An accident is defined as a sudden, unintentional, harmful impact of an unusual external factor on the human body resulting in impairment of physical, mental and psychological health, according to both the Swiss National Insurance Fund (Suva) and Swiss law. This definition includes recreational and work-related accidents but distinguishes accidents from illness and other forms of injury (Egli, 2018). Although many people who suffer an accident recover well and quickly, a significant proportion experience decreased well-being, prolonged working disability and emotional distress even in cases of mild-to-moderate accident-caused injuries (Kendrick, Coupland, et al., 2017; Kendrick, Kelllezi, et al., 2017). However, injured people often receive only fragmented care (Kendrick, Kelllezi, et al., 2017). Since accident-caused injuries may encompass numerous issues and can lead to just as many physical and psychological sequelae, there is a need for more flexible and individually tailored treatment options to meet the heterogeneity of complications.
The results of previous injury rehabilitation studies suggest that interventions should promote collaborative care, thereby broadening the treatment focus and applying a holistic biopsychosocial perspective (e.g. Bültmann et al., 2009; Cullen et al., 2018; Zatzick et al., 2004). This is further supported by findings that injured or ill workers can best return to work when involved individuals and stakeholders work collaboratively (Russell & Kosny, 2019).
In an effort to address this need, we conducted a randomised controlled trial (RCT) to investigate the efficacy of a highly tailored psychological counselling intervention (Hegy et al., 2021). Despite applying collaborative care and tailoring, we only found significant improvement in one of the five assessed domains of well-being: participants in the intervention group (IG) showed a significant decrease in negative feelings up until 18 months post-injury, with a moderate effect size (d = 0.74), compared with the participants in the control group (CG). Due to this overall rather limited effect, we decided to investigate possible moderators of the treatment. In addition to examining an intervention's effectiveness, the question of potential moderators, that is, what works best for whom, is a key aspect of intervention research (Grawe, 1997, 2004; Kraemer et al., 2002; Tornås et al., 2019). Knowledge of patient characteristics that moderate treatment outcomes could help personalise psychosocial rehabilitation treatment. However, to the best of our knowledge, no studies have evaluated treatment moderators for psychosocial rehabilitation interventions for the heterogeneous population of mild to moderately injured workers. In accordance with the biopsychosocial model of disability (Wade & Halligan, 2004, 2017), the evaluation of treatment moderators may provide unique, new and valuable information to guide further treatment decisions.
An additional factor supporting the examination of potential treatment moderators is the high attrition rates often reported in injury rehabilitation interventions (Giummarra et al., 2018; De Silva et al., 2009; Tecic et al., 2011). For example, in a review of five studies of psychosocial injury rehabilitation interventions, De Silva et al. (2009) reported attrition rates ranging from 47% to 66%. To prevent early treatment termination, the authors recommend conducting a reliable screening of the injured individuals and gaining a deeper understanding of differential treatment effects.
We implemented both of these recommendations, with the recommendation to gain a deeper understanding of treatment effects constituting the aim of the current study. More specifically, we examined moderators of treatment outcome by means of secondary exploratory analyses of the data of our aforementioned RCT (Hegy et al., 2021). Due to the lack of studies regarding moderators of treatment success of injury rehabilitation interventions, we adopted a hypothesis-generating approach with an exploratory analysis. We selected seven well-established predictors of adaptation to health-related adversities and of psychosocial treatment success that could generate specific hypotheses for further studies of differential treatment effects in injury rehabilitation (Livneh & Martz, 2014; Skogstad et al., 2014; Tough et al., 2017; Vassend et al., 2011). Those seven predictors consisted of five coping styles, dispositional optimism and dispositional pessimism.
Coping has been shown to influence the relationship between stressful life events and physical and psychological functioning by mitigating how a stressful life event is perceived and handled (Archer et al., 2019; Higgins & Endler, 1995; Langford et al., 2017; Tein et al., 2000). Since all people encounter challenges at some point in their lives, the way in which stressful events are dealt with and, related to this, how well-being is achieved or regained, is of great importance (Marroquín et al., 2017). In their seminal work, Lazarus and Folkman define coping as ‘constantly changing cognitive and behavioral efforts to manage specific external and internal demands that are appraised as taxing or exceeding the resources of the person’ (Lazarus & Folkman, 1984, p. 141). Based on this definition, it follows that coping can take different forms. These different forms are referred to as coping styles. Thus, we examined different coping styles, more specifically (a) task orientation, (b) emotion orientation and (c) avoidance orientation: social diversion as possible moderators of treatment on psychological well-being.
Coping styles can be influenced by different factors such as personality dispositions and traits, personal resources and beliefs about the self and the world (Lazarus, 2006). Defined as a personality trait reflecting a favourable orientation to the future (Carver et al., 2010), dispositional optimism has been found to be a resource for different work and health-related factors. For example, higher levels of dispositional optimism have been associated, among other things, with improved psychological functioning, adjustment following injury and earlier return to work (Cancelliere et al., 2016; Myhren et al., 2010; Wadey et al., 2013). Therefore, we decided to assess generalised dispositional optimism and its counterpart, generalised dispositional pessimism, as the sixth and seventh possible moderators.
2 METHODSThe data used in the current study were collected in an RCT investigating the effects of a tailored multidisciplinary counselling intervention with the aim to support the adjustment process of injured workers (Hegy et al., 2021). We obtained ethics approval from the Ethics Committee of the University of Bern (No. 2011-04-172) and registered the study at the ISRCTN registry (ISRCTN05534684). The Clinical Trial Unit Bern, an independent national clinical trial management facility to coordinate patient-oriented clinical research, monitored and assessed the study.
2.1 Recruitment and eligibility criteriaThe study population consisted of German-speaking adult workers (≥18 years) who suffered an accident within three months prior to study participation. Participants were consecutively recruited in the main agency of Suva, the largest accident insurance company in Switzerland, with an average coverage of about 50% of all employees. Suva case managers were requested to screen all eligible claimants for the risk of a complicated rehabilitation process with an evaluated screening tool (Abegglen et al., 2017) within the first three months post-injury. Claimants were excluded if they were suffering from (a) severe injuries (e.g. head or spinal cord injuries), (b) occupationally related illnesses (e.g. pulmonary illness) or (c) degenerative conditions (e.g. rheumatoid arthritis). To be included, claimants further had to (a) be at least 18 years old, (b) have a working disability causing a complete working incapacity, (c) have a permanent employment contract and (d) live no more than 20 km away from Berne to ensure convenient accessibility to the intervention.
2.2 ProcedureSuva claimants whose screening showed an increased risk for a complicated rehabilitation process were asked to participate in the study. Of those, claimants who gave written informed consent, fulfilled all inclusion criteria and did not fulfil any of the exclusion criteria were eligible to participate in the RCT and thus randomised to either the IG or CG. Participants in the CG received only conventional case management according to Suva's case management procedure (Scholz et al., 2016), which comprised the standard treatment (care as usual, CAU). Trained and experienced case managers provided support and personal assistance in all aspects of rehabilitation and work reintegration, with the primary aim of a fast and long-lasting work reintegration. In addition to CAU, participants in the IG also received a tailored counselling intervention and collaborative care.
The intervention was created individually for each participant based on the screening results and thus tailored to their requirements. If the screening results mainly indicated work-related distress, the participant received occupational counselling, which consisted of work-related diagnostics and a discussion of the participant's life and work-related goals, followed by a structured observation of the workplace and tailored job counselling. If the screening results mainly indicated psychological distress, the participant received mental health counselling, which consisted of integrative counselling including educational, cognitive and behavioural elements to support the psychosocial adaptation process to the accident-caused injury. If the screening indicated both work-related and psychological distress, the participant received both occupational and mental health counselling. Both the occupational and the mental health intervention focused on individual resource activation (Flückiger et al., 2010; Grawe, 2004) and life goal setting (Rose et al., 2017).
After the randomisation, participants were asked to complete the baseline questionnaire (T0) and were assessed again 12 (T1) and 18 months (T2) post-injury.
2.3 OutcomesWe assessed five different aspects of subjective well-being as main outcomes. The first two aspects of subjective well-being were life satisfaction and negative feelings, which we both assessed with the two uncorrelated subscales of the Bern Questionnaire on Well-Being, adult form (BSW/A; Grob et al., 1991). Items are rated on 6-point Likert scales and 4-point Likert scales. The questionnaire has obtained satisfactory psychometric qualities concerning stability and validity (Grob et al., 1991). The internal consistency of the subscales is satisfactory, with Cronbach's α = 0.82 (life satisfaction) and α = 0.77 (negative feelings). As a third aspect of subjective well-being, we assessed job satisfaction by a single item of the Short Job Satisfaction Questionnaire (AZK; Baillod & Semmer, 1994): ‘If there is no change of my work conditions sooner or later, I will look for a new job’. The answer was rated on a 7-point Likert scale. To assess the fourth and fifth aspects of subjective well-being, namely family-related satisfaction and health-related satisfaction, we used the two corresponding subscales of the Rehab Status Questionnaire Version 3 (IRES-3; Bührlen et al., 2005). All items of these two subscales are rated on 5-point Likert scales with high scores indicating lower family-related satisfaction and higher health-related satisfaction, respectively. The internal consistencies of all the questionnaire's subscales range from good to very good, with Cronbach's α between 0.75 and 0.94.
Of particular relevance to the present study are the potential moderators of treatment outcomes that were assessed. These included different coping styles and generalised dispositional optimism and pessimism. We assessed three different coping styles: (a) task-oriented coping (Cronbach's α = 0.83), (b) emotion-oriented coping (Cronbach's α = 0.80) and (c) avoidance by social diversion (Cronbach's α = 0.80) with the German short version of the Coping Inventory for Stressful Situations (CISS; Kälin, 1995). Participants rated the extent to which they use these coping styles with 18 items using 5-point Likert scales. The generalised dispositional optimism and pessimism were assessed with the German Version of the Life Orientation Test Revised (LOT-R; Glaesmer et al., 2008). The LOT-R consists of 10 items that are rated on a 5-point Likert scale, of which 3 items each are analysed for optimism (Cronbach's α = 0.69) and pessimism (Cronbach's α = 0.59), respectively. The rest are filler items.
2.4 Statistical analysisParticipants' characteristics were calculated at baseline using means and standard deviations. Following the Consolidated Standards of Reporting Trials (CONSORT), analyses were performed according to an intention-to-treat principle using all available data from all randomised participants (Chambless & Hollon, 1998). To accommodate between and within effects considering missing data and unequal numbers of observations, we fitted linear mixed models to the longitudinal measures of outcomes (Singer & Willett, 2003). At level I, the within-person level, time was specified using the measurement points: the baseline measurement (4–6 months after injury) was defined as 0, the post-measurement (12 months after injury) was defined as 1, and follow-up measurement (18 months after injury) was defined as 2. By doing so, the intercept could be interpreted as an outcome score at the baseline measurement. At level II, the between-person level, treatment conditions were specified as 0 for the CG and 1 for the IG. The analyses were conducted in R Statistical Language with the R package nlme (Pinheiro et al., 2021) using full maximum likelihood estimation. The normal distribution of the outcome variables was confirmed by inspecting the residual diagnostics of the fitted models.
For each outcome variable, the analysis proceeds through different steps according to the techniques described by Tasca and Gallop (2009). First, we estimated a null model (intercept-only model), which allowed an estimation of the proportion of variation between and within persons in the outcome variable. Then, we examined the within-person trajectories of change across sessions with the first model (unconditional growth model with random intercept). The second model (conditional growth model with random intercept and cross-level interaction) allowed us to examine the effect of the study conditions, that is, to evaluate whether the different study conditions had different rates of change across the three assessments.
Subsequent exploratory models were used to examine whether individual coping abilities and dispositional optimism and pessimism moderated the treatment efficacy of the intervention compared with the CG. For this purpose, we fitted four separate multilevel models for the subscales of the CISS, and two separate models for the two subscales of the LOT-R. The moderator variables were grand-mean-centred to create a meaningful null point. All these models include the main effect of (a) the respective moderator, (b) time, (c) condition, (d) all three two-way interactions and (e) the three-way interaction of the respective moderator variable with condition and time. To show that a variable is a moderator of the treatment success, this variable must not be correlated with the treatment (Beutler et al., 1991). Table 1 shows baseline values of the putative moderators. Our analyses revealed no significant differences between the two groups.
TABLE 1. Demographics and clinical characteristics of participants CG IG t (df) χ2 (df) p Age Mean (SD) 50.50 10.353 49.04 10.362 0.94 (166.7) 0.35 Gender (%) Female 31 31.00 23 25.00 0.58 (1) 0.45 Male 69 69.00 69 75.00 Level of education (%) No high school diploma 70 70.70 64 69.56 0.70 (2) 0.71 High school and above 25 25.25 26 28.26 Others 4 4.05 2 2.17 Missing values 1 0 Annual income at baseline Up to CHF 40,000 6 6.19 7 8.14 4.34 (4) 0.36 Up to CHF 60,000 21 21.65 16 18.60 Up to CHF 80,000 30 30.92 30 34.88 Up to CHF 100,000 27 27.84 15 17.44 Over CHF 100,000 13 13.40 18 20.93 Missing values 3 14 Occupational classification (%) Blue-collar worker 60 60.60 64 69.57 1.31 (1) 0.25 White-collar worker 39 39.39 28 30.43 Missing values 1 0 Accident type (%) Recreational 70 76.09 60 67.42 1.28 (1) 0.26 Work-related 22 23.91 29 32.58 Missing values 8 3 Outcome variables Well-being (BWQ) at baseline Life satisfaction Mean (SD) 4.58 0.70 4.43 0.80 1.28 (176) 0.20 Negative feelings Mean (SD) 2.68 0.83 2.73 0.83 0.35 (160) 0.72 Job satisfaction (AKZ) at baseline Mean (SD) 4.73 1.18 4.70 1.19 0.18 (172.4) 0.86 Family-related satisfaction (IRES) at baseline Mean (SD) 3.16 0.92 3.12 0.87 0.28 (182) 0.78 Health-related satisfaction (IRES) at baseline Mean (SD) 4.13 0.70 4.07 0.61 0.64 (183.7) 0.52 Moderator variables Coping abilities (CISS) at baseline Task-orientated Mean (SD) 3.77 0.59 3.82 0.55 0.50 (181.7) 0.62 Emotion-orientated Mean (SD) 2.45 0.75 2.556 0.67 −0.96 (178.6) 0.34 Avoidance Mean (SD) 2.51 0.68 2.44 0.72 0.631 (177.6) 0.53 Distraction Mean (SD) 1.96 0.77 1.90 0.75 0.54 (181) 0.59 Social diversion Mean (SD) 3.06 0.87 3.01 0.90 0.43 (180.9) 0.67 Optimism (LOT-R) at baseline Optimism Mean (SD) 8.92 2.14 8.50 2.33 1.28 (177.9) 0.20 Pessimism Mean (SD) 4.72 2.23 5.08 2.39 −1.05 (179.2) 0.30 Note CG = control group (n = 100), IG = intervention group (n = 92), comparison between CG and IG is performed by two-sided Welch's t test for continuous data and Yates continuity correction for the chi-squared test for categorical variables. Degrees of freedom (df) (of the respective test) are given in italics.In case of a significant three-way interaction, we plotted the adjusted means of the subgroups to facilitate the interpretation of this effect. To guide our interpretation, we further conducted simple slopes analyses to test which slope differed significantly from zero (Preacher et al., 2006). We also conducted post hoc tests of the mean differences of these interactions for Time × Condition one standard deviation above (i.e. high level) and below (i.e. low level) the mean of the moderator, using the R package phia (De Rosario-Marinez et al., 2015). These follow-up analyses serve to illustrate the specific nature of the interactions.
We estimated all models as linear because of the sparse number of measurement points (Singer & Willett, 2003). The slopes in all models were fixed, as no model yielded significantly lower global fit indices when including random slopes. As a global effect size, we calculated Nagelkerke's pseudo-R2 statistics, and as a local effect size, we calculated Cohens' d.
3 RESULTSThe majority of the 192 participants of the final sample were male (n = 138; 71.9%) with a mean age of 49.8 years (SD = 10.4). Of the randomised participants whose screening results suggested mental health counselling (n = 75), 42 participants (56%) refused to participate. The remaining 33 participants received an average of 2.23 (SD = 6.94) mental health counselling sessions of approximately 50 min duration per session. Of the 35 participants whose screening results indicated a work-related high-risk profile, 30 participants (85.7%) received one session of occupational counselling and, if the employer agreed, a structured observational analysis of the workplace. We found no significant association with any sociodemographic variables or non-compliance. Figure 1 shows a CONSORT diagram of the flow of participants throughout the study.
CONSORT flowchart of participants
Of the five evaluated coping styles, only social diversion and emotion-oriented coping were significant moderators of treatment success. We found that social diversion moderated the effect of treatment condition on changes in life satisfaction (b = −0.10, SE = 0.048, p = .045; Table 2). This model explained 52% of the variance (pseudo-R2, adjusted by Nagelkerke).
TABLE 2. Results of the multilevel models for change in life satisfaction across time and conditions and significant moderators Fixed effectsModel 1
Unconditional growth model
Model 2
Conditional growth model (cross-level interaction)
Model 3
Conditional growth model (treatment and moderator)
B SE B t B SE B t B SE B t Intercept γ00 4.48 0.06 81.37*** 4.54 0.08 59.39*** 4.54 0.08 58.45*** Treatment γ01 −0.11 0.11 −1.03 −0.11 0.11 −1.01 Time γ10 −0.01 0.02 −0.56 −0.05 0.03 −1.70 −0.05 0.09 −1.83† Social diversion γ02 0.10 0.09 1.12 Time × Treatment γ11 0.08 0.04 1.88† 0.09 0.03 2.02* Time × Diversion γ12 0.03 0.04 0.91 Treatment × Diversion γ03 0.17 0.03 1.32 Time × Treatment × Diversion γ13 −0.10 0.13 −2.01* Random effects Intercept σ20 0.44 [0.35–0.55] 0.44 [0.35–0.55] 0.44 [0.35–0.56] Residual σ2ε 0.15 [0.13–0.18] 0.15
留言 (0)