs. The findings of this approach are not a technical exercise using sophisticated statistical methods, neither are they limited to showing relationships between a set of variables in a large longitudinal data-set of humanitarian aid workers. The analyses used provide a clearer understanding into how and why iHAWs maintain their health or lose their capacity to remain healthy.

We will evaluate five health indicators that showed high variance in their longitudinal mean scores in previous research:Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 symptomatology of post-traumatic stress disorder (PTSD), emotional exhaustion, anxiety, depression, and work engagement as indicators of well-being. To further support and improve iHAWs’ health, predictors associated with the various health trajectories are identified. We expect different levels of sense of coherence, coping self-efficacy and perceived social support to predict membership of beneficial and detrimental health trajectories.Reference Veronese and Pepe6

Method

Participants

The current study was a prospective survey of 609 iHAWs of Médecins Sans Frontières Operational Centre Amsterdam (MSF OCA). Additional study details can be found elsewhere, including detailed psychometric information about each measure and field assignment information;Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 participant information is presented in Table 1. Independent non-MSF researchers contacted all iHAWs going to a field assignment between December 2017 and February 2019 to inform them about the study; data collection ended in February 2020. Participants signed an informed consent. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human subjects/patients were approved by Ethics Review Board of MSF (ID 1642).

Table 1 Participant information (n = 609)

PCL-5, PTSD Checklist for DSM-5; MBI, Maslach Burnout Inventory; UWES-9, Utrecht Work Engagement Scale; HSCL-25, Hopkins Symptom Checklist; SOC, Sense of Coherence scale; CSE-7, Coping Self-Efficacy Scale; MSPSS, Multi-Dimensional Scale of Perceived Social Support.

Procedure

Participants completed online questionnaires pre-assignment (T1: 0–14 days pre-departure; response rate 98%), post-assignment (T2: within 4 weeks of returning; response rate 82%) and follow-up (T3: 2 months after T2, or 4–8 weeks after T2 in case of a new assignment; response rate 61%). We found no evidence that results on changes in health outcomes were influenced by those who declined to participate or dropped out of the study, based on decliner, non-responder and sensitivity analyses.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1

Instruments

Health outcome measures

The PTSD Checklist for DSM-5 (PCL-5) (past month; range 0–80, scores above 31–33 indicate probable PTSD)Reference Blevins, Weathers, Davis, Witte and Domino11,Reference Bovin, Marx, Weathers, Gallagher, Rodriguez and Schnurr12 measures the DSM-5 symptoms of PTSD. In the current sample, the scale had good internal consistency (α = 0.89).

The emotional exhaustion subscale of the Maslach Burnout Inventory – Human Services Survey (MBI-HSS) (range: 0–6)Reference Maslach, Jackson and Leiter13 measures burnout-related complaints. High emotional exhaustion is defined as a score of 2.1 or above based on the critical boundaries calculation for population norms.Reference Leiter and Maslach14 The internal consistency of this subscale was good (α = 0.84).

The Utrecht Work Engagement Scale (UWES-9) (range 0–6) measures work engagement.Reference Schaufeli, Bakker and Salanova15 Threshold scores indicate very low (<1.77), low (1.78–2.88), average (2.89–4.66), high (4.67–5.50) and very high (>5.50) work engagement.Reference Schaufeli and Bakker16 High scores correspond to a positive and fulfilling work-related state of mind. Internal consistency was good (α = 0.84).

The Hopkins Symptom Checklist (HSCL-25) (25-item self-report questionnaire, rated on a 1–4 Likert scale) assesses symptoms of anxiety and depression during the past week.Reference Derogatis, Lipman, Rickels, Uhlenhuth and Covi17 A cut-off score of 1.75 was used to screen for elevated symptoms of depression or anxiety.Reference Derogatis, Lipman, Rickels, Uhlenhuth and Covi18 The internal consistency in the current sample was good for both the depression (α = 0.90) and anxiety (α = 0.87) subscales.

Profile membership (resilience) variables

The Sense of Coherence (SOC) scale (range 13–91) measures the concept ‘sense of coherence’, which includes three properties: (a) comprehensibility (capacity to understand the situation or problem), (b) manageability (the belief that one can master the situation or problem by oneself) and (c) meaningfulness (experience and awareness of sufficient meaning and motivation to manage the situation or problem).Reference Antonovsky7 People with a high sense of coherence are able to manage (extreme) stressors in a way that maintains and/or protects their good health. The internal consistency of the scale in our sample was good (α = 0.81).

The Coping Self-Efficacy (CSE-7) scale (range 7–49) assesses seven trauma-related coping behaviours.Reference Bosmans, Komproe, van Loey, van der Knaap, Benight and Van der Velden19 A high score implies high confidence in one's ability to cope with potentially traumatic events. The internal consistency of the scale was high in this sample (α = 0.85).

The 10-item Multidimensional Scale of Perceived Social Support (MSPSS) (range 1–7) measures perceived social support on three dimensions (family, friends, significant others).Reference Zimet, Dahlem, Zimet and Farley20 A high score implies good perceived social support. The internal consistency of the instrument was very strong (α = 0.91). Perceived support refers to the subjective experience of being supported.

Statistical analysis

After examining the individual health trajectory plots, measurement occasion (T1, T2 or T3) was chosen as the time metric, ranging from 0 to 2 (pre-assignment to follow-up). Separate growth mixture modelling (GMM) was performed for each of the five health indicators using the four-step GMM approach.Reference Ram and Grimm21 The four steps are (a) problem definition, (b) model specification, (c) model estimation and (d) model selection and interpretation. The specified growth model is a statistical model used to describe individuals’ change over time and examines the between-person differences in those changes in unobservable (i.e. latent) subgroups within a population.Reference Grimm, Ram and Estabrook22

We tested three baseline (single-group) growth curve models – intercept, linear and latent basis – to detect the best representation of health change using the χ² model fit test (P > 0.05), root mean square error of approximation (RMSEA) (<0.05), comparative fit index (CFI) (>0.95), Tucker–Lewis index (TLI) (>0.95) and standardised root mean squared residual (SRMR) (<0.05). A baseline model indicates that it is the simplest model for analysing and understanding the relationships between the different study indicators of the expected changes. It provides a baseline to compare and determine whether more complex models (e.g. by introducing subgroups) better fit the data.

To determine the number of different subgroups (profiles/classes), their longitudinal trajectories and individual profile-membership probabilities, we specified three group difference models for 1–5-profile solutions. The first group difference model is the means model (M2). It allows the means of each group to differ. Second is the means and covariances model (M3), which allows the means, intercept and slope variances and covariance to differ. Third is the means, covariances and residual variances model (M4). It allows all model parameters to differ (the means, intercept, slope variances, covariance and residual variance). Additional detailed information about the terminology can be found in Grimm and colleagues.Reference Grimm, Ram and Estabrook22

The best fitting group difference GMM models were selected using a decision tree based on model convergence, lowest fit indices (Bayesian Information Criterion (BIC); Akaike's Bayesian information criterion (ABIC); Akaike information criterion (AIC); −2 log likelihood (−2LL) difference test), significant bootstrap likelihood ratio test (BLRT) P-value (P < 0.05), entropy value (>0.75) to adequately distinguish between profiles, meaningful profile size (>1%), model interpretation of iHAWs’ health, and model parsimony.

We added the time-invariant covariates (biological sex, assignment duration, and pre-assignment SOC, CSE-7, UWES-9 and MSPSS scores) to each of the selected best fitting group difference GMM models, to determine which predictors were associated with each health trajectory. We used the automatic three-step approach to study these covariates.Reference Grimm, Ram and Estabrook22,Reference Muthén and Asparouhov23 The three steps consist of: (1) an estimation of the latent class growth model using only the latent class indicator variables; (2) assigning participants to their most likely profile based on their latent class posterior distribution; and (3) regressing the most likely profile on each predictor variable while taking into account the misclassification in the second step. The model that includes covariates is called the conditional model.

All analyses were performed in Mplus (version 8 for Windows). The GMMs were estimated using maximum likelihood methods. Missing data were handled using full maximum likelihood. A prior study utilising this sample noted no substantive bias in parameter estimation based on sensitivity analyses using multiple imputation and Little's Missing Completely at Random test.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1

Results

The distributions of the observed health indicator scores were within the parameters of normality, except for PTSD symptoms, which had kurtosis values between 2.6 and 5.6. Switching to a GMM model with non-normal distributionsReference Muthén and Asparouhov23 increased convergence problems and did not outperform the regular normal-distributions GMM models for PTSD symptoms. Thus, we used the regular normal-distributions approach.

Unconditional model

Based on the general linear model (GLM) fit indices (Table 2), the latent basis model was considered the best fitting baseline model for PTSD, emotional exhaustion and anxiety symptomatology. The baseline linear model was the best fit for work engagement and depression indicators. The 1–5-profile (class) solution M2 fit indices are presented in Table 3. A visual representation of the trajectories for each GMM model is shown in Supplementary Fig. 2, available at https://dx.doi.org/10.1192/bjo.2023.58. The slope term variance had to be constrained in all M2 models to counter convergence errors that frequently occur in unconstrained models. M3 and M4 group difference models did not converge or remained unidentified.

Table 2 Fit indices for unconditional baseline general linear modelsa

AIC, Akaike information criterion; BIC, Bayesian information criterion; ABIC, Akaike's Bayesian information criterion; RMSEA, root mean square error of approximation; CFI, comparative fit index; TLI, Tucker–Lewis index; SRMR, standardised root mean squared residual.

a. The optimal model is shown in bold.

b. The slope term variance was constrained to adjust for small non-significant negative variance.

Table 3 Model fit comparisons for the one-, two-, three-, four- and five-profile (class) solutions for each health indicatora

AIC, Akaike information criterion; BIC, Bayesian information criterion; ABIC, Akaike's Bayesian information criterion; BLRT, bootstrapped likelihood ratio test; −2LL difference, two times log-likelihood difference between an n-profile solution and an n − 1 profile solution.

a. The optimal model is shown in bold. Group difference models M3 and M4 are not reported owing to convergence problems for all health indicators. Fixed slope term variance was used to adjust for convergence problems.

b. A number of bootstrap draws had a smaller LRT value than the observed LRT value and therefore the P-value may not be trustworthy.

The relative fit indices of the health indicators indicated a preference for higher-profile solutions, with the 5-profile solutions providing the optimal fit. However, the 5-profile solutions for PTSD, work engagement, anxiety and depressive symptomatology, and the 4-profile solution for work engagement, were unsatisfactory owing to classes accounting for ≤1% of the sample. The 4-profile and 5-profile solutions for work engagement also contained unacceptable out-of-bounds trajectories. The entropy value was acceptable (≥0.75) to good (≥0.80) for all profile solutions, except for the 5-profile solution for work engagement (0.70) and the 2-, 3- and 5-profile solutions for depression (<0.75). Only the 4-profile depression solution had an acceptable entropy value (0.76). Consequently, only the 2- to 4-profile solutions for PTSD and anxiety symptoms, the 2- and 3-profile solutions for work engagement, the 2- to 5-profile solutions for emotional exhaustion and the 4-profile solution for depression were interpreted. We also interpreted the 3-profile solution for depression because of its near acceptable entropy level (0.73) and similarities to the 4-profile solution.

The most meaningful and parsimonious models were presented (Fig 1). Note that the time coefficient (b) for each trajectory has different meanings for linear and latent basis models. For linear models, b_1 = 0 reflects pre-assignment baseline scores, b_2 = 1 reflects the change between pre- and post-assignment scores, b_3 = 2 reflects the change between pre-assignment and follow-up scores. In latent basis models, b_1 = 0 is set to 0, b_3 = 1 is set to 1. They reflect the change between pre-assignment and follow-up scores. The estimated post-assignment slope term b_2 is variable and differs for each latent basis model. We show the latent time basis coefficients in Fig. 1. For each health indicator, its associated T1–T3 health trajectory mean scores and estimated slopes (B) are given in Table 4.

Fig. 1 Optimal model class solutions for each health trajectory. UWES-9, Utrecht Work Engagement Scale-9.

Table 4 Latent trajectory health indicator means and slopesa

PTSD, post-traumatic stress disorder; T1, pre-assignment; T2, post-assignment; T3, follow-up.

a. The unstandardised regression coefficient (B) for the health outcome indicators ‘PTSD’, ‘emotional exhaustion’ and ‘anxiety’ reflects the change between pre-assignment (T1) and follow-up (T3) scores (latent models). The unstandardised regression coefficient (B) for the health outcome indicators ‘work engagement’ and ‘depression’ reflects the change between pre-assignment (T1) and post-assignment (T2), and the change between post-assignment (T2) and follow-up scores (T3) (linear models).

PTSD

For PTSD symptoms, the 4-profile solution was selected as the best model because the additional health trajectories were more informative compared with the 2-profile and 3-profile solutions. The first profile (82% of the sample) was considered ‘healthy/normative’: it had a low pre-assignment severity level and a stable slope trajectory. The second profile (7%, ‘worsening’) had low pre-assignment severity levels and significant pre-assignment to follow-up increases in severity, with moderate subthreshold severity levels at post-assignment and follow-up. The level of PTSD symptomatology continued to rise between post-assignment and follow-up. The third profile (7%, ‘ill health’) had a moderate, subthreshold pre-assignment severity level and a stable slope trajectory. The fourth profile (5%, ‘improving’) had a moderate, subthreshold pre-assignment severity level that significantly decreased between pre-assignment and follow-up. Post-assignment and follow-up severity scores were low, demonstrating that the greatest symptom decrease took place between pre- and post-assignment.

Emotional exhaustion

For emotional exhaustion, the 3-profile solution was considered optimal. It was more informative than the 2-profile solution and more parsimonious than the 4-profile solution. The first profile (77%, ‘healthy/normative’) had low pre-assignment exhaustion scores that remained stable over time. The second profile (17%, ‘ill health’) had moderate severity levels that remained stable over time. The third profile (6%, ‘worsening’) showed a low pre-assignment severity, spiking to a high post-assignment severity level and decreasing to borderline moderate-to-high levels of emotional exhaustion at follow-up.

Work engagement

The linear 3-profile solution was considered the best model for work engagement because it was more informative than the 2-profile solution. It consisted of a ‘healthy/normative’ majority profile (86%), characterised by high engagement scores that slightly, but significantly, decreased over time, remaining highly engaged at follow-up. The second profile (9%) reported an average work engagement score and a stable slope over time. It can be considered a less engaged profile. The third profile (4%, ‘worsening’) exhibited high pre-assignment engagement scores, significantly decreasing to a moderate post-assignment and low follow-up work engagement level. The level of work engagement continued to decrease between post-assignment and follow-up.

Anxiety

The 3-profile solution for anxiety symptoms was considered the best model. It was more informative than the 2-profile model in identifying participants who worsened over time. The 4-profile solution was considered less parsimonious. The first profile (76%, ‘healthy/normative’) had low pre-assignment anxiety severity that slightly, but significantly, decreased to even lower levels over time. The second profile (14%, ‘improving’) had moderate-to-high pre-assignment severity levels and significant slope decreases over time, with post-assignment and follow-up scores dropping to a moderate severity level. The third profile (11%, ‘worsening’) reported low pre-assignment anxiety symptom scores that increased significantly over time, with above clinical threshold post-assignment anxiety symptomatology. At follow-up, anxiety symptomatology decreased although it remained above the clinical threshold. The anxiety ‘improving’ profile differs from the PTSD ‘improving’ profile. At follow-up, anxiety levels were still substantially elevated and above threshold norms, unlike PTSD severity scores.

Depression

The 3-profile solution for symptoms of depression was considered the best model. Although it was slightly below an acceptable entropy value (0.73), it had acceptable to high profile membership probability scores (0.91–0.76). It was considered more parsimonious than the 4-profile solution. The first profile (73%, ‘healthy/normative’) reported low severity, decreasing over time. The second profile (15%, ‘worsening’) reported a subthreshold pre-assignment depression severity that increased over time to above clinical threshold levels at post-assignment and follow-up. The level of depressive symptomology continued to deteriorate between post-assignment and follow-up. The third profile (12%, ‘ill health’) reported stable pre-assignment depression severity levels above clinical threshold.

Conditional model

Table 5 provides an overview of all class membership predictor analyses, their predictive value and log odds. In all cases, the ‘healthy/normative’ profile served as the reference class for each health indicator. We used biological sex, number of prior assignments and assignment duration as covariates, and we examined the predictive value of the pre-assignment sense of coherence (SOC), coping self-efficacy (CSE-7) and perceived social support (MSPSS) scores as health mechanisms to remain healthy in times of (extreme) stress. Supplementary Table 6 (Supplementary Appendix 3) provides the latent trajectory descriptive for all predictors at each measurement.

Table 5 Multinomial logistic regression outcomes for pre-assignment (T1) predictors of class membership for each health indicatora

CI, confidence interval (lower/upper limit); B, unstandardised regression coefficient.

a. The optimal unconditional model for each health indicator was used for predictor analyses. The reference profile is always the majority group, characterised as ‘healthy/normative’.

Biological sex

The odds for being in the anxiety symptoms ‘worsening’ profile versus the ‘healthy/normative’ profile were 3.14 times higher among women compared with men (P = 0.02, OR = 3.14). The odds for being in the symptoms of depression ‘worsening’ profile versus the ‘healthy/normative’ profile were 3.85 times higher among women compared with men (P = 0.01, OR = 3.85). The odds for being in the symptoms of depression ‘ill health’ profile versus the ‘healthy/normative’ profile were 2.31 times higher among women compared with men (P = 0.02, OR = 2.31).

Duration of field assignment

Longer assignment duration was significantly (P = 0.04) associated with an increased risk of being in the PTSD ‘ill health’ profile compared with the ‘healthy/normative’ profile. The odds were 13% higher (OR = 1.13) for each additional month on assignment. Similarly, a longer assignment duration was related to higher odds of being in the symptoms of depression ‘worsening’ profile (P = 0.003; OR = 1.16).

Sense of coherence

Relative to the ‘healthy/normative’ profile, the pre-assignment (low) SOC score was the most consistent predictor of membership of the ‘worsening’ profile for all health indicators. In each instance, a 1 unit increase in SOC score was significantly (P < 0.05) associated with lower odds of an individual belonging to the ‘worsening’ profile (by 4–7%; OR = 0.96–0.93). Compared with the ‘healthy/normative’ reference profiles, higher pre-assignment SOC scores were associated with significantly (P < 0.001) lower odds of an individual belonging to the PTSD and anxiety symptoms ‘improving’ profiles. A 1 unit increase in SOC score was associated with lower odds of membership (by 16% (OR = 0.84) and 10% (OR = 0.90) respectively). Compared with the ‘healthy/normative’ profile, higher SOC scores were associated with lower odds of belonging to the PTSD, emotional exhaustion and depression symptoms ‘ill health’ profiles and the ‘less engaged’ profile for work engagement. A 1 unit increase in SOC score was associated with 17%, 11%, 16% and 9% lower odds respectively.

Coping self-efficacy

Compared with the ‘healthy/normative’ reference profile, higher pre-assignment CSE-7 scores were associated with significantly (P < 0.001) lower odds of an individual belonging to the PTSD and anxiety symptoms ‘improving’ profiles. A 1 unit increase in CSE-7 score was associated with lower odds of membership (by 19% (OR = 0.81) and 14% (OR = 0.86) respectively). Compared with the healthy/normative profile, higher CSE-7 scores were associated with lower odds of belonging to the PTSD, emotional exhaustion and depression symptoms ‘ill health’ profiles and the ‘less engaged’ profile for work engagement. A 1 unit increase in CSE-7 score was associated with 17%, 13%, 15% and 8% lower odds respectively.

Social support

Perceived social support was associated with membership of the depression ‘worsening’ profile. An increase in MSPSS score by 1 unit was significantly (P < 0.05) associated with lower odds of membership of this profile (36% lower (OR = 0.64)). Compared with the ‘healthy/normative’ profile, higher perceived social support was related to lower odds of belonging to the PTSD, emotional exhaustion and depression symptoms ‘ill health’ profiles and the ‘less engaged’ profile for work engagement. A 1 unit increase in MSPSS score was associated with 38%, 38%, 39% and 27% lower odds respectively.

Discussion

This study investigated whether iHAWs experience different assignment-related health trajectories and what variables predict membership of those trajectories. The findings specify how humanitarian aid assignments affect health and which mechanisms are used to protect it.

Health trajectories

All health and work engagement indicators consisted of a healthy and a worsening profile (class). Most health indicators had an ‘ill health’ profile. Only PTSD and anxiety symptoms indicators had an ‘improving’ profile. These findings are consistent with research in other populations exposed to extreme stress,Reference Galatzer-Levy, Huang and Bonanno3 including iHAWs.Reference Greene-Cramer, Hulland, Russell, Eriksson and Lopes-Cardozo24

Most iHAWs remained healthy:Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 the largest trajectory for all health indicators was ‘healthy’ (73–86%). These and other findings in iHAW researchReference Greene-Cramer, Hulland, Russell, Eriksson and Lopes-Cardozo24 and general populations demonstrate the human capacity for dealing with high levels of adversity-related stress.Reference Norris, Tracy and Galea25

A minority of iHAWs (4–15%) worsened on the health indicators during their humanitarian work. PTSD, depressive symptoms and work engagement continued to deteriorate post-assignment, demonstrating that these assignment-related health issues do not resolve quickly. This may be indicative of future pathology (delayed onset). IHAWs’ emotional exhaustion symptoms improved after return from assignment, although they did not return to baseline level. Emotional exhaustion, being a dimension of work-related burnout, may improve post-assignment, as a result of distance from the emotionally intense work environment.Reference Ram and Grimm21

All health indicators, except anxiety, had a stable (moderate) ‘ill health’ profile (7–17%). Some iHAWs improved their pre-assignment elevated levels of PTSD and anxiety during their assignment (respectively: 7%, 14%). The distress from previous assignments and life/work experiences may explain the stable ‘ill health’ trajectories of most health indicators. The improvement of the pre-assignment anxiety levels of some iHAWs, which is likely related to a temporary, mission-related, anticipatory nervousness, may explain the absence of a stable ‘ill health’ profile of anxiety as well as the post-assignment improvement of PTSD symptomatology.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1

Some iHAWs, belonging to the ‘improving’ trajectory, decreased their pre-assignment elevated levels of PTSD and anxiety during their assignment (by 7% and 14% respectively). Pre-assignment anticipatory anxiety may have temporarily increased anxiety levels, only to diminish post-assignment after their actual exposure to the humanitarian emergency.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 Some iHAWs belonged to the PTSD (7%) and anxiety (14%) ‘improving’ trajectories, which were associated with high levels of PTSD and anxiety symptoms pre-assignment that decreased to low levels after their assignment. The ‘improving’ trajectories were also associated with lower levels of pre-assignment sense of coherence and coping self-efficacy (i.e. higher pre-assignment sense of coherence and coping self-efficacy levels were associated with lower odds of belonging to the ‘improving’ trajectory). Given these findings, pre-assignment anticipatory anxiety may have temporarily increased anxiety levels, only to diminish at post-assignment after their actual exposure to the humanitarian emergency.Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 The anticipatory anxiety may have undermined participants’ belief in their ability to understand and manage assignment-related demands, translating into the lower pre-assignment CSE-7 and SOC scores that were associated with membership of the ‘improving’ profile. Or, vice versa, a pre-assignment lack of belief in their ability to understand and manage upcoming assignment-related demands may have increased anticipatory anxiety and (anxiety-related) post-traumatic stress levels. In the latter case, it is hypothesised that coping self-efficacy and sense of coherence levels increased during assignment after better understanding of the humanitarian aid context, leading to post-assignment health gains (decreases in PTSD and anxiety symptoms). Moreover, in accordance with our hypothesis, those with high levels of sense of coherence and coping self-efficacy pre-assignment are more likely to stay healthy (higher odds of membership of the ‘healthy/normative’ trajectory). Considering that high levels of pre-assignment sense of coherence and coping self-efficacy were associated with fewer symptoms of PTSD and anxiety (Reference de Jong, Martinmäki, Brake, Kleber, Haagen and Komproe26 and Supplementary Appendix 1), it is less likely that those who score high on the CSE-7 and SOC have any need for health indicators to improve.

Predictor findings

Higher levels of sense of coherence were associated with the ‘healthy’ trajectory and lower levels were associated with increased probability of belonging to any other profile, regardless of the health indicator in question. This shows the importance of sense of coherence as a mechanism for explaining the health condition of iHAWs.

Consistent with global prevalence rates,Reference Muthén and Asparouhov23 findings among deployed military personnelReference Jones, Jones, Greenberg, Phillips, Simms and Wessely27 and iHAWs,Reference Greene-Cramer, Hulland, Russell, Eriksson and Lopes-Cardozo24 female sex was an important predictor of worsening depression and anxiety trajectories. Higher prevalence of sexual harassment/violence and other negative interactional experiences among female iHAWs,Reference de Jong, Martinmäki, te Brake, Haagen and Kleber1 less control over their jobs,Reference Jones, Jones, Greenberg, Phillips, Simms and Wessely27 hormonal differences and gender-specific cultural expressionsReference Greene-Cramer, Hulland, Russell, Eriksson and Lopes-Cardozo24 may explain this.

Unlike most of the literature on biological sex a predictor of PTSD pathology,Reference Olff28 sex did not predict membership of the PTSD ‘worsening’ trajectory in the present study. The lack of significant findings may be due to the distinction between experiencing symptoms and having a formal PTSD pathology. Another explanation might be that differences in copings strategiesReference Olff28 allow women to cope better with potentially traumatic events in humanitarian aid settings. For example, women are more likely to use tend-and-befriend responses and emotion-focused, defensive and palliative coping strategies. Men are in general more likely to use fight-or-flight responses and problem-focused coping strategies.Reference Olff28

Increased assignment duration was associated with the likelihood of developing assignment-related depressive symptoms. The importance of assignment length has been demonstrated in many occupational groups operating in highly threatening environments, including the military, non-governmental organisations and even the diplomatic core.Reference Dunn, Williams, Kemp, Patel and Greenberg29 Longer assignment means prolonged exposure to high psychological job demands (e.g. excessive workload) and increased potential exposure to traumatic events. These demands may cause chronic stress, wear down individuals’ ability to cope and cause depression.Reference Melchior, Caspi, Milne, Danese, Poulton and Moffitt30 Alternatively, findings on long-term assignments in other populations (navy and astronauts) suggest that depressive symptoms develop as a result of extended isolation and separation from loved ones, loneliness and a lack of sense of belonging.Reference Kanas, Sandal, Boyd, Gushin, Manzey and North31,Reference Kruse, Hagerty, Byers, Gatien and Williams32 Being in close contact with colleagues does not necessarily prevent loneliness or provide adequate received social support.Reference Kruse, Hagerty, Byers, Gatien and Williams32 In the absence of partner and family obligations, single people may be more able to form close relationships during missions, whereas those in committed relationships profit more from their existing partner and family support after missions.Reference Greene-Cramer, Hulland, Russell, Eriksson and Lopes-Cardozo24

Implications

Health monitoring

A minority of iHAWs’ health worsens during aid assignments. Pre- and post-assignment health screenings enable healthcare professionals to detect this minority, as well as to distinguish between individuals with overall ill health and those whose health worsens during assignment. Screening helps iHAWs to reflect on whether they have taken sufficient time to recuperate before accepting new assignments. Monitoring women is important because they were more prone to develop anxiety and depression issues. IHAWs on long-term assignments can best be monitored for early detection of depressive symptoms during the assignment. Multiple post-assignment screenings, a watchful waiting approach, are useful to detect iHAWs at risk, because assignment-related health problems may manifest fully over time. Some health indicators, such as PTSD symptomatology, may also be less likely to remit spontaneously.Reference Nasveld, Cotea, Pullman and Pietrzak33

Strengthening sense of coherence

In the process of staying healthy, meaningfulness is a key component of sense of coherence. To support iHAWs making sense of their work, ongoing communication during assignments on the purpose of the aid work, justification of choices and priorities, a culture of appreciation and management actively seeking feedback from iHAWs are important mechanisms.Reference van den Berg, Bakker and ten Cate34 It is also important to develop interventions that at pre-assignment (e.g. realistic preparation) and peri-assignment (e.g. the fostering of a positive, social working environment) strengthen sense of coherence during field assignments. General interventions that improve sense of coherence and well-being,Reference Vastamäki, Moser and Paul35 such as physical workouts and mindfulness-based meditation practices, including mobile apps,Reference Ando, Natsume, Kukihara, Shibata and Ito36,Reference Kekäläinen, Kokko, Sipilä and Walker37 can be encouraged.

Strengths and limitations

The present study contributes to new insights into the health of iHAWs. We used a prospective design, multiple health indicators and measurement moments, a large sample and advanced statistical techniques to strengthen the quality of findings.

There are also limitations. The study focused on iHAWs and the results cannot be generalised to other groups of aid workers, such as locally contracted staff. We used English language questionnaires. This may have been challenging for non-native speakers, despite being highly educated and proficient in the English language, with onsite support available to clarify the interpretation of the wording and questions. A strength of this study is that it used repeated measures and an emphasis on change scores, rather than on absolute cut-off criteria. Owing to the brief follow-up period, the possibility of a ‘delayed onset’ trajectory, which typically takes a long time to emerge,Reference Norris, Tracy and Galea25 could not be investigated within the present data-set. Sense of coherence was regarded as a stable concept,Reference Antonovsky7 and therefore pre-/post-assignment changes in sense of coherence outcomes were not analysed for the purpose of this study. We cannot exclude the possible impact of the instability of sense of coherence, which might affect both its intensity and nature. Multiple testing increases the risk of false-positive findings (type I errors). Considering the exploratory nature of GMM, no adjustments for multiplicity were made.Reference Li, Taljaard, Van den Heuvel, Levine, Cook and Wells38 Some findings may be attributable to a response shift concerning how iHAWs interpreted the questionnaires. The impact of response shift effects is generally small;Reference Schwartz, Bode, Repucci, Becker, Sprangers and Fayers39 a previous study detected no response shift among iHAWs.Reference Young, Pakenham and Norwood40

Future research could also include more specific non-psychopathological outcome measures, for example quality of life or optimal psychological functioning. Overall, we want to conclude that the present findings show how and why and with what ‘resources’ health workers maintain their health or lose their capacity to remain healthy based on humanitarian aid worker and aid assignment characteristics, as well as several health-promoting theories. The theory of salutogenesis and its associated mechanism of sense of coherence appears to be particularly relevant as a protective mechanism for humanitarian aid workers working in highly stressful and potentially dangerous emergency settings. It enables new insights for possibilities to prevent health worsening and help strengthen iHAWs’ ability to remain healthy under stress.

Supplementary material

Supplementary material is available online at https://doi.org/10.1192/bjo.2023.58.

Data availability

The data that support the findings of this study are available on request from the corresponding author (K.d.J.). The data are not publicly available because they contain information that could compromise the privacy of research participants.

Author contributions

K.d.J.: conceptualisation, methodology, formal analysis, writing original draft, project administration, supervision, funding acquisition. S.E.M.: data curation, data collection, methodology, formal analysis, software, visualisation, writing (review, editing). H.t.B.: methodology, research quality supervision, writing (review, editing). I.K.: conceptualisation, methodology, formal analysis, software, research quality supervision, writing (review, editing). R.J.K.: conceptualisation, methodology, research quality supervision, writing (review, editing). J.F.G.H.: data curation, data collection, methodology, formal analysis, software, writing (review, editing).

Funding

The Management Team of MSF Operational Centre Amsterdam funded this study into the health of its international staff. It was not involved in: study design; collection, analysis, and interpretation of data; writing of the articles; and decisions to submit papers for publication.

Declaration of interest

K.d.J. is employed currently by MSF Operational Centre Amsterdam.

References

de Jong, K, Martinmäki, SE, te Brake, H, Haagen, JFG, Kleber, RJ. Mental and physical health of international humanitarian aid workers on short-term assignments: findings from a prospective cohort study. Soc Sci Med 2021; 285: 114268.CrossRefGoogle ScholarPubMed
Stoddard, A, Harvey, P, Czwarno, M, Breckenridge, M. Aid Worker Security Report 2019. Speakable: Addressing Sexual Violence and Gender-Based Risk in Humanitarian Aid. Humanitarian Outcomes, 2019.Google Scholar
Galatzer-Levy, IR, Huang, SH, Bonanno, GA. Trajectories of resilience and dysfunction following potential trauma: a review and statistical evaluation. Clin Psychol Rev 2018; 63: 4155.CrossRefGoogle ScholarPubMed
Bonanno, GA, Mancini, AD, Horton, JL, Powell, TM, Leardmann, CA, Boyko, EJ, et al. Trajectories of trauma symptoms and resilience in deployed U.S. military service members: prospective cohort study. Br J Psychiatry 2012; 200: 317–23.CrossRefGoogle ScholarPubMed
Van der Wal, S, Gorter, R, Reijnen, A, Geuze, E, Vermetten, E. Cohort profile: the Prospective Research in Stress-Related Military Operations (PRISMO) study in the Dutch Armed Forces. BMJ Open 2019; 9(3): e026670.CrossRefGoogle ScholarPubMed
Veronese, G, Pepe, A. Sense of coherence as a determinant of psychological well-being across professional groups of aid workers exposed to war trauma. J Interpers Violence 2017; 32: 1899–920.CrossRefGoogle ScholarPubMed
Antonovsky, A. Unraveling the Mystery of Health: How People Manage Stress and Stay Well. Jossey-Bass Publishers, 1987.Google Scholar
Lazarus, RS, Folkman, S. Transactional theory and research on emotions and coping. Eur J Personal 1987; 1: 141–69.CrossRefGoogle Scholar
van der Velden, PG, van Loon, P, Benight, CC, Eckhardt, T. Mental health problems among search and rescue workers deployed in the Haïti earthquake 2010: a pre–post comparison. Psychiatry Res 2012; 198: 100–5.CrossRefGoogle ScholarPubMed
Eriksson, CB, Bjorck, JP, Larson, LC, Walling, SM, Trice, GA, Fawcett, J, et al. Social support, organisational support, and religious support in relation to burnout in expatriate humanitarian aid workers. Ment Health Relig Cult 2009; 12: 671–86.CrossRefGoogle Scholar
Blevins, CA, Weathers, FW, Davis, MT, Witte, TK, Domino, JL. The Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5): development and initial psychometric evaluation. J Trauma Stress 2015; 28: 489–98.CrossRefGoogle ScholarPubMed
Bovin, MJ, Marx, BP, Weathers, FW, Gallagher, MW, Rodriguez, P, Schnurr, PP, et al. Psychometric properties of the PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders-fifth edition (PCL-5) in veterans. Psychol Assess 2016; 28: 1379–91.CrossRefGoogle ScholarPubMed
Maslach, C, Jackson, SE, Leiter, MP. Maslach Burnout Inventory: MBI. Consulting Psychologists Press, 1981.Google Scholar
Leiter, MP, Maslach, C. Latent burnout profiles: a new approach to understanding the burnout experience. Burn Res 2016; 3: 89100.CrossRefGoogle Scholar
Schaufeli, WB, Bakker, AB, Salanova, M. The measurement of work engagement with a short questionnaire: a cross-national study. Educ Psychol Meas 2006; 66: 701–16.CrossRefGoogle Scholar
Schaufeli, W, Bakker, A. UWES Utrecht Work Engagement Scale Preliminary Manual. 1.1. Utrecht University: Occupational Health Psychology Unit, 2004.Google Scholar
Derogatis, LR, Lipman, RS, Rickels, K, Uhlenhuth, EH, Covi, L. The Hopkins Symptom Checklist (HSCL). Psychol Measurem Psychopharmacol 1974; 7: 79110.Google ScholarPubMed
Derogatis, LR, Lipman, RS, Rickels, K, Uhlenhuth, EH, Covi, L. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 1974; 19: 115.CrossRefGoogle ScholarPubMed
Bosmans, MWG, Komproe, IH, van Loey, NE, van der Knaap, LM, Benight, CC, Van der Velden, PG. Assessing perceived ability to cope with trauma: a multi-group validity study of a 7-item coping self-efficacy scale. Eur J Psychol Assess 2017; 33: 5564.CrossRefGoogle Scholar
Zimet, GD, Dahlem, NW, Zimet, SG, Farley, G. The multidimensional scale of perceived social support. J Pers Assess 1988; 52: 3041.CrossRefGoogle Scholar
Ram, N, Grimm, KJ. Growth mixture modeling: a method for identifying differences in longitudinal change among unobserved groups. Int J Behav Dev 2009; 33: 565–76.CrossRefGoogle ScholarPubMed
Grimm, KJ, Ram, N, Estabrook, R. Growth Modeling Structural Equation and Multilevel Modeling Approaches. Guilford Press, 2016.Google Scholar
Muthén, B, Asparouhov, T. Growth mixture modeling with non-normal distributions. Stat Med 2015; 34: 1041–58.CrossRefGoogle ScholarPubMed
Greene-Cramer, BJ, Hulland, EN, Russell, SP, Eriksson, CB, Lopes-Cardozo, B. Patterns of posttraumatic stress symptoms among international humanitarian aid workers. Traumatology 2021; 27: 177–84.CrossRefGoogle Scholar
Norris, FH, Tracy, M, Galea, S. Looking for resilience: understanding the longitudinal trajectories of responses to stress. Soc Sci Med 2009; 68: 2190–8.CrossRefGoogle ScholarPubMed
de Jong, D, Martinmäki, S, Brake, HT, Kleber, R, Haagen, J, Komproe, I. How do international humanitarian aid workers stay healthy in the face of adversity? PLoS One 2022; 17(11): e0276727.CrossRefGoogle ScholarPubMed
Jones, N, Jones, M, Greenberg, N, Phillips, A, Simms, A, Wessely, S. UK military women: mental health, military service and occupational adjustment. Occup Med 2020; 70: 235–42.CrossRefGoogle ScholarPubMed
Olff, M. Sex and gender differences in post-traumatic stress disorder: an update. Eur J Psychotraumatology 2017; 8(suppl 4): 1351204.CrossRefGoogle Scholar
Dunn, R, Williams, R, Kemp, V, Patel, D, Greenberg, N. Systematic review: deployment length and the mental health of diplomats. Occup Med (Lond) 2015; 65(1): 32–8.CrossRefGoogle ScholarPubMed
Melchior, M, Caspi, A, Milne, BJ, Danese, A, Poulton, R, Moffitt, TE. Work stress precipitates depression and anxiety in young, working women and men. Psychol Med 2007; 37: 1119–29.CrossRefGoogle ScholarPubMed
Kanas, N, Sandal, G, Boyd, JE, Gushin, VI, Manzey, D, North, R, et al. Psychology and culture during long-duration space missions. Acta Astronaut 2009; 64: 659–77.CrossRefGoogle Scholar
Kruse, JA, Hagerty, BM, Byers, WS, Gatien, G, Williams, RA. Considering a relational model for depression in Navy recruits. Mil Med 2014; 179: 1293–300.CrossRefGoogle ScholarPubMed
Nasveld, P, Cotea, C, Pullman, S, Pietrzak, E. Effects of deployment on mental health in modern military forces: a review of longitudinal studies. J Mil Vet Health 2012; 20: 2436.Google Scholar
van den Berg, BAM, Bakker, AB, ten Cate, T. Key factors in work engagement and job motivation of teaching faculty at a university medical centre. Perspect Med Educ 2013; 2: 264–75.CrossRefGoogle Scholar
Vastamäki, J, Moser, K, Paul, KI. How stable is sense of coherence? Changes following an intervention for unemployed individuals. Scand J Psychol 2009; 50: 161–71.CrossRefGoogle ScholarPubMed
Ando, M, Natsume, T, Kukihara, H, Shibata, H, Ito, S. Efficacy of mindfulness-based meditation therapy on the sense of coherence and mental health of nurses. Health (N Y) 2011; 3: 118–22.Google Scholar
Kekäläinen, T, Kokko, K, Sipilä, S, Walker, S. Effects of a 9-month resistance training intervention on quality of life, sense of coherence, and depressive symptoms in older adults: randomized controlled trial. Qual Life Res 2018; 27: 455–65.CrossRefGoogle ScholarPubMed
Li, G, Taljaard, M, Van den Heuvel, ER, Levine, MA, Cook, DJ, Wells, GA, et al. An introduction to multiplicity issues in clinical trials: the what, why, when and how. Int J Epidemiol 2017; 46: 746–55.Google ScholarPubMed
Schwartz, CE, Bode, R, Repucci, N, Becker, J, Sprangers, MAG, Fayers, PM. The clinical significance of adaptation to changing health: a meta-analysis of response shift. Qual Life Res 2006; 15: 1533–50.CrossRefGoogle Scholar
Young, TK, Pakenham, KI, Norwood, MF. Thematic analysis of aid workers’ stressors and coping strategies: work, psychological, lifestyle and social dimensions. J Int Humanit Action 2018; 3: 19.CrossRefGoogle Scholar
Figure 0

Table 1 Participant information (n = 609)

Figure 1

Table 2 Fit indices for unconditional baseline general linear modelsa

Figure 2

Table 3 Model fit comparisons for the one-, two-, three-, four- and five-profile (class) solutions for each health indicatora

Figure 3

Fig. 1 Optimal model class solutions for each health trajectory. UWES-9, Utrecht Work Engagement Scale-9.

Figure 4

Table 4 Latent trajectory health indicator means and slopesa

Figure 5

Table 5 Multinomial logistic regression outcomes for pre-assignment (T1) predictors of class membership for each health indicatora

Supplementary material: File

De jong et al. supplementary material

De jong et al. supplementary material

Download De jong et al. supplementary material(File)
File 124 KB
Submit a response

eLetters

No eLetters have been published for this article.