Task engagement as a mediator between the cognitive demands of sustained attention and musculoskeletal complaints: A structural equation modelling approach

Abstract

BACKGROUND:

Cognitive demands in the form of sustained attention are prevalent in automated manufacturing where operators undertake monitoring tasks for prolonged periods. Sustained attention tasks are stressful and could have potential health consequences for employees including contributing to Work Related Musculoskeletal Disorders (WRMSDs).

OBJECTIVE:

The study investigates if lowered task engagement (low task motivation/focus) is a mediator of the relationship between sustained attention and musculoskeletal complaints.

METHOD:

A structural equation modelling technique was used to determine if task engagement mediates the relationship between attention demand/attention supply and self-reported musculoskeletal complaints (MSCs) in manufacturing employees (n = 235).

RESULTS:

Study participants were 5.16 times (OR, odds ratio) more likely to have neck complaints, 7.27 (OR) times more likely to have upper back complaints and 3.9 (OR) times more likely to report lower back complaints (OR 2.05) when attention demands were high and task engagement was low. When task engagement was introduced as a mediator between attention supply and MSCs, odds ratios decreased for neck (from 1.01 to 0.60), shoulder (from 0.95 to 0.47) (p < 0.01), upper back (from 1.01 to 0.70) (p < 0.01) and lower back (from 0.94 to 0.40)^*.

CONCLUSION:

Task engagement was a significant mediator of the relationship between attention demand and neck, upper back and lower back musculoskeletal complaints for study participants.

Keywords

Occupational ergonomics stress vigilance

1 Introduction

Manufacturing is going through a renaissance due to advancements in technology, resulting in a higher level of complexity within automation design [1]. This has directed changes in working conditions for employees, especially process operatives who often spend much of their work shift monitoring an automated process via a computer interface. This monitoring work requires a high level of mental workload in the form of sustained attention [2].

1.1 Sustained attention

Sustained attention or vigilance refers to the ability to maintain focus of attention and remain alert to stimuli over prolonged periods of time [3]. There are two main opposing theories, which have been used to explain the concept of sustained attention. The first theoretical viewpoint deems sustained attention tasks as being under stimulating and placing low demands on cognitive resources. The mindlessness model [4] is based on this theory and posits that sustained attention tasks are monotonous, leading to mind-wandering. The alternate attention depletion theory, which is increasing in popularity, suggests that sustained attention tasks demand a high level of cognitive resources and deplete these resources quickly. This theory forms the basis of this study and is based on the attentional resources model [5] which postulates that monitoring tasks require extensive mental workload. Research has also proven that sustained attention tasks are stressful for humans [6]. It is therefore possible that the stress due to sustained attention work may contribute to workplace health conditions such as WRMSDs.

1.2 Work related musculoskeletal disorders

Across the 27 EU Member States, WRMSDs represent the most common work-related health disorders. These disorders have been defined as impairments of body structures such as muscles, joints, tendons, ligaments, nerves, bones or localised blood circulation system caused or aggravated primarily by the performance of work and by the effects of the immediate work environment [7]. Mechanisms that may contribute to WRMSDs include exposure to physical (e.g. repetitive work, awkward postures) and psychosocial factors (e.g. high job demands, low job control) at work as well as individual factors (e.g. age, gender). Several models postulate mechanisms that help to explain how psychosocial stressors contribute to WRMSDs. Smith and Smith et al. [8] suggest that physiological stress responses including increased blood pressure, corticosteroids, muscle tension and decreased immune response, are associated with the aetiology of WRMSDs. Other theories [9] propose that increased muscular tension and activity due to stress plays a role in the development of WRMSDs particularly in the upper body affecting the neck, shoulder, upper back and lower back [10, 11]. While there are many other theories, no singular theory is universally accepted.

1.3 Occupational stress

Work-related stress was found to be the second most common occupational health problem across the EU15 [12]. A poor psychosocial climate at work due to psychosocial stressors can contribute to the development of many health conditions, among these being work-related musculoskeletal disorders [13]. Psychosocial stressors include excessive work demands, low job control and high job uncertainty. Sustained attention work as a dominant and often demanding aspect of job content in modern manufacturing environments can be viewed as an occupational psychosocial factor, but research has been carried out to determine its effects on employee health.

1.4 Task engagement

Visual vigilance tasks have been found to lower task engagement [14]. For the purposes of this study, task engagement is defined as temporary engagement in a task or activity within a work context. Task engagement positively and significantly correlates with performance on attention tasks [15]. Key factors of high task engagement include challenge, task interest, personal control and positive feedback, while factors of low task engagement include monotony, long task duration, system automation and passive fatigue [16]. Disengagement (i.e. withdrawing from involvement in activity) can occur where lower levels of attentional resources are required by the task [17], such as with highly automated processes. Disengagement can in turn result in a decline in performance efficiency through lowered attention on the task which is an additional source of stress for the operator [18]. Monitoring of highly automated processes in modern manufacturing environments is often a monotonous task resulting in lowered task engagement. Multivariate analysis suggests a mediating role of task engagement between sustained attention and stress [19], but this has not been tested by a statistical model in an industrial context.

1.5 Study aims

This study aims to investigate the effects of attention demand and attention supply on neck, shoulder, upper back and lower back complaints. Using a structural equation modelling approach to do this, allows inclusion of task engagement as a mediator. A mediation model of this nature has not been previously employed to link attention to musculoskeletal complaints. The presence of task engagement as a necessary mediating variable within this relationship helps to emphasize the importance and role of this construct in job design.

Hypothesis:

Task engagement mediates the relationship between attention supply/attention demand and upper body musculoskeletal complaints in industrial workers.

2 Method

2.1 Survey background

A cross-sectional survey (described below) was administered across five companies: one medical devices, one semi-conductor, one electronics and two pharmaceutical. Companies were selected on the basis that they had high levels of automation. While the production layout and tasks varied across companies, each employee surveyed worked 12–hour shifts and undertook monitoring-based work for a minimum of 70% of the work shift. In total, 235 individuals (188 male, 47 female), completed the hard copy questionnaire fully. An additional 31 incomplete questionnaires were excluded from analysis. There was an average response rate of 80%, with the main reason for non-completion being insufficient time. Convenience sampling [20] was undertaken due to restricted access to employee information. Convenience sampling (also known as availability sampling) is a specific type of non-probability sampling method that relies on data collection from population members who are conveniently available to participate in study. Participants were asked to partake in the survey on a voluntary basis. All parts of the survey were approved by the Ethics Committee, University of Limerick, Ireland with all participants giving their consent totake part.

2.2 Procedure

Access was given to the researchers to distribute the survey to the employees that had consented to participate. Questionnaires were distributed and collected in person by the researcher (F.W.). A paper version of the questionnaire was completed by each participant at the end of their shift. Participants were given adequate time to complete the questionnaire which took 10–15 minutes for most people. Numbers or names were not included on the questionnaires to ensure data was not attributed to specificparticipants.

2.3 Companies surveyed

The 105 respondents from the semi-conductor company were based in a control room completely removed from the automated process. The seated workstations each had 4-5 computer screens from which workers monitored the manufacturing process for 100% of the working shift. Physical tasks (beyond sitting/standing while monitoring screen) were not carried out by this group.

The electronics company employees (34 responses) had sit/stand workstations located along the process assembly line and they alternated between standing and seated position during the shift. Each workstation had 1-2 computer screens through which the operator monitored that section of the process. Monitoring took place for approximately 70% of the working shift.

The medical devices company (49 responses) had seated workstations with 2–4 computer screens located in front of the main production line. The operators spent approximately 70% of the time monitoring the process.

Both pharmaceutical companies (39 + 7 responses) had control rooms located within the area of the process being monitored. Each control operator was seated and attended to between 2 and 4 computer screens, for approximately 75% of the work shift.

In the case of the electronics, medical devices and pharmaceutical companies, light physical tasks were undertaken along with the monitoring work. This mostly involved light maintenance and troubleshooting tasks (e.g., inserting new labels cartridges into the process equipment, and physical inspection of finished products). In the pharmaceutical company, the light physical work included opening and closing valves and taking samples from the process.

2.4 Survey inclusion criteria

The questionnaire was distributed to personnel who met the following inclusion criteria:

Participants were between 18 and 65 years of age,

Participants spent at least 70% of their working shift monitoring an automated manufacturing process via a computer.

Participants were employed in the respective companies for at least 12 months

2.5 Survey questionnaires

A survey was compiled, encompassing validated questionnaires to measure the following variables of interest:

Perceived musculoskeletal complaints were measured using an amended version of the standardised Nordic Musculoskeletal Questionnaire (NMQ) [21]. This questionnaire allows comparison of low back, shoulder, neck and general complaints in occupational study groups and the NMQ reliability has been found to be acceptable [21, 22]. For this questionnaire, participants indicated whether they have had trouble with a particular area in their body in the past 12 months by ticking a check box to indicate ‘Yes’ or ‘No’.

Task engagement was measured in the industrial study using one subscale of the Short State Stress Questionnaire (SSSQ) [23]. The SSSQ is a 24-item scale based on the longer 96-item Dundee State Stress Questionnaire (DSSQ) [24], which has been widely used to measure the stress associated with sustained attention tasks. Matthews et al. [16] proposed a state-mediation model based on transactional stress approaches, in which stressors influence stress states and these states mediate the relationship between stressors, cognition and the DSSQ [25]. The results of the task engagement subscale from the SSSQ were analysed using confirmatory factor analysis.

As there is no validated test to specifically measure sustained attention in an industrial setting, attention measures from a validated situational awareness scale were used. Attention demand and supply were measured using two subscales of the situational awareness to response test (SART). This is a self-rating subjective measure of situational awareness (SA) developed by Taylor [26]. The questions within the attention subscales are particularly relevant to operators monitoring automated processes. Attentional demand refers to the demand placed on attentional resources and encompasses the instability, variability and complexity of the situation. Attention demand is a useful measure of sustained attention in that, increasing the attentional demands of a task results in deterioration of sustained attention performance [4].

Attentional supply, which refers to the amount of attention resources required by a task, includes measures of stimulation, spare mental capacity, concentration and division of attention. This is essentially the attention-related mental workload required to complete the tasks. The ability to perform sustained monitoring tasks depends on the stimulation and concentration of the individual, making attention supply a valid measure in this context [27].

Salmon et al. [28] found that the SART method was useful in relation to subjective assessments measuring how operators felt during task performance in relation to the supply, demand and understanding of information. Participants are asked to rate each test dimension on a Likert scale of 1 to 7. The scores from three questions relating to attention demand were averaged for each participant with the total potential score for attention demand of 21. The scores for attention supply were based on averaging the results of four questions with the total score for attention supply of 28.

Data were analysed using SPSS version 22.0 and Mplus version 7. Although observed variables were non-normally distributed, skewness (+/–1) and kurtosis (+/–3) were slight and sample size was large enough to prevent a reduction in overall power [29]. A series of regression models were specified and tested to determine if task engagement was a significant mediator of the relationship between attention demand/attention supply and musculoskeletal complaints of the neck, shoulder, and upper and lower back. Multivariate binary logistic regression models were estimated in Mplus using robust maximum likelihood. The latent variable, task engagement, was calculated with confirmatory factor analysis using Mplus. The independent variables were attentional demand and attentional supply. The dependent variables were upper body musculoskeletal complaints (neck, shoulder, upper back and lower back) and the mediating variable was task engagement. The effects from attention demand and attention supply to task engagement were linear regression estimates and the effects from task engagement to musculoskeletal complaints were logistic estimates reported as odds ratios. The adequacy of each model was assessed using the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC) and the sample-size adjusted Bayesian Information Criterion (ssaBIC), with lower values indicating better model fit. These fit statistics balance model fit with parsimony to determine the optimum model [30].

Three variations of each model were specified and estimated. The first model (Fig. 1a) tested the direct effect of the independent variables on the dependent variables. The second (Fig. 1b) was the mediated model, which tested all possible relationships between the independent and dependent variable via the mediator (task engagement). The third model (Fig. 1c) included both the direct effects of the independent variables on the dependent variables along with the mediatedeffects.

Fig. 1

Direct effects (a), mediated effects (b) and combined direct & mediated effects (c) models.

3 Results

3.1 Characteristics of the observed variables

Of the 235 respondents, eighty percent (188) were male and 20% were female (47), 20% (47) were in the 21–30 age group, 41.3% (97) were in the 31–40 age group, 27.7 % (65) in the 41–50 age group and 11.1 % (26) were in the 51–60 age group.

Figure 2 details the prevalence of neck, shoulder, and upper and lower back complaints across the industrial sectors. The electronics company had the highest complaint rates for shoulder (51.4%), upper back (37.1%) and lower back (68.8%), while the pharmaceutical workers reported the highest rates of neck complaints (46.8%). Back complaints (upper and lower) were lower for the medical devicescompanies.

Fig. 2

Mean (SD) prevalence of self-reported musculoskeletal complaints in the previous 12 months. The % prevalence represents a proportion of the respondents who reported WRMSDs in each body area. The error bars denote the standard error, which is a measure of the variability of the mean in each group.. WRMSD: Work related musculoskeletal disorder.

One-way analysis of variance results between the key study variables (Table 1) show that there is no significant difference between companies in relation to musculoskeletal complaints. The Kruskall wallis test was used for upper back results as this variable was not normally distributed and there is no significant difference between companies in relation to this variable.

Table 1

One-way analysis of variance for the musculoskeletal variables

Variable name	Mean (SD)	Frequency	One-way ANOVA	P-value
Neck	0.41 (0.49)	96	Welch (3,96.16) = 0.815	P = 0.48
Shoulder	0.46 (0.49)	108	F(3,234) = 0.334	P = 0.80
Upper back	0.28 (0.45)	66	χ2(3) n = 235, = 5.120	P = 0.16
Lower back	0.57 (0.50)	134	Welch (3,98.53) = 1.7	P = 0.16

Figure 3 compares mean (SD) attention demand, attention supply scores across the four industry sectors. Attention supply scores were similar for the semi-conductor ( $\bar{x}$ = 22.21, SD = 3.33), pharmaceutical ( $\bar{x}$ = 22.57, SD = 3.01)) and medical devices companies ( $\bar{x}$ = 20.39,SD = 3.60), but were slightly lower in the case of the electronics company ( $\bar{x}$ = 18.56, SD = 3.97). Attention demand scores were also similar for the semi-conductor ( $\bar{x}$ = 14.91, SD = 3.96), pharmaceutical ( $\bar{x}$ = 14.51, SD = 4.11) and medical devices ( $\bar{x}$ = 15.14, SD = 3.11) with the electronics company again having lower mean values ( $\bar{x}$ = 9.88, SD = 4.03).

Fig. 3

Mean (SD) attention demand and attention supply scores across the industry sectors. The error bars denote the standard error, which is a measure of the variability of the mean in each group. SART: Situational Awareness to Response Test.

Table 2 shows the fit indices for all models. The results show that the mediated models demonstrate the best fit statistics in both the attention demand and attention supply models. The AIC, BIC and ssaBIC scores are lowest and most desirable for the mediated models in both cases.

Table 2

Comparison of fit indices for the direct model, mediated model, and direct-mediated (combined) models

	Model	Log likelihood	No. of free parameters	AIC	BIC	ssaBIC
Attention Demand	1. Direct	–2990	22	6023	6099	6030
	2. Mediated	–2933	30	5926	6031	5935
	3. Direct and mediated	–2933	34	5933	6051	5943
Attention Supply	1. Direct	–2920	22	5884	5960	5890
	2. Mediated	–2697	30	5454	5558	5463
	3. Direct and mediated	–2698	34	5460	5577	5469

AIC = Akaike information criterion, BIC = Bayesian information criterion, ssaBIC = sample-size-adjusted BIC with a lower value for each indicating a better model fit.

The Odds Ratios (ORs) from the multivariate binary regression analyses are presented in Table 3 for the attention demand model and Table 4 for the attention supply model. An odds ratio of more than 1 means that there is a higher odds of task engagement having an impact on the measured musculoskeletal complaints. For the attention demand model, introducing task engagement as a mediator increased the ORs for each of the four musculoskeletal complaints. For the attention demand model, significant indirect effects are shown for neck (p < 0.01), upper back (p < 0.01) and lower back (p = 0.01) complaints. For the attention supply model, the introduction of the mediator resulted in decreasing odds ratios for the four musculoskeletal complaints. With the attention supply model significant indirect effects are shown for shoulder (p < 0.01) and lower back (p < 0.01) complaints. Table 3 & 4 also include the linear regression estimates for attention demand (p = 0.18) and attention supply (p < 0.001) to task engagement within the mediated model.

Table 3

Logistic regression odds ratios (OR) and indirect effect (via task engagement) of attention demand in upper body musculoskeletal complaints

	Musculoskeletal complaint	Direct OR (95% CI)	Mediated by task engagement OR (95% CI)	Direct and Mediated by task engagement OR (95% CI)	Indirect effect B ¹(S.E.) based on Mediated model, P-value
Attention Demand Model
Direct effects	Neck	0.86 (0.72–1.02)		0.93 (0.75–1.14)
	Shoulder	1.00 (0.80–1.26)		0.90 (0.81–1.33)
	Upper back	0.83 (0.70 –0.96)		0.86 (0.72–1.12)
	Lower back	0.81 (0.66–0.99)^*		1.04 (0.69–1.07)
Mediated effects	Neck		5.16 (1.68–15.88)*	3.52 (1.15–10.79)^*	0.43(0.11) p < 0.001
	Shoulder		1.70 (0.68–4.26)	1.50 (0.63–3.28)	0.29(0.32) p = 0.29
	Upper back		7.27 (1.55–34.11)^*	4.54 (1.10–18.66)^*	0.50(0.47) p < 0.001
	Lower back		3.90 (1.51–13.19)^*	2.62 (0.97–7.03)	0.36 (0.35) p = 0.01
Task engagement	Attention demand				–0.23(0.09) p = 0.01

MSC: Musculoskeletal complaint, OR, odds ratio, C.I., Confidence Interval. ¹Indirect effects are based on standardized model results. Direct models depicted the direct effects of the independent variable on the MSC. Mediated models show the indirect effect of the independent variable on the MSC. Direct and mediated models combine all effects on MSCs.

Table 4

Logistic regression odds ratios (OR) and indirect effect (via task engagement) of attention supply in upper body musculoskeletal complaints

	Musculoskeletal complaint	Direct OR (95% CI)	Mediated by task engagement OR (95% CI)	Direct and Mediated by task engagement OR (95% CI)	Indirect effect B ¹(S.E.) based on Mediated model, P-value
Attention Supply Model
Direct effects	Neck	1.01 (0.95 –1.07)		1.03 (0.97–1.10)
	Shoulder	0.95 (0.89 –1.01)		0.97 (0.91–1.04)
	Upper back	1.01 (0.95 –1.08)		1.03 (0.97 –1.10)
	Lower back	0.94 (0.88 –1.00)		0.97 (0.91 –1.04)
Mediated effects	Neck		0.60 (0.35–1.02)	0.58(0.31–0.99)^*	–0.14 (0.09) p = 0.10
	Shoulder		0.47 (0.26–0.86)^*	0.50 (0.27–0.94)^*	–0.20 (0.09) p = 0.02
	Upper back		0.70 (0.40–1.20)	0.43 (0.23 –0.80)^*	–0.10 (0.09) p = 0.26
	Lower back		0.40 (0.22–0.70)^*	0.64 (0.36–1.14)	–0.25 (0.08) (p < 0.01)
Task engagement	Attention Supply				0.29(0.11) p = 0.01

4 Discussion

Despite the fact that operators participating in this study carried out little or no physical work, the prevalence of upper body musculoskeletal complaints appeared high compared to previous studies [31]. Over 40% of those surveyed (n = 96) reported back complaints, 46% (n = 108) shoulder complaints, 28.1% (n = 66) upper back and 57.6% (n = 134) lower back complaints. These findings are comparable with a recent Eurofound report [32] which reported that 49.2% of plant and process operators in industry had reported musculoskeletal disorders. However, this rate is considerably higher than a recent multi-sector review which reported average prevalence rates of 14.9% [31]. Overall, levels of attention demand and supply of attention tended to be quite high with the maximum attention supply score being 28 and the maximum attention demand score being 21. Task engagement scores were analysed by factorial analysis making these scores difficult to compare with other studies.

The study model (model 2, Fig. 1b) indicated significant links between attention supply attention demand and self-reported neck, upper and lower back complaints via the mediator of task engagement. This supports the study hypothesis. Overall, the results indicate that attention demand had a greater effect on reporting MSCs than attention supply. Odds ratios for the direct effects attention demand model show that attention demand had a significant direct effect on upper and lower back complaints. Further, introducing the mediator into the relationship increased the odd ratios, suggesting that the effect of attention demand on MSCs was more pronounced in the presence of low task engagement. For the attention demand model, significant indirect effects (via task engagement) were shown for neck (p < 0.01), upper back (p < 0.01) and lower back (p < 0.01) complaints. The measure of attention demand used in this study focuses on the instability, variability and complexity of the work, which measures uncertainty within the environment. Uncertainty at work is a psychosocial stressor, particularly when it occurs for prolonged periods [33], and the default physiological response to uncertainty is the sympathoexcitatory preparation for action or ‘fight or flight’ response. Workplaces included in the study tended to have high levels of instability and variability associated with the manufacturing process, which helps to explain why attention demand predicted MSCs when operators were engaged with thetask.

For the attention supply model, inclusion of the mediator decreased the odd ratios for the MSCs significantly in the case of shoulder and lower back complaints. Significant indirect effects are also shown for shoulder (p < 0.01) and lower back (p < 0.01) complaints within this model.

Attention supply measures arousal (energy), concentration and spare mental capacity, which could be viewed as positive performance measures of sustained attention. Attention in the form of arousal and concentration is supplied to meet demands until attentional resources are depleted. The results suggest that when the supply of attention to a task is high, the inclusion of even lowered task engagement can reduce the level of stress experienced, making reporting MSCs less likely. These results are somewhat in conflict with previous research. As high levels of attention supply require increased mental demand, an increase in physiological stress effects, and consequently, MSCs would have been expected with low task engagement [34]. The modelling of an ambiguous construct such as task engagement within psychosocial climates is therefore useful in order to elucidate its role in the reporting of MSCs.

While Helton et al. found that the effect of increasing workload appears to improve task engagement [35], it has also been shown that task engagement correlates with resource availability [36]. This would suggest that, although cognitive workload levels are high in these work environments, cognitive resources are depleted during monitoring tasks in accordance with the attention resource theory, and task engagement is subsequently lowered. A recent meta-analysis found several motivational constructs to be positively associated with task engagement, including autonomy, task variety, feedback, problem solving, and task significance [37]. These variables, particularly autonomy, task variety and feedback from the automated system, have been found to be deficient in highly automated environments [38], which further corroborates the low task engagement scores within the survey.

These findings should be considered in light of some limitations. As cross-sectional data were used, the temporal ordering of task engagement and WRMSDs cannot be unequivocally determined. It is acknowledged that the use of self-report questionnaires as opposed to objective data may introduce self-reporting bias [39] where one might expect someone to report negatively in one dimension and negatively in another. Attempts were made during the study design to reduce common method variance including the use of well validated questionnaires and varying the response formats for each questionnaire which has been found useful in reducing method bias [40]. Additionally, it is possible that the psychosocial factors reported were influenced by confounders (e.g. physical discomfort and fatigue), which were not included within the model. Non-work activities could also contribute to musculoskeletal symptoms, but these were beyond the scope of this study. While these measures of attention have been used in previous studies, they have more often been used in relation to estimating levels of situational awareness. Although the correlation between task engagement and attention supply is weak (r = 0.24, p < 0.01), the introduction of some collinearity into the model may have made it more difficult to detect an effect in the case of attention supply. High collinearity can inflate the variance of at least one estimated regression coefficient.

The study had a specific target of plant and process operators in highly automated manufacturing environments, so the model results may not be directly relevant to other occupational groups. This group is unique in that it encompasses lower skilled workers with high cognitive demands and low physical demands. It would be useful to apply this model to other occupational groups with high cognitive demands, such as office-based workers.

The model does not account for the varying types of levels of automation across the participating companies. However, the study focused on the fact that, regardless of the type and level of automation, the cross-company outcome was an increase in the extent of supervisory monitoring and consequently cognitive demands for the human operator.

5 Conclusion

The study highlights attention demand as a potential psychosocial stressor in highly automated environments. It suggests that task engagement has a role in the relationship between attention and MSCs. Our results suggest that lowered engagement coupled with high attentional demand work is linked to an increased incidence of musculoskeletal complaints. Existing research has linked attention and task engagement, but to our knowledge, these measures have not been studied as precursors of WRMSDs previously.

Where possible, sustained attention work should be minimised during job design and it should be included as a hazard in risk assessments. Improving task engagement may also help to counteract the physiological stress effects of sustained attention.

6 Funding

Research was conducted within the Department of Design and Manufacturing Technology in the University of Limerick, Ireland. This work was supported by the 7th Framework Programme for Research and Technological Development of the European Union (FP7), Robo-Mate project under grant agreement n° 608979.

Conflict of interest

There are no conflicts of interest to be declared.

Footnotes

Appendix

References

Considine

, Considine

. Standard handbook of industrial automation: Springer Science & Business Media; 2012.

Hancock

, Jagacinski

, Parasuraman

, Wickens

, Wilson

, Kaber

. Human-automation interaction research past, present, and future. Ergon Design: Quart Hum Factor Appl. 2013;21(2):9–14.

Warm

, Warm

. Anintroduction to vigilance In: Sustained attention in human performance. John Wiley & Sons Inc, 1984:1–14

Thomson

, Besner

, Smilek

. A resource-control account of sustained attention evidence from mind-wandering and vigilance paradigms. Perspect Psych Sci. 2015;10(1):82–96.

Wickens

. Information processing, decision-making, and cognition In: Salvendy G, editor. Handbook of human factors. John Wiley & Sons; 1987;72-107.

Warm

, Parasuraman

, Matthews

. Vigilance Requires Hard Mental Work and Is Stressful. Hum Factors: J Hum Factors and Ergon Soc. 2008;50(3):433–41.

Podniece

, Heuvel

, Blatter

. Work-related musculoskeletal disorders: Prevention report. Luxembourg: Office for Official Publications of the European Communities, EU-OSHA; 2008.

Smith

, Carayon

. Work organization, stress, and cumulative trauma disorders. In: Beyond Biomechanics: Psychosocial Aspects of Musculoskeletal Disorders in Office Work, Moon SD and SL Sauter SL London: Taylor & Francis; 1996, p 23-42.

Melin

, Lundberg

. A biopsychosocial approach to work-stress and musculosketal disorders. J Psychophysiol. 1997;11(3):238–24.

10.

Van den Heuvel

, Van der Beek

, Blatter

, Hoogendoorn

, Bongers

. Psychosocial work characteristics in relation to neck and upper limb symptoms. Pain. 2005;114(1–2):47–53.

11.

Bongers

, Ijmker

, Van den Heuvel

, Blatter

. Epidemiology of work related neck and upper limb problems: Psychosocial and personal risk factors (part I) and effective interventions from a bio behavioural perspective (part II). J Occup Rehabili. 2006;16(3):272–95.

12.

Romans

, Preclin

. European Union Labour Force Survey–Annual results 2007. Eurostat Data en Focus. 2008;(27)

13.

Golubovich

, Chang

C-H

, Eatough

. Safety climate, hardiness, and musculoskeletal complaints: A mediated moderation model. Appl Ergon. 2014;45(3):757–66.

14.

Matthews

, Campbell

, Falconer

, Joyner

, Huggins

, Gilliland

, et al. Fundamental dimensions of subjective state in performance settings: Task engagement, distress, and worry. Emotion. 2002;2(4):315.

15.

Langheim

, Matthews

, Warm

, Reinerman

, Shaw

, Finomore

, et al. editors. The long pursuit: In search of predictors of individual differences in vigilance. Thirteenth Meeting of the International Society for the Study of Individual Differences, Giessen, Germany; 2007.

16.

Matthews

, Szalma

, Panganiban

, Neubauer

, Warm

. Profiling task stress with the Dundee Stress State Questionnaire In: Psychology of stress: New research. 2013;49-90.

17.

Cheyne

, Solman

, Carriere

, Smilek

. Anatomy of an error: A bidirectional state model of task engagement/disengagement and attention-related errors. Cognition. 2009;111(1):98–113.

18.

Helton

, Russell

. Feature absence–presence and two theories of lapses of sustained attention. Psychological research. 2011;75(5):384–92.

19.

Helton

, Warm

, Matthews

, Corcoran

, Dember

, editors. Further tests of an abbreviated vigilance task: Effects of signal salience and jet aircraft noise on performance and stress. Proceedings of the Human Factors and Ergonomics Society Annual Meeting; 2002: SAGE Publications.

20.

Fink

. Howto conduct surveys: A step-by-step guide: Sage Publications; 2015.

21.

Kuorinka

, Jonsson

, Kilbom

, Vinterberg

, Biering-Sørensen

, Andersson

, et al. Standardised Nordic questionnaires for the analysis of musculoskeletal symptoms. Appl Ergon. 1987;18(3):233–7.

22.

Mesquita

, Ribeiro

, Moreira

. Portuguese version of the standardized Nordic musculoskeletal questionnaire: Cross cultural and reliability. J Pub Health. 2010;18(5):461–6. 2

23.

Helton

, editorValidation of a short stress state questionnaire. Proceedings of the Human Factors and Ergonomics Society Annual Meeting; 2004: SAGE Publications.

24.

Matthews

, Joyner

, Gilliland

, Campbell

, Falconer

, Huggins

. Validation of a comprehensive stress state questionnaire: Towards a state big three In: Personality and Psychology in Europe. 1999;7:335–50

25.

Matthews

, Coyle

, Craig

. Multiple factors of cognitive failure and their relationships with stress vulnerability. J Psychopathol Behav Assess. 1990;12(1):49–65.

26.

Taylor

. Situational Awareness Rating Technique(SART): The development of a tool for aircrew systems design. AGARD, Situational Awareness in Aerospace Operations 17 p(SEE N 90-28972 23-53). 1990.

27.

Sarter

, Givens

, Bruno

. The cognitive neuroscience of sustained attention: Where top-down meets bottom-up. Brain Res Rev. 2001;35(2):146–60.

28.

Salmon

, Stanton

, Walker

, Jenkins

, Ladva

, Rafferty

, et al. Measuring Situation Awareness in complex systems: Comparison of measures study. Int J Ind Ergon. 2009;39(3):490–500.

29.

Hair

, Anderson

, Babin

, Black

. Multivariate data analysis: A global perspective: Pearson Upper Saddle River, NJ; 2010.

30.

Vrieze

. Model selection and psychological theory: A discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Psychol Meth. 2012;17(2):228.

31.

da Costa

, Baptista

, Vaz

. Incidence and prevalence of upper-limb work related musculoskeletal disorders: A systematic review. Work. 2015;51(4):635–44.

32.

Eurofound. Fourth European Survey on Working Conditions (ESWC 2005). Luxemburg; 2007.

33.

Warr

. Job features and excessive stress,. In: Prevention of Mental Ill Health at Work London: HMSO; 1992.

34.

Matthews

, Warm

, Reinerman

, Langheim

, Saxby

. Task engagement, attention, and executive control In: Handbook of individual differences in cognition: Springer; 2010, p. 205-30.

35.

Helton

, Shaw

, Warm

, Matthews

, Hancock

. Effects of warned and unwarned demand transitions on vigilance performance and stress. Anxiety, Stress, & Coping. 2008;21(2):173–84.

36.

Reinerman

, Matthews

, Warm

, Langheim

, Parsons

, Proctor

, et al. editors. Cerebral blood flow velocity and task engagement as predictors of vigilance performance. Proc Hum Factors Ergon Soc Annu Meet; 2006: Sage Publications.

37.

Christian

, Garza

, Slaughter

. Work engagement. Pers Psychol. 2011;64(1):89–136.

38.

Truxillo

, Cadiz

, Rineer

, Zaniboni

, Fraccaroli

. A lifespan perspective on job design: Fitting the job and the worker to promote job satisfaction, engagement, and performance. Organiz Psych Rev. 2012;2(4):340–60.

39.

Adams

, Soumerai

, Lomas

, Ross-Degnan

. Evidence of self-report bias in assessing adherence to guidelines. Int J Qual Health Care. 1999;11(3):187–92.

40.

Kothandapani

. Validation of feeling, belief, and intention to act as three components of attitude and their contribution to prediction of contraceptive behavior. J Pers Soc Psychol. 1971;19(3):321–333.