Occupational stress and strain in the Royal Navy 2007

Occupational Medicine 2008;58:534 539 Published online 28 October 2008 doi:10.1093/occmed/kqn136 Occupational stress and strain in the Royal Navy 2007 R. S. Bridger, K. Brasher, A. Dew and S. Kilminster Background Previous surveys of psychological strain in the Naval Service (NS) have shown higher than expected levels of strain when compared to the general population. Aim To repeat the survey last carried out in 2004 and to obtain further information on the nature of the occupational stressors associated with strain. Methods General Health Questionnaire-12 strain rates and job/life stressors were measured using a Work and Well-Being Questionnaire. Models of strain were developed for male and female personnel in the Royal Navy (RN) and males in the Royal Marines (RM). Results The response rate was 57%. The psychological strain rate was 31.5% overall. Personnel suffering from strain tended to be overcommitted to work, had low levels of commitment to the NS and had suffered stressful life events (SLEs) in the previous 12 months. Strain rates declined with age and rank in males, but not in females. Strain was significantly positively correlated with levels of overcommitment, effort reward imbalance (ERI), role conflict, work family conflict, organizational commitment and exposure to SLEs. Models of strain in the males and females in the RN and in the RM accounted for between 37 and 44% of the variance in strain. Conclusions The survey provides evidence for both the demand control and ERI models components of these models contribute independently to strain. High levels of commitment to the organization were associated with lower strain and exposure to SLEs to higher strain. Key words Occupational stress; overcommitment; Royal Navy; stressfull life events. Introduction The Naval Service (NS), consisting of the Royal Navy (RN) and Royal Marines (RM), had a psychological strain prevalence rate of 32% in 1999 and 33.5% in 2004 [1] higher than the rate in the general UK population (18% [2]) and comparable to that in the UK Police [3]. Strain, measured by the General Health Questionnaire (GHQ) [4], is a response to external stress, characterized by symptoms of anxiety and depression. The rate was highest in females and in younger personnel serving at sea. Lower levels of commitment to the organization were associated with strain in males and females. Dissatisfaction with the physical work environment was associated with strain in RN females and RM. Role conflict was associated with strain in all groups. Work family conflict was associated with strain in males, but not in females. Statistical models accounted for 20 30% of the variance in strain. Institute of Naval Medicine, Crescent Road, Alverstoke PO12 2DL, UK. Correspondence to: R. S. Bridger, Institute of Naval Medicine, Crescent Road, Alverstoke PO12 2DL, UK. Tel/fax: 144(0)2392 768220/504823; e-mail: hhfd@inm.mod.uk Although these findings were of use in the formulation of stress management policy, it was considered that the nature of the strain was not well understood. In the NS, deployments and work patterns change both ashore and afloat. Periods of high workload may be interspersed with periods of low workload. A longitudinal study over a 6-month deployment on a warship at sea [5] found a statistically significant drop in the strain rate of males over the course of the deployment. It is conceivable that strain is a transient problem for many and that there is a significant amount of churn (e.g. those reporting high strain in 1999 are unlikely to be the same as those reporting it in 2004). The link between chronic psychological strain and health is well established. For example, Salonen et al. [6] found that prolonged strain predicted musculoskeletal disorders, nervous system, eye, ear and metabolic diseases and mental disorders in Finnish factory workers studied over a 20-year period. The Whitehall II study showed that an individual s position within the organizational hierarchy was a better predictor of heart attack than obesity, smoking or high blood pressure [7]. Employees in low-grade jobs were almost four times as likely to suffer a heart attack as a permanent secretary at the top of the Ó The Author 2008. Published by Oxford University Press on behalf of the Society of Occupational Medicine. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

R. S. BRIDGER ET AL.: OCCUPATIONAL STRESS AND STRAIN IN THE ROYAL NAVY 2007 535 hierarchy. Schei [8] found that strain was associated with physical inactivity and the consumption of alcohol, tobacco, cannabis and junk food in Norwegian conscripts. The existence of similar associations in the NS would shed light on the processes by which mental and physical disability develop over time. Stansfield et al. [9] measured job characteristics and psychiatric disorder in 10 308 civil servants in the Whitehall II study. Decision latitude (decision authority and skill discretion), job demands and social support were measured in three successive phases 1985 88, 1989 and 1991 93. Effort reward imbalance (ERI) was also measured. Psychiatric disorder was measured using the GHQ-30. At Phase III, 21% of men and 25% of women were GHQ cases. High efforts combined with low rewards were associated with a high risk of psychiatric disorder. The two well-researched models of the link between external job demands (occupational stress) and psychological strain are Karasek s [10] Demand Control Support (DCS) model and Siegrist s [11] ERI model. Karasek s model predicts strain when demands are high and when employees have little control over their work. Social support from management and/or co-workers acts as a buffer to mitigate the effects. The ERI model focuses on the relation between the extrinsic effort required to do a job and the rewards. Strain occurs when employees perceive an imbalance between effort and reward. This paper reports the first phase of a cohort study to determine the prevalence of chronic versus acute strain in the NS and to investigate links between strain and adverse health outcomes. The scope of the study was expanded to include a wider range of stressors than in 1999 and 2004, including exposure to stressful life events (SLEs). It was not designed to capture data on operational stress, bullying or sexual harassment because these are managed differently. Methods This cross-sectional study comprised Phase I of a 5-year cohort study of strain in NS personnel. The Work and Well-Being Questionnaire (WWBQ) used in the 1999 and 2004 surveys [1] was modified to include scales from the ERI Questionnaire (with permission) [11]. Nine items on satisfaction with living conditions were included after consultation with subject matter experts. Six items on coping technique were included on an empirical basis to distinguish between those who habitually utilize strategies based on avoidance and those who take a more active approach towards work problems. The Positive and Negative Affect Scale (PANAS) is a valid, reliable scale that is used to measure mood [12]. It was included to control for the confounding effects of mood on stressor strain relationships. A modified version of the Holmes and Rahe [13] Social Readjustment Scale was included to measure exposure to SLEs. Thirty copies of the modified WWBQ were circulated to a convenience sample of volunteer NS personnel on a Monday morning and re-administered to the same individuals on the following Friday to determine test retest reliability and factor structure. In the case of the ERI scales and PANAS, Cronbach s alpha and item-remainder correlations were calculated enabling the scales to be reduced while retaining their psychometric properties. Factor analysis of the modified WWBQ revealed a very similar factor structure to that of the previous WWBQ regarding physical work environment, autonomy and control, provision of resources, role conflict, organizational commitment, leader support, peer support, work family conflict, domestic support and intention to leave. The original ERI scale had 17 items and was successfully reduced to four. Cronbach s alpha for the new reduced scale was maintained from 0.726 to 0.720 and the test retest reliability improved from 0.82 to 0.84. The original overcommitment Scale (OCS) had six items, successfully reduced to three (Cronbach s alpha for the reduced scale was improved from 0.56 to 0.83 and the test retest reliability increased from 0.86 to 0.87). The SLE scale had good test retest reliability (0.93 for major events scoring.50 and 0.89 minor events scoring,50). Ethical approval was obtained from the Ministry of Defence independent ethics committee. The sampling procedure and contact methods were as used in the original survey [1] in which the sampling rate was 4.2% for males and 20% for females (the latter to ensure adequate subsample size for analysis). The sample was stratified by branch. The sample sizes were chosen such that the maximum errors on unknown percentages indicating X (e.g. strain case ) were 1.5% for males and 2.5% for females. Because the 2007 survey was the first phase of a 5-year cohort study, the sample sizes were increased substantially to account for attrition over the 5- year period (to 10% of males and 60% of females). Questionnaires were posted in January 2007. Nonrespondents were targeted in two subsequent mail-outs in February and April. The survey was closed in July 2007 to allow for returns from distant locations. All data were initially entered into Microsoft Excel and validated by double data entry cross validation in addition to outlier and range checks. Subsequently, all data were analysed by Statistica v6 (Tulsa, OK, USA). In addition, oblique and orthogonal varimax factor analyses were carried out, confirming the factor structure previously given by the main author of the WWBQ [14]. Strain was presented as prevalence rates and 95% confidence intervals (CIs) broken down into relevant subsamples e.g. RN versus RM. Forward stepwise regression procedures were carried out by regressing GHQ score as either a continuous variable on the WWBQ predictors

536 OCCUPATIONAL MEDICINE or as a strain case by dichotomous discriminant function analyses and confirmatory logistic regression. Residual plots were examined. Only models with significant predictors are reported. Means and 95% CIs for the WWBQ scale scores were calculated and compared to these previous surveys in order to identify any trends between response rate and scale means. Results The sample size was 4949 NS personnel, 407 of whom could not be contacted, resulting in a total of 4542 questionnaires which were sent to personnel. The response rate was 57% (2596 returns). The mean age of the respondents was 34.7 years (compared to 27.9 years for non-respondents, which was significantly lower). Table 1 shows the response rates. Table 1 suggests that males,25 had the poorest response rate. Overall, the response rates were lower than previous surveys conducted in 2004 (68%) and 1999 (78%) [1]. Table 2 shows that, although some of the 2007 scale scores do differ from previous surveys, there was no systematic trend with declining response rate. The male ratings data are presented separately to demonstrate that although this group had a particularly low response rate, this did not bias their scores. The 2007 male strain rate was 27.8% (95% CI of 26 30%), which did not differ from the combined strain rate for males in 1999 and 2004 (30.9%, 95% CI of 29 33%). For females, the strain rate was 37.3% (95% CI of 34 40), which also did not differ from the combined 1999/ 2004 strain rate (42.7%, 95% CI of 39 46%). As in previous surveys, the strain rate in females was significantly higher than the strain rate in males. Exposure to stressors was measured using a five-point Likert scale a score of 5 indicated a negative affective response to the item and a score of 1 the opposite. A score of 3 indicated neutral. The WWBQ scale scores were summed and the means and standard deviations were calculated (Table 3). Overall, the mood of Naval personnel was positive, with commitment to the service, support from leaders and peers and a degree of autonomy and control in their work. Work family conflict and lack of resources were rated negatively, as was the balance between effort and reward. There was a high level of overcommitment to task or role. Table 4 presents the results of multiple regression analyses of strain in RN males and females and in the RM. These analyses yielded additive models that identify the variables associated with strain and the percentage of variability in strain accounted for. The right hand column, percentage of variance in strain, shows the increase in variance explained as each stressor is included in the model. In total, the models explain 44% of the variance in strain Table 1. Total sample, number of respondents and response rates by gender, rank and age for the 2007 survey All in RN females, 41% of the variance in strain in RN males and 37% of the variance in strain in the RM. Discussion Original full sample (n) Response numbers (n) 4542 2596 57 Male Rank RN officer 530 441 83 RN rating 1823 959 53 RM officer 69 46 67 RM other ranks 469 153 33 Age (years),25 619 154 25 25 34 1011 403 40 $35 1261 1042 83 Female Rank RN officer 334 287 86 RN rating 1275 686 54 RM other ranks 42 24 57 a Age (years),25 623 238 38 25 34 682 425 62 $35 346 334 97 a RM bandswomen. Response rate (%) The main predictors of psychological strain were overcommitment to work role and under-commitment to the NS. The strain rate was unchanged since 1999. A small percentage of the explained variance in strain was due to exposure to SLEs. The main weakness of the present study was the response rate of 57% and the particularly low response rate for younger males. Response rates to the NS surveys have declined markedly since 1999. However, comparing the present data with that of previous surveys with much higher response rates, there appears to be no systematic bias due to low response. A plausible explanation for the declining response rate is increased operational tempo. The rapid rotation of personnel through different deployments may render contact addresses obsolete. The study has major strengths. The WWBQ is psychometrically robust and grounded on well-validated theories of occupational strain. The sample size is large and previous surveys with high response rates have enabled norms to be established for many of the variables. The models for RN males, RN females and for the RM accounted for 41, 44 and 37% of the variance in strain in these groups. This is a considerably larger proportion of the variance than was accounted for in the analysis of

R. S. BRIDGER ET AL.: OCCUPATIONAL STRESS AND STRAIN IN THE ROYAL NAVY 2007 537 Table 2. Mean stressor Likert scale scores and 95% CIs for the total sample compared with male ratings, in 1999, 2004 and 2007 (high scores indicate greater negativity) 1999 Mean (CI) 2004 Mean (CI) 2007 Mean (CI) All Physical work environment 1.95 (1.91 1.99) 1.94 (1.91 1.97) 2.68 (2.65 2.71) a Autonomy and control 3.00 (2.95 3.05) 2.89 (2.84 2.94) 2.88 (2.84 2.92) Role conflict 3.08 (3.03 3.13) 3.03 (2.98 3.08) 3.05 (3.01 3.09) Lack of resources 3.42 (3.38 3.46) 3.41 (3.37 3.45) 3.55 (3.52 3.58) a Organizational commitment 2.83 (2.79 2.86) 2.70 (2.67 2.73) 2.45 (2.42 2.48) a Leader support 2.24 (2.19 2.29) 2.25 (2.21 2.29) 2.25 (2.21 2.29) Peer support 2.52 (2.48 2.56) 2.51 (2.47 2.55) 2.56 (2.53 2.59) Work family conflict 3.50 (3.46 3.54) 3.45 (3.41 3.49) 3.46 (3.43 3.49) Male ratings Physical work environment 2.14 (2.08 2.20) 2.10 (2.04 2.15) 2.87 (2.82 2.93) a Autonomy and control 3.10 (3.03 3.18) 2.94 (2.87 3.01) 2.93 (2.86 2.99) Role conflict 3.22 (3.14 3.29) 3.10 (3.03 3.18) 3.16 (3.10 3.23) Lack of resources 3.54 (3.49 3.60) 3.52 (3.47 3.58) 3.71 (3.65 3.76) Organizational commitment 2.98 (2.93 3.03) 2.85 (2.81 2.90) 2.63 (2.58 2.68) a Leader support 2.27 (2.20 2.33) 2.26 (2.20 2.33) 2.28 (2.22 2.34) Peer support 2.62 (2.56 2.68) 2.60 (2.55 2.66) 2.59 (2.53 2.65) Work family conflict 3.67 (3.62 3.73) 3.58 (3.53 3.64) 3.56 (3.50 3.61) a 2007 95% CIs do not overlap with one or more previous surveys. Table 3. Mean stressor Likert scale scores and SDs listed in terms of positive, neutral and negative affective responses Mean Positive ratings Leader support 2.24 0.92 Organizational commitment 2.45 0.92 Peer support 2.56 0.89 Physical work environment 2.69 0.85 Physical living environment 2.78 0.94 Autonomy and control 2.88 1.01 Positive mood 3.75 0.71 Negative mood a 1.60 0.72 Neutral ratings Approachability of leader 3.07 1.05 Role-conflict 3.06 1.12 Financial reward 2.07 1.16 Negative ratings Work family conflict 3.45 0.83 Lack of resources 3.55 0.87 ERI b 2.49 0.67 Overcommitment 2.58 0.67 SD, standard deviation. a The negative mood mean score is positive because it indicates an absence of negative mood. b Scores.1 on the ERI scale indicate dissatisfaction. the combined 1999 and 2004 data [1] and is due to the inclusion in the WWBQ of scales from the ERI model, the control of the confounding effects of dispositional mood and the measurement of exposure to SLEs. Whereas the previous surveys [1] found some support for the DCS model with role conflict, a lack of autonomy and control and leader and peer support accounting for a significant proportion of the variance in strain, the present findings present a different perspective. Personnel suffering from strain is likely to be overcommitted to their work role, while lacking commitment to the RN, and, independently, to have been exposed to SLEs outside of work. The OCS accounted for a considerable proportion of the explained variance in strain (18, 14 and 9% of the variance in RN males, females and the RM, respectively). The only strain-related factor from the previous surveys, best conceptualized as a strain buffer, which remained in the 2007 models was organizational commitment. Again, organizational commitment, as a component of the psychosocial work environment, is separate from the DCS model and the findings strongly suggest that attempts to engender commitment to the NS as an organization should be central to future stress management. Overcommitment is regarded as intrinsic effort in the ERI model. Overcommitted people have difficulty disengaging from work. The survey findings suggest that it is very common in the naval population. Some 54% showed signs of overcommitment, while.9% were classified as strongly overcommitted. Overcommitment may share trait-like features with strain that are derived from neuroticism. People with high levels of neuroticism may have a predisposition to both overcommitment (excessive striving, inability to disengage from work outside of working hours) and strain. This argument is similar to the suggestion that the correlation coefficients were inflated by common method variance due, for example, to response bias such as social desirability or stoicism [15]. A similar SD

538 OCCUPATIONAL MEDICINE Table 4. Percentage of variance in strain explained by predictor variables in the multiple linear regression models a Regression coefficient Multiple correlation coefficient RN females Overcommitment 0.24 0.42 18 Negative mood 0.25 0.51 27 Organizational 0.11 0.58 34 commitment SLE a 0.18 0.62 38 Positive mood 20.17 0.63 40 ERI 0.12 0.65 42 Work family 0.12 0.66 44 conflict RN males Overcommitment 0.22 0.38 14 Negative mood 0.24 0.48 23 Organizational 0.12 0.56 31 commitment SLE a 0.16 0.59 35 ERI 0.12 0.61 37 Positive mood 20.11 0.62 39 Leader support 0.07 0.63 40 Role conflict 0.06 0.63 40 Autonomy and 0.05 0.63 41 control Work family 0.05 0.63 41 conflict RM Negative mood 0.27 0.39 15 Organizational 0.22 0.51 25 commitment Overcommitment 0.27 0.58 34 Positive mood 20.16 0.6 36 SLE 0.13 0.61 37 a All models statistically significant, P, 0.001. Cumulative percentage of variance in strain (%) argument can be made about the mood data. Several authors have investigated the possibility that the presence of employees with a negative disposition may inflate the relationships between work stressors and strain [16]. This is because such people tend to view everything in a negative light and would be more likely to find work stressors distressing. The mood scores do indeed account for a significant proportion of the explained variance in these models and this may offer a partial explanation of why work stressors such as role conflict that were found to account for significant variance in strain in the prior NS surveys [1] no longer appear in the models. The present findings indicate that the DCS model offers an incomplete description of stress in the NS. Stress will be managed more effectively by focusing on commitment and intrinsic effort. Those most susceptible to strain would be relatively uncommitted to the NS as an organization while having a high level of commitment to their work and low mood. Given the large impact of low organizational commitment, it may be advantageous to explore ways of fostering commitment to the NS and to the peer group. Overcommitment to role should be managed by giving positive feedback, reducing long hours and ensuring appropriate disengagement by, for example, ensuring that annual leave allowances are taken. Recent research in the Dutch Police Force [17] has indicated that work home interference is a precursor of burnout and depression. Previous NS surveys have shown work family conflict to be an important stressor in RN males and in RM. The present version of the WWBQ contained sections on both SLEs and work family conflict. SLEs were found to be a generic stressor with strain increasing according to the number and severity of events experienced in the last 12 months (events ranged in severity from change in personal habits to death of a family member). Thus, it would seem that for NS personnel, it is not separation from families that is stressful but separation when things go wrong. Recent research on stress in university academics [18] found that ERI was associated with physical ill-health and low job satisfaction (high efforts associated with ill-health and low rewards with low satisfaction). Overcommitment was strongly related to physical symptoms. The NS cohort is currently being followed until 2012 and a subsequent paper will report on the 6-month prevalence of strain and strain, psychosocial stressors and physical health. Key points The ERI model provides a better description of acute strain in Naval personnel than the DCS model. Neither model provides a complete explanation commitment to the organization and exposure to stressful events outside of work are also important. Generic models of job strain may be of limited applicability to specialized occupations, which may require their own explanatory models if psychological strain at work is to be managed effectively. Conflicts of interest None declared. References 1. Bridger RS, Kilminster S, Slaven G. Occupational stress and strain in the naval service: 1999 and 2004. Occup Med (Lond) 2007;57:92 97. 2. Dollard MF, Winefield MA, Winefield HR. Occupational Stress in the Service Professions. London: Taylor and Francis, 2003.

R. S. BRIDGER ET AL.: OCCUPATIONAL STRESS AND STRAIN IN THE ROYAL NAVY 2007 539 3. CollinsPA, Gibbs ACC. Stress in police officers: a studyof the origins, prevalence and severity of stress-related symptoms within a county police force. Occup Med (Lond) 2003;4:256 264. 4. Goldberg D, Williams P. A Users Guide to the General Health Questionnaire. London: NFER Nelson, 1998; 129. 5. Slaven G. Work-related Strain in the Naval Service. London: University of East London, 2002 Dissertation. 6. Salonen PH, Arola H, Nygard CH, Huhtala H. Long-term associations of stress and chronic diseases in ageing and retired employees. Psychol Health Med 2008;13:55 62. 7. MarmottMG, DaveySmithG, Stansfield S, PatelC, NorthF, Head J. Health inequalities among British civil servants: the Whitehall II study. Lancet 1991;337:1387 1393. 8. Schei E. A strengthening experience? Mental distress during military service. Soc Psychiatry Psychiatr Epidemiol 1994;29: 40 45. 9. Stansfield SA, Fuhrer R, Shipley MJ, Marmot MG. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II Study. Occup Environ Med 1999; 56:302 307. 10. Karasek RA. Job demands, job decision latitude, and mental strain: implications for job redesign. Adm Sci Q 1979;24: 285 308. 11. Siegrist J. Adverse health effects of high effort/low reward conditions. J Occup Health Psychol 1996;1:27 41. 12. Watson D, Clark LA, Tellegen A. Development and validation of brief measures of positive and negative affect: the PANAS scales. J Pers Soc Psychol 1988;54: 1063 1070. 13. Holmes D, Rahe J. The social readjustment scale. J Psychosom Res 1967;11:213 218. 14. Slaven G, Shariff A, Pethybridge R. Stress in the Naval Service: Stage 1 Identification of Workplace Stressors, 1999 Unpublished MoD Report. 15. Burton-Jones A, Straub DW. Minimizing method variance in measures of system usage. In: Papp R, ed. Proceedings of the Southern Association for Information Systems (SAIS). Savannah, GA: Southern Association for Information Systems, 2004; 336 342. 16. Jex SM, Spector PE. The impact of negative affectivity on stressor-strain relations: a replication and extension. Work Stress 1996;1:36 45. 17. van Hooff MLM, Geurts SAE, Taris TW et al. Disentangling the causal relationships between work-home interference and employee health. Scand J Work Environ Health 2005;3:15 29. 18. Kinman G, Jones F. Effort-reward imbalance and overcommitment: predicting health and job satisfaction in UK academic employees. The British Psychological Society Division of Occupational Psychology Conference, January 9 11, Stratford upon Avon 2008. BPS publications, 2008.