Ergonomics Vol. 53, No. 6, June 2010, 739–747
Job strain related to cognitive failure in naval personnel* Robert S. Bridger{, Kate Brasher, Angela Dew, Kathy Sparshott and Shaun Kilminster Institute of Naval Medicine, Crescent Rd, Alverstoke PO12 2DL, UK (Received 10 February 2009; final version received 10 December 2009) The Naval Service Stress Study (2007–2012) is investigating job strain, its characteristics, causes and distribution in the Service. Data from phases I, II and III of the study (January 2007, June 2007 and January 2008) were analysed to determine the relationship between General Health questionnaire scores and a score on the Cognitive Failures Questionnaire (CFQ) completed at phase III. Of 791 personnel who completed questionnaires at all phases, 43.6% had no job strain at any phase, whereas 9.9% had strain on all three occasions (‘chronic strain’). 27% had strain at one of the three phases and 19% had strain at two of the three phases. The particular phase at which job strain was experienced was not related to CFQ score at phase III, whereas the total strain experienced over the period was related. High strain over the year was the strongest predictor of high CFQ score. A ‘strain dose’ variable, which combined both the amount of strain exposure and the timing of the exposure, explained little additional variance in CFQ score. The findings might be interpreted to indicate that a high CFQ score is a vulnerability factor for adverse reactions to work stress. The hypothesis that recent job strain elevates CFQ score was not supported. Statement of Relevance: Current models of occupational stress focus on psychosocial factors and much of the advice about stress management in organisations is centred on the identification and control of psychosocial risk factors. The present paper provides evidence that cognitive factors are also important and suggests that support for those with poor executive function should be part of stress management in complex environments. Keywords: cognitive failure; Navy; occupational stress; psychological strain
1. Introduction The Maritime Coastguard Agency in the UK states that one of the effects of job stress is increased error due to lapses of attention, memory problems and slower reaction time (Maritime Coastguard Agency 2007). This article presents data from the Naval Service Cohort Study of occupational stress in order to explore this proposed relationship between problems of memory, attention and action in naval personnel (hereafter ‘cognitive failure in naval personnel’) and adverse psychological reaction to job stress – hereafter ‘job strain’ Job strain has been investigated in a wide variety of occupational groups, often appearing in the form of anxiety and depression, problems sleeping and so on (Stansfield et al. 1999). Chronic strain may lead to physical ailments such as coronary heart disease (Bosma et al. 1998). Karasek and Theorell (1990) proposed that job strain is more likely when job demands are high, when employees have little control over the demands and receive little social support (Karasek 1979). Questionnaires, such as the 12-item General Health Questionnaire (GHQ-12; Goldberg and Williams 1988), have been used to measure the prevalence of job strain in a wide variety of {
occupational groups, including civil servants (e.g. Kuper and Marmot 2003), the UK Police (Collins and Gibbs 2003) and the Royal Navy (Bridger et al. 2007). The prevalence rate of job strain in the Royal Navy is approximately 30% at any time and similar to the rate in the UK Police (Slaven 2002, Bridger et al. 2007, 2008a). Occupational stressors such as low control, conflicting work demands, conflict between work and family life and lack of support from superiors have been shown to be associated with acute job strain in both the Navy and the Police. Job demands, seen as ‘occupational stressors’ in these models, are often measured using self-report questionnaires such as the ‘Work and Well-Being Questionnaire’ (WWBQ) used in the current Royal Navy cohort study. The development of this questionnaire, including its validity and reliability, is described in Slaven (2002) and Kilminster and Bridger (2007). 1.1.
Cognitive failures
One of the most commonly used tools to investigate cognitive failure in a wide variety of occupational and
Corresponding author. Email:
[email protected] *Published with the permission of the Controller of Her Majesty’s Stationery Office. ISSN 0014-0139 print/ISSN 1366-5847 online Ó British Crown Copyright 2010 MOD DOI: 10.1080/00140131003672031 http://www.informaworld.com
740
R.S. Bridger et al.
patient groups is the Cognitive Failures Questionnaire (CFQ; Broadbent et al. 1982, Kass et al. 2002). The CFQ measures self-reported failures of perception, memory and action and has been shown to be psychometrically robust with good test–retest reliability. The high test–retest reliability of the CFQ led Broadbent et al. (1982) to conclude that the CFQ score was more like a trait variable (a stable characteristic of a person) than a temporary state (Weeks et al. 1980 found a test–retest correlation of 0.82 over 21 weeks and 0.80 over 65 weeks for the CFQ). Furthermore, Broadbent presented evidence that a high CFQ score was not the result of high anxiety but more likely a ‘vulnerability factor’ for adverse reactions to stressful work. In other words, people with a high propensity to cognitive failure would be more likely to suffer job strain when job demands were high, but not when work was undemanding. 1.2. CFQ and job strain The relationship between job strain and cognitive failure has been demonstrated in several studies, but the nature of the relationship over time is not clear. Although Broadbent believed that proneness to cognitive failure makes job strain more likely when job demands are high (Broadbent et al. 1982), others have suggested that chronic strain may have adverse effects on the mechanisms of attention. For example, van der Linden et al. (2005) compared CFQ scores and tests of cognitive performance in three groups: teachers with no symptoms of burnout; working teachers reporting burnout symptoms; hospital patients with burnout symptoms who could no longer work. CFQ scores were higher in the burnout groups than in the controls and highest in patients. CFQ scores correlated positively with the number of response inhibition errors on cognitive tasks but not with reaction time, which was indicative of diminished executive control in those with burnout. Because this study of job outcomes and CFQ, in common with other studies such as that of Wadsworth et al. (2008) was cross-sectional, it could not be determined whether a high CFQ score was a cause or an effect of the outcome of interest. For example, if the teachers in van der Linden et al.’ s study recovered from burnout, would their CFQ scores then decline? It is known that exposure to physical stressors, such as surgery, can cause an elevated CFQ score (after general anaesthesia, for example, Tzabar et al. 1997) but it is not known whether GHQ (job strain) and CFQ scores are simply covariates, in the sense of being measures of the same underlying process or deficit, or whether one causes or moderates the incidence of the other.
1.3.
Naval Service cohort study
The present paper is part of a larger study – the Naval Service cohort study of occupational stress (Bridger et al. 2008a) the purpose of which is to: identify those characteristics of demanding work (‘occupational stressors’) that cause strain in naval personnel; investigate the relationships between work demands, health and the retention of personnel; identify groups of personnel at high risk of job strain. To date, the study has identified occupational psychosocial and demographic risk factors for job strain, determined the role of stressful life events outside of work and enabled the prevalence of acute and chronic job strain to be estimated. Job strain has been shown to be associated with over-commitment to work role, perceptions of effort–reward imbalance (ERI), conflicting work demands, work family conflict (in male personnel) and dissatisfaction with the physical work environment (in female personnel). Strain prevalence is greater in females than in males and in non-officers compared to officers (Bridger et al. 2007). Due to the nature of naval deployments, with time spent working at sea and ashore in a variety of work roles, work demands change frequently. Recent findings have shown that strain levels in naval personnel vary greatly over a 6-month period, particularly when work roles change (Bridger et al. 2008b). In summary, these findings show that job strain in naval personnel is not randomly distributed. It is systematically linked to work exposures that have been measured using valid and reliable scales designed for the purpose (Slaven 2002). 1.4. Aims The present paper explores the relationship between cognitive failure, measured using the CFQ in January– March 2008 and job strain measured at the same time and at 6-month intervals in the preceding year. Current CFQ score could then be related to strain levels experienced in the past, as well as strain measured at the same time as CFQ, and the total strain experienced over the past year. If cognitive failure were a result of job strain, particularly cumulative strain, then it might be possible to identify the kinds of demands that are responsible and design or curtail workload accordingly. Conversely, if cognitive failure were a risk factor for adverse reactions to stressful work, then it might be preferable to propose job aids to support those prone to cognitive failure when work demands are high. The hypothesis that job strain, measured at three phases from January 2007 to January 2008, predicted CFQ score tested at phase III (January 2008), was
741
Ergonomics tested with specific reference to the temporal proximity of strain episodes to CFQ administration. 2. Method Ethical approval was obtained from the Ministry of Defence Research Ethics Committee. A retrospective longitudinal study design was employed. At phase III (January–March 2008), 2596 questionnaires were sent to all personnel who responded to the naval stress survey at phase I (January– March 2007). Full details of phase I of the study may be found in Bridger et al. (2008a). The questionnaires sent at phase III were the GHQ-12 to measure job strain and the CFQ to measure cognitive failures. 2.1.
WWBQ
The WWBQ was used to measure occupational stressors in the naval sample at phase I of this longitudinal survey. The WWBQ was developed following the conduct of focus groups of five to eight personnel of equivalent rank in three naval establishments (Slaven 2002). A pilot questionnaire was then circulated to 600 randomly selected personnel. Factor analysis of the returns was conducted to identify items that loaded discretely on to different factors. The item set was reduced by calculating alpha coefficients for each of the scales and removing items until the value of alpha exceeded 0.7. Additional scales were added in 2006 (Kilminster and Bridger 2007) and psychometric reduction and validation was carried out using a sample of naval personnel. The purpose of the psychometric reduction was to confirm the factor structure of the original WWBQ and to reduce the length of the new WWBQ by extracting three to five items from parent scales measuring ERI, over-commitment, positive and negative mood (from the PANAS scale of Watson et al. 1988) and coping style. The new ERI scale had a Cronbach’s alpha value of 0.72 and the over-commitment scale had a value of 0.83. A 5-d test–retest reliability study was conducted and unreliable items were removed. Most of the items in all scales of the final WWBQ were highly reliable with Spearman correlation coefficients over 0.7. The new WWBQ (Table 1) measured occupational stressor exposure, job strain exposure to stressful life events and main confounding factors and effect modifiers. 2.2. Questionnaire scoring The WWBQ data were scored by summing individual items to obtain totals for the parent scales used to
Table 1. Summary of all variables measured by the Work and Well-Being Questionnaire at phase I. Stressors (Occupational psychosocial factors) Financial reward satisfaction Effort–reward imbalance Over-commitment Role conflict Work–family conflict Physical work environment Autonomy and control Adequacy of resources Physical job demands Satisfaction with ship habitability* Stress buffers/modifiers Leader support Leader approachability Peer support Commitment to the service Coping style
Demographics Job specialisation Location and preference Marital status Health lifestyle factors BMI Smoking Alcohol intake (units) Exercise participation Health complaints Sea-sickness Stress outcomes GHQ-12 Score (‘strain’) General health rating Absenteeism Expectations of future service (years)
Confounding factors Mood Stressful life events GHQ-12 ¼ 12-item General Health Questionnaire. *For personnel at sea only.
measure the variables listed in Table 1. Stressor exposures were self-reported on 5-point Likert scales, where high scores indicated greater negativity. Stressful life event scores were simple counts of the number of prompted events to which personnel had been exposed in the previous 12 months. The GHQ-12 measure of strain was scored in three different ways: (1) Strain prevalence rates were generated using a binary classification of strain ‘caseness’. A strain case is a respondent who reports four or more strain symptoms on the GHQ-12. The strain prevalence rate is given by the percentage of strain cases in the total sample. Goldberg and Williams (1988) advise that the choice of the number of symptoms used to describe a strain case depends on the particular population. In the present study, a cut-off score of 4 was used because it has been shown to have reasonable test–retest reliability (Kilminster and Bridger 2007) and to enable comparison with the strain prevalence rates reported in the literature for military and other populations in which a cut-off score of 4 is used (e.g. Hotopf 2006). (2) The number of strain symptoms that an individual reported was counted. Note, this is an integer score that ranges from ‘0’ (no
742
R.S. Bridger et al.
symptoms) to ‘12’ (strain symptoms on all 12 items on the GHQ-12). (3) A ‘Likert’ strain score was obtained. Given that the 12 items on the GHQ-12 can each be assigned a Likert score from zero to 3, the GHQ-Likert score ranges from ‘0’ to ‘36’ (higher numbers indicating increased strain). CFQ score was calculated by summing the 5-point scales (ranging 0–4) for each of the 25 items on the CFQ. The CFQ score can range from 0–100, with higher numbers indicating increased cognitive failure. 2.3.
Strain dose variable
A strain dose variable was constructed to enable strain at the three phases to be analysed in a simple manner. The concept of strain dose relates to primacy and recency effects and the notion of decay and accumulation of stress. The GHQ-12 case scoring method (in which respondents with a score of 4 or above are labelled strain ‘cases’) was used to divide the sample into eight subgroups (Table 2) varying in frequency of strain episodes over the year and strain recency (closeness in time to phase III). The grouping method is described in Table 2. The resulting ‘strain dose’ variable was considered as a quasi-interval scale of 8 points, with higher strain dose scores indicating greater exposure to strain and greater temporal proximity of strain and CFQ measurement. Mean CFQ scores were calculated for the different groups. 2.4.
Data analysis
Correlation matrices (Pearson’s r) for CFQ, GHQ-12 (Likert score), psychosocial scale scores (phase I) and demographic data were generated. A pragmatic approach to the stepwise multiple regression analysis was conducted to identify predictors of CFQ score, with predictors entered in order of strength of bivariate
Table 2.
Summary of strain categories.
Strain category 1. 2. 3. 4. 5. 6. 7. 8.
No strain Strain once Strain once Strain once Strain twice Strain twice Strain twice Chronic strain
Phase I
Phase II
Phase III
n
0 1 0 0 1 1 0 1
0 0 1 0 1 0 1 1
0 0 0 1 0 1 1 1
345 79 48 90 58 39 54 78 791
0 ¼ No strain; 1 ¼ strain present.
correlation with CFQ. A combined GHQ-12 Likert score (GHQ grand total) was calculated by summing the GHQ-12 Likert scores for each respondent from all three phases. The ‘strain dose’ score was also entered as a predictor combining both strain frequency and strain recency (the temporal proximity of strain to CFQ measurement at phase III). Before doing so, w2 tests were conducted to determine whether there were any associations between the strain dose score and rank and gender. 3. Results Of the 2596 personnel who returned questionnaires at phase I, 2468 were still serving by phase III. Of the questionnaires sent, 1305 were returned, which equated to a 61% response rate (following correction for 315 non-contactables). The mean CFQ score for females (mean 39.7, SD 15.0) was higher than that for males (mean 36.3, SD 13.5) and the difference was statistically significant (t ¼ 4.17, degrees of freedom (df) ¼ 1257, p 5 0.001). The mean CFQ score for officers (mean 35.9, SD 12.8) was lower than that for ratings (mean 38.7, SD 14.9) and the difference was statistically significant (t ¼ 3.3, df ¼ 1257, p 5 0.002). Those with strain at phase III were, for the most part, not the same individuals who had strain at phase I: . 15% had psychological strain on both occasions. . 35% had strain at either phase I or phase III but not on both occasions. . 50% had no strain on either occasion. Further analysis was restricted to respondents to all three phases of the study. A total of 791 personnel (319 females and 472 males, 317 officers and 474 non-officers) completed the questionnaire at all three phases and 44% of these had no strain at any phase, whereas 10% had strain on all three occasions (‘chronic strain’). Only 27% had strain at one of the three phases and 19% had strain at two of the three phases. The prevalence of strain at phase III was 31% for males and 36% for females. Only 31% of officers had strain at phase III compared to 34% of non-officers. The mean CFQ scores were calculated for the different strain dose groups (Figure 1). w2 tests were conducted to determine whether the strain dose category was associated with rank or gender. For gender, w2 ¼ 13.3, df ¼ 7, p 4 0.05. For rank, w2 ¼ 0.47, df ¼ 7, p 4 0.05). In both cases, there were no statistically significant associations.
743
Ergonomics 3.1. Correlations between GHQ-12 at phases I, II and III and CFQ Table 3 summarises Pearson product moment (PPM) correlation coefficients between GHQ-12 Likert scores at phases I, II and III. As can be seen, the 6-month test–retest reliability of the GHQ-12 was greater than the 12-month reliability. Statistically significant positive correlations were found between CFQ and GHQ-12 score at all three phases, suggesting that job strain is associated with high CFQ score irrespective of when the strain was experienced. With a sample size of 4790, statistical power is great and thus pairwise PPM correlations are all significant even as low as when r ¼ 0.1, where p ¼ 0.004. The minimum PPM correlation that can be detected as significant is r ¼ 0.075, which is practically
Figure 1. Relationship between strain dose variable and Cognitive Failures Questionnaire (CFQ) score (range 0–100). Table 3. Pearson product moment correlation coefficients, 12-item General Health Questionnaire (GHQ-12) scores at phases I, II and III and Cognitive Failures Questionnaire (CFQ) score. Correlation coefficients greater than 0.3 are in bold for ease of identification. GHQ
Phase III Phase II Phase I
Phase III
Phase II
Phase I
CFQ
1.00 0.41 0.34
0.41 1.00 0.48
0.34 0.48 1.00
0.40 0.33 0.38
no relationship. Correlations above r ¼ 0.3 are thought to have the minimum practical relevance and beyond chance and are highlighted in bold. This also applies to the PPMs in Table 4. 3.2. Correlations between CFQ scores, total strain and job stressors Table 4 presents a correlation matrix of CFQ, job strain and psychosocial variables. The highest correlation was between GHQ-grand total and strain dose, followed by CFQ and GHQ-grand total. 3.3.
Prediction of CFQ score
Multiple regression analysis (forward stepwise) was conducted to identify variables that predicted CFQ score at phase III. Table 5 summarises the results of the analysis. The best predictor of CFQ score was GHQ-grand total, accounting for 23% of the variance. Although statistically significant, the remaining factors had small effects on CFQ score. The strain dose variable added nothing to the predictive power of the model. Table 6 presents the CFQ data in the context of other studies to facilitate interpretation of the findings. Overall, the Naval Service has a mean CFQ score similar to that of other occupational groups, such as skilled workers, and lower than that of groups experiencing psychological difficulty (van der Linden et al. 2005). The chronic strain group in the RN (strain dose score 8 in Table 1) has a mean CFQ score similar to that of other non-clinically depressed groups, as might be expected. 4. Discussion Cross-sectional studies of job strain in naval personnel conducted in 1999, 2004 and 2007 have shown that the prevalence rate of strain is relatively constant over time (approximately 30–32%), higher in females than in males and in non-officers than in officers (Bridger et al. 2007, 2008a). Respondents to phases II and III had similar rates of strain and exhibited similar demographic differences in strain prevalence compared to previous surveys. This suggests that those who responded to all three phases of the cohort study were representative of the phase I cohort and of the naval personnel studied in 1999 and 2004. Importantly, the presence of strain did not bias response to subsequent administrations of the questionnaires. Tests of association between the strain dose categories, rank and gender, indicated that the strain dose categories were not surrogate measures of rank or gender and therefore that the differences in CFQ score between the
744
R.S. Bridger et al.
Table 4. Correlation matrix of Cognitive Failures Questionnaire (CFQ), General Health Questionnaire (GHQ) and psychosocial risk factors. GHQ Strain grand CFQ dose total ERI OC
RC
WFC
WH
AC
Res
L supp
P Coping Pos Neg supp OrgC. style mood mood
CFQ 1.00 Strain dose 0.36 1.00 GHQ-grand 0.46 0.81 1.00 total ERI 0.12 0.22 0.28 1.00 OC 0.27 0.25 0.37 0.29 1.00 RC 0.23 0.22 0.30 0.41 0.32 1.00 WFC 0.19 0.22 0.30 0.24 0.19 0.27 1.00 WH 0.20 0.21 0.23 0.22 0.10 0.30 0.31 1.00 AC 0.13 0.15 0.20 0.15 0.03 0.22 0.17 0.12 1.00 Res 0.14 0.13 0.17 0.25 0.24 0.37 0.22 0.28 0.11 1.00 LSupp 0.10 0.16 0.23 0.25 0.10 0.28 0.16 0.16 0.29 0.14 1.00 PSupp 0.07 0.12 0.18 0.26 0.13 0.15 0.16 0.16 0.16 0.08 0.34 1.00 OrgC. 0.19 0.22 0.32 0.24 0.08 0.25 0.31 0.21 0.24 0.18 0.28 0.31 1.00 Coping style 70.01 70.08 70.09 70.06 0.07 70.06 70.06 70.07 70.29 0.01 70.24 70.29 70.21 1.00 Pos mood 0.24 0.19 0.28 0.06 0.03 0.12 0.12 0.16 0.34 70.03 0.20 0.19 0.36 70.29 1.00 Neg mood 0.31 0.25 0.36 0.05 0.25 0.10 0.08 0.09 0.09 0.09 0.02 0.05 0.04 70.04 0.15
1.00
ERI ¼ effort–reward imbalance; OC ¼ over-commitment; RC ¼ role conflict; WFC ¼ work–family conflict; WH ¼ work environment; AC ¼ autonomy and control; Res ¼ resource adequacy; Lsupp ¼ support form leaders; Psupp ¼ support from peers; OrgC ¼ commitment to organisation; Pos mood ¼ positive mood; Negmood ¼ negative mood. Correlation coefficients greater than 0.3 are in bold for ease of identification.
Table 5. Predictors* of Cognitive Failures Questionnaire score at phase III for respondents to phases I, II and III. B GHQ grand total 0.66 GHQ Phase 2 70.40 Strain Dose 70.31
Multiple R Multiple R2 0.48 0.49 0.49
0.23 0.24 0.24
DR2 0.23 0.01 0.00
*The major predictor is General Health Questionnaire (GHQ) grand total and this table shows the additional advantage of separate GHQ phase 2 data, adds only 1% to the prediction and strictly speaking is collinear. Similarly, strain dose with additional temporal information adds nothing, being less than 1%. The negative B weighting of the latter two is simply a very small rebalancing to the overall score, which is, in variance terms, extremely small and of no practical additional value.
categories were not confounded by these demographic factors. The best predictor of CFQ was total job strain (GHQ-12 score summed over three phases), which suggests that adverse reactions to work demands were associated with increased cognitive failure. However, CFQ was similarly related to GHQ-12 score at each phase (e.g. CFQ accounted for 14.4% of the GHQ variance at phase I and 16% of the GHQ variance at phase III), implying that it made little difference whether the adverse reaction was occurring at the same time the CFQ was completed or whether it had occurred 12 months before. As can be seen from Figure 1, CFQ score increased steadily as the number of episodes of strain increased, but there was no evidence of any effect due to the temporal proximity of strain.
Table 6. Cognitive Failures Questionnaire (CFQ) scores in the Naval Service – comparison with other demographic groups in ascending order of CFQ score.
Teachers in a vocational institute (low burnout)1 University students (not depressed)2 Gulf War veterans (well)3 Car factory production workers4 Skilled men4 RN (all) University students (with SAD)2 RN (chronic strain cases) University students (depressed)2 Student nurses4 Gulf War veterans (ill)3 Diagnosed burnout patients1
Mean
SD
26.8
7.5
31.1 33.3 35.0 36.7 37.6 43.2 45.5 50.2 52.5 56.4 59.5
1.4 14.3 11.5 9.4 14.0 2.3 14.7 6.0 14.5 16.8 8.3
SAD ¼ seasonal affective disorder. Sources: 1van der Linden et al. 2005; 2Sullivan and Payne 2007; 3 David et al. 2002; 4Broadbent et al. 1982.
This is borne out by the findings of the regression analysis – the addition of temporal information in terms of onset, duration, summation and elimination of a ‘strain dose’ did not add anything to the prediction of CFQ above a simple summation of strain. On the contrary, the act of summating the GHQ-12 scores over the three time periods eliminated temporal information about the pattern of exposure to job strain. The elimination of this information had no effect on the predictive power of the model.
745
Ergonomics Figure 1 summarised the mean CFQ scores for personnel grouped according to how many phases they had been categorised as a strain case. Visual inspection shows that the mean CFQ scores are higher for personnel who were strain cases more often. However, there was no obvious linear trend in relation to how recently the strain was experienced. For those who had been strain cases at one of the three phases, mean CFQ score was not higher when strain was concurrent than when strain had occurred 12 months before. The same observations can be made about the mean CFQ scores of the groups categorised as strain cases twice in the 12 month period. If job strain caused cognitive failure, such a time-dependent trend would be expected. The linear trend between CFQ score and strain category appears to be due to strain frequency, rather than strain recency (i.e. temporal proximity to CFQ measurement at phase III). This finding concurs with the view of Broadbent et al. (1982) that personnel with a high CFQ score are more likely to exhibit adverse psychological reactions to work stress than those with a low CFQ score. Alternative explanations are possible. It may be that work stress causing psychological strain increases the tendency to make mistakes and this increased tendency persists subsequently for long periods after the strain has subsided. Alternatively, those with chronic or repeated strain may be taking medication and the high CFQ score is an effect of the medication. Data on medication were not available under the current research protocol. Of the explanations outlined above, the former seems more likely. A previous report (Bridger et al. 2008b) demonstrated that change in job strain over time is linked to change in work role and that a period of 6 months was long enough for personnel to experience significant recovery from strain. There is no obvious explanation as to why symptoms of anxiety and depression should resolve over this time period whilst cognitive deficits as a result of the symptoms should persist. The most parsimonious explanation is that CFQ score mediates the interaction between job stress and the emotional reaction to it. Personnel with a propensity to cognitive failures may be more likely to experience strain when exposed to high occupational demands than personnel with low cognitive failure propensity – under high workload, they have difficulty coping with the increased demands on attention. To the extent that CFQ score reflects executive function, this argument accords with Williams et al. (2009), who state that executive function is vital for solving novel problems, modifying behaviour in the light of new information, generating strategies for complex actions, following through with plans and over-riding impulses to engage in goal-directed behaviour. Executive
control is a core neurocognitive domain (Suchy 2009). Good executive control would appear to be needed to deal with stressful situations at work and a lack of control will manifest itself as an inability to cope – work inefficiency and very likely an adverse emotional response. The precise mechanism by which CFQ might mediate the relationship between job stress and emotional reaction is not yet known. Schmidt et al. (2007) suggested that there might be direct or indirect effects. Tasks that place high demands on executive function may be perceived as more demanding or more threatening by those with high CFQ scores. Alternatively, cognitive failure at work may damage interpersonal relations with co-workers and supervisors. 4.1.
Practical implications
The current findings are in accord with research carried out in other occupational groups. High CFQ scores in otherwise healthy individuals indicate increased vulnerability to exhibiting psychoneurotic symptoms when exposed to a stressful environment. In nonstressful environments, no such difference is observed. Females tend to have higher CFQ scores and greater strain levels than males (Hood et al. 1987) and, indeed, the current study found statistically significant differences in CFQ score between males and females, and officers and non-officers (‘ratings’). However, these differences were not large. Individuals with good executive function demonstrate the ability to monitor their actions in relation to their intentions at all times (to pay attention to what they are doing). Those with poor executive function tend to lapse into a mode of behaviour characterised by increased ‘automaticity’ (carrying out simple tasks without thinking, allowing the mind to wander and so on), are less able to resist sudden impulses (to take short cuts, to deviate from a planned course of action in return for temporary gain and so on) There is empirical evidence, for example, that computer users with high CFQ scores are more likely to lose computer work because of a failure to save it (Jones and Martin 2003). The error occurs when the user’s attention is captured by inappropriate stimuli, suggesting that people with high CFQ scores are less able to resist distraction. Wallace and Chen (2005), for example, found that several indicators of organisational safety were associated with CFQ scores – including supervisor ratings of individuals’ safety behaviour. They also suggested that it would be worthwhile to examine the influence of anxiety and stress on cognitive failure. Whilst not providing a definitive explication of the direction causality, the
746
R.S. Bridger et al.
present findings do suggest that, in healthy employees, CFQ score is likely to mediate the adverse reaction to stress and that high stress does not increase the risk of cognitive failure within the range of responses studied (whilst accepting that more extreme physical stressors applied to persons receiving medical treatment may do so, e.g. Tzabar et al. 1997). The trend in ship design is for more complex technology, increased automation and fewer personnel aboard. Teams will be smaller, reducing the scope for peer support while many of the existing work stressors (such as work–family conflict and role conflict) will remain. That there is a link between job strain and cognitive failure is of practical importance and of theoretical interest. If job strain caused cognitive failure in otherwise healthy people, then the imposition of high work demands on personnel might compromise safety at sea. If, however, high CFQ score mediates the stress–strain reaction, then CFQ might be used to identify people likely to react adversely to stressful deployments. Where it is not possible to ‘fit the job to the person’ in the sense of traditional ergonomics, then the identification of vulnerable individuals may be a preferred option. Indeed, Finomore et al. (2009) cite evidence that high CFQ scores are indicative of increased propensity for boredom, mind-wandering and the vigilance decrement and suggest that tests of executive function might have use in the selection of operators for vigilance tasks. 4.2.
Caveats
The suggestion based on the present data that high CFQ scores indicate increased susceptibility to adverse reactions to stressful work exposures is open to criticism. It might be argued that both high GHQ-12 scores and high CFQ scores are common indicators of nothing more than a general tendency of some individuals towards negative self-evaluation (David et al. 2002). However, the multiple regression analysis indicated that the total GHQ-12 score predicted CFQ at phase III independently of negative mood. There is also the possibility that GHQ-12 and CFQ scores are simply correlates of a third, unmeasured variable, such as conscientiousness, and that conscientious people cope better when work is stressful and tend not to make errors in daily life. A recent study by Wallace and Vodanovich (2003) found that cognitive failure does account for workplace safety behaviour and accidents, independently of conscientiousness, although the effect was most marked when conscientiousness was low. A second weakness of the study is the low participation rate in all three phases and the possibility of response bias. Analysis of response bias at phases I and II has already been conducted (Bridger et al.
2008b), indicating only 5% bias in repeated measures (due to age and participation in physically demanding work). The prevalence rate of strain at all three phases was the same (approximately 30%) and almost identical to that found in previous studies of job strain in the Naval Service. It is unlikely that there is any response bias due to job demands or job strain. Furthermore, previous studies of naval personnel all indicate a large work component to job strain: the main occupational stressors linked to strain being role conflict, work conflict and dissatisfaction with the physical conditions (Bridger et al. 2007). In addition, change in job strain in naval personnel is linked to change in work role (Bridger et al. 2008b) in the expected direction (increased work demands increase job strain levels and vice versa). So, despite these reservations, a case can be made that it was the link between high CFQ score and occupational stressors that resulted in the strain. Finally, reservations must be expressed about the extension of these findings to other groups. The Naval Service has a mean CFQ score similar to that of occupational groups, such as skilled workers, and lower than that of groups experiencing psychological difficulty. The chronic strain group in the RN (strain index score 8 in Table 2) had a mean CFQ score similar to that of other non-clinically depressed groups, as might be expected. It is quite possible that exposure to more extreme kinds of stress can cause temporary cognitive failure but this lies beyond the present scope. In a working population, subjected to varying levels of job stress, high CFQ appears to be a vulnerability factor for job strain. Further studies will be carried out to determine whether CFQ score predicts job strain prospectively. In phase V of the study (2012), GHQ-12, occupational stressor exposure and CFQ will be measured in those of the cohort still serving in the Naval Service. This will provide a better understanding of the direction of the causal association between CFQ score and job strain. At present, the findings show a strong and enduring association between CFQ score and job strain. CFQ score certainly seems to mediate the interaction between job stress and the emotional response to stress. Previous surveys (Bridger et al. 2007) have shown that two of the main stressors linked to job strain in naval personnel are the presence of conflicting work demands and work–family conflict, representing precisely the kinds of challenges that personnel with poor executive function would be challenged by. 5. Conclusions Job strain is related to CFQ score over time periods of up to 12 months. Those reporting a greater
Ergonomics propensity to commit errors in daily life were more likely to experience strain. CFQ scores for naval personnel were in the normal range. For chronic strain sufferers, they were in the range for depressed (but not clinically depressed) personnel. Theoretically, the findings are of interest because they suggest that scores on the CFQ and GHQ are not simply measures of the same underlying process or deficit. It is more likely that CFQ score measures an enduring trait in naval personnel, a trait that mediates the psychological stress–strain response measured by the GHQ-12. From a practical viewpoint, the findings contribute to understanding the nature of the health and safety risk in stressful work.
References Bosma, H., et al., 1998. Two alternative job stress models and the risk of coronary heart disease. American Journal of Public Health, 88, 68–75. Bridger, R.S., Kilminster, S., and Slaven, G., 2007. Occupational stress and strain in the Naval Service: 1999 and 2004. Occupational Medicine, 57, 92–97. Bridger, R.S., et al., 2008a. Occupational stress and strain in the Royal Navy, 2007. Occupational Medicine, 58, 534– 539. Bridger, R.S., et al., 2008b. Naval Service Cohort Study Phase II: Chronic strain, acute strain and recovery. Unpublished MoD Report. Broadbent, D.E., et al., 1982. The cognitive failures questionnaire and its correlates. British Journal of Clinical Psychology, 21, 1–16. Collins, P.A. and Gibbs, A.C.C., 2003. Stress in police officers: a study of the origins, prevalence and severity of stress-related symptoms within a county police force. Occupational Medicine, 4, 256–264. David, A.S., et al., 2002. Cognitive functioning and disturbances of mood in UK veterans of the Persian Gulf War: a comparative study. Psychological Medicine, 32, 1357–1370. Finomore, V., et al., 2009. Predicting vigilance: a fresh look at an old problem. Ergonomics, 52, 791–208. Goldberg, D. and Williams, P., 1988. A user’s guide to the General Health Questionnaire. Basingstoke, Hampshire, UK: NFER Nelson, The Basingstoke Press, Ltd. Hood, B.M., Maclachlan, M., and Fisher, 1987. The relationship between cognitive failures, psychoneurotic symptoms and sex. Acta Psychiatrica Scandinavica, 76, 33035. Hotopf, M., 2006. The health of UK military personnel who deployed to the 2003 Iraq way: a cohort study. The Lancet, 367, 1731–1741. Jones, G.V. and Martin, M., 2003. Distribution of attention and failure to save computer work. In: M. Rauterberg et al., Human computer interaction INTERACT 03. Tokyo, Japan: IOS Press @ IFIP 2003. Karasek, R., 1979. Job demands, job decision lassitude and mental strain: implications for job redesign. 1979. Administrative Science Quarterly, 24, 285–307.
747
Karasek, R.A. and Theorell, T., 1990. Healthy work: stress, productivity and the reconstruction of working life. New York: Basic Books. Kass, S.J., Stanney, C.J., and Wallace, J.C., 2002. Cognitive failures questionnaire revisited: dimensions and correlates. Journal of General Psychology, 129, 238–256. Kilminster, S. and Bridger, R.S., 2007. Modified work and well-being questionnaire for a cohort study of occupational stress in the naval service: Psychometric validation. Unpublished MoD Report. Kuper, H. and Marmot, M., 2003. Job strain, job demands, decision latitude and risk of coronary heart disease within the Whitehall II study. Journal of Epidemiology and Community Health, 57, 147–153. Maritime Coastguard Agency, 2007. Fatigue in seafarers. MCA/181/2007. Southampton, UK: Maritime Coastguard Agency. Schmidt, K.-H., Neubach, B., and Heuer, H., 2007. Self control demands, cognitive control deficits and burnout. Work and Stress, 21, 142–154. Slaven, G., 2002. Stress audit of the Royal Navy. Thesis (DOccPsych). University of East London. Stansfield, S.A., et al., 1999. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II Study. Occupational and Environmental Medicine, 56, 302–307. Suchy, Y., 2009. Executive function: overview, assessment and research issues for neuropsychologists. Annals of Behavioural Medicine, 37 (2), 106–116. Sullivan, B. and Payne, T., 2007. Affective disorders and cognitive failures: a comparison of seasonal and nonseasonal depression. The American Journal of Psychiatry, 164, 1663–1667. Tzabar, Y., Asburg, A.J., and Millar, A.J., 1997. Cognitive failures alter general anaesthesia. British Journal of Anaesthesia, 76, 194–197. van der Linden, D., et al., 2005. Work stress and attentional difficulties: An initial study on burnout and cognitive failures. Work and Stress, 19, 23–36. Wadsworth, E.J., et al., 2008. Fatigue and health in a seafaring population. Occupational Medicine, 58, 198– 204. Wallace, J.C. and Chen, G., 2005. Development and validation of a work specific measure of cognitive failure: Implications for organisational safety. Journal of Occupational and Organisational Psychology, 78, 615– 632. Wallace, J.C. and Vodanovich, S.J., 2003. Workplace safety performance: conscientiousness, cognitive failure and their interaction. Journal of Occupational Health Psychology, 8, 316–327. Watson, D., Clark, L.A., and Tellegen, A., 1988. Development and validation of brief measures of positive and negative affect: The PANAS scale. Journal of Personality and Social Psychology, 54, 1063–1070. Weeks, D., Freeman, C.P.L., and Kendell, R.E., 1980. Enduring cognitive deficits? British Journal of Psychiatry, 137, 26–37. Williams, P.G., Suchy, Y., and Rau, H.K., 2009. Individual differences in executive functioning: implications for stress regulation. Annals of Behavioural Medicine, 37, 126–140.
