Do Sex Differences in the Association between Work Exposure and Health in the Manufacturing Industry Depend on Work Context? Results from the WOLF-Study ()
1. Introduction
Men and women differ in many aspects such as lifestyle, work exposure, and health. One of the reasons for this difference in Sweden is that the Swedish labor market is highly gender segregated both vertically and horizontally. Women work foremost in the service sector with responsibilities for caring and teaching, whereas men are mostly found within the private industries and in higher managerial positions (Statistics Sweden, 2012; Melkas & Anker, 2001). Consequently, men and women are exposed to different work-related factors (Messing, Punnett, Bond, Alexandersson, Pyle, Zahm et al., 2003; Bekker, Rutte, & van Rijswijk, 2009). Women report being more frequently on sick-leave (Statistics Sweden, 2014), they report having more troubles sleeping (Johnsson, 2006), being more tired, and they report more burnout (Norlund, Reuterwall, Höög, Lindahl, Janlert, & Slunga Birgander, 2010) than men. Whether these differences are due to biological or contextual reasons is hard, if not impossible, to disentangle. However, in order to understand this controversy better, studying the relationship between work exposure and health in the same type of industry and the same type of occupation is important. Therefore, this study aims at investigating the associations between work exposure, such as psychosocial work factors and work posture, and health in men and women with the same type of occupation working in the same type of industry; in this case the manufacturing industry.
Physical work environment has dominated the research on the association between work exposure and health in the manufacturing industry. Sex differences were found when studying injuries in the US aluminum manufacturing industries. Compared to men with the same job descriptions, women had more injuries in general and more musculoskeletal-related injuries in particular (Taiwo, Cantley, Slade, Pollack, Vegso, Fiellin, & Cullen, 2008). Kines, Hannerz, Lyngby Mikkelsen, and Tüchsen (2007) also concluded that there was a stronger association between work exposure, operationalized as industry sector, and upper extremity injuries in women than in men. It is interesting to note that even if women often have been found to be weaker in their shoulders (Miller, Mac Dougall, Tarnopolsky, & Sale, 1993) and wrists (Nordander, Ohlsson, Balogh, Hansson, Axmon, Persson, & Skerfving, 2008; Nordander, Ohlsson, Åkesson, Arvidsson, Balogh, Hansson, Strömberg et al., 2013), they work with their hands more often above shoulder height even when having the same job description as men (Dahlberg, Karlqvist, Bildt, & Nykvist, 2004). This work posture is a risk factor for neckand shoulder problems in general and given that women on average may be weaker than men in the exposed muscles, it is not surprising that they report more problems in the neck and the back than men (Widanarko, Legg, Stevenson et al., 2011), especially in the manufacturing industry (Försäkringskassan, 2011).
Research has pointed to a relationship between psychosocial factors and musculoskeletal health in women. For instance, Bildt, Backstig, & Andersson Hjelm (2006) showed that musculoskeletal health was associated with demand, control and sense of coherence in a manufacturing setting. However, when it comes to research on psychosocial work factors in relation to health, the manufacturing industry has been somewhat overshadowed by the growing service sector. Traditionally, psychosocial work factors have been defined according to the demandcontrol-support model (Karasek & Theorell, 1990; Johnson & Hall, 1988). Plenty of studies have shown a relationship between health and demand, control and social support (Belkic, Landsbergis, Schnall, & Baker, 2004; Cohen, Underwood, & Gottlieb, 2000). However, the results, especially on the relationship between demand, control and health are inconclusive (Kasl, 1996; Kristensen, 1996), and working life has changed since the 1970’s and 1980’s when the model was postulated. Today’s working life is very much characterized by smeared boundaries between working and family life thus increasing the risk of a conflict between the two (Allvin, Aronsson, Hagström, Johansson, & Lundberg, 2011). In a recent study, Hämmig and Bauer (2014) concluded that work-life conflict was the largest risk factor of both physical and psychological health. They also addressed this aspect of psychosocial work as a neglected area in occupational medicine.
Another conflict that is commonly reported in today’s working life is role conflict. Experiencing role conflict at work is stressful (Fenlason & Beehr, 1994). Being expected to perform certain tasks but, for instance not having enough resources to complete these or experiencing incongruent expectations and demands, will elicit negative emotions and decrease job satisfaction (Burke, 2002). For instance, Piko (2006) showed that role conflict contributed to emotional exhaustion and depersonalization, which are prominent aspects of Maslach’s burnout construct (Maslach & Jackson, 1981).
In an industrial setting, the relationship between psychosocial work factors and health have mostly been studied in the Finnish private industry. Factors such as social support, job autonomy, and job complexity were found to be associated over time with sickness absenteeism in a multinational forest industry (Väänänen, Toppinen-Tanner, Kalimo, Mutanen, Vahtera, & Peiró, 2003). Väänänen, Kalimo, Toppinen-Tanner, Mutanen, Peiró, Kivimäki and Vahtera (2004) also studied role conflict, organizational justice, and fairness in relation to sickness absence in blueand white collar male and female workers in the same industry. Here socioeconomic group was important for the association between the psychosocial factors and health. For instance, the authors (ibid.) found that role ambiguity increased the risk of sickness absence in white-collar men whereas organizational climate was more important to health in blue-collar women.
Sickness absence is a common outcome when studying psychosocial factors and health in the manufacturing industry. However, a promotive perspective may reduce sickness absence and save companies and the society from costs due to sick leave and loss of productivity. Therefore, this study focuses not only on factors supposed to be indicative of ill-health such as work overcommitment and fatigue, but also on neck and back pain since that is a common complaint in women in general and in women in the manufacturing industry in particular as described above. Many studies have shown that work overcommitment predicts ill-health. Work overcommitment indicates an inability to let go of work when not working (Siegrist, 1996). It has been shown to contribute to stress and stress-related diseases (Bakker, Killmer, Siegrist, & Schaufeli, 2000). Moreover, work overcommitment has been shown to predict disturbed sleep (Åkerstedt, Nordin, Alfredsson, Westerholm, & Kecklund, 2012), musculoskeletal complaints (Joksimovic, Starke, Knesebeck, & Siegrist, 2002), cardiovascular disease, and depression (Dragano, Ying, Moebus, Jöckel, Erbel, & Siegrist, 2008). In order to understand this precursor’s own predictors, it is important to study as an outcome variable.
Sickness absence is also associated with fatigue. For example, in a Dutch study, Janssen, Kant, Swaen, Janssen and Schröer (2003) showed that fatigue was related to both short and long term sickness absence and that the more fatigue the shorter the time to onset of the first sickness absence spell. Moreover, fatigue increased the risk of long term sickness absence by 35% in a study on a national Swedish sample (Åkerstedt, Kecklund, Alfredsson, & Selén, 2007), and even if there seems to be a shortage of studies investigating the causal pathways between fatigue and health, research indicates that it is an important predictor of ill-health.
By using the longitudinal WOLF study from Northern Sweden we investigate the relationship between psychosocial and physical work factors and ill-health in men and women working in the same occupations in the manufacturing industries.
2. Methods
2.1. Participants
To investigate sex differences in the relationship between psychosocial and physical work environment and health in the manufacturing industry, the WOLF (WOrk Lipids and Fibrinogen)-cohort from the Northern Sweden was used. This cohort is part of a larger cohort including civil servants in the Stockholm region. However, since manufacturing industries are most prevalent in the northern part of Sweden, the WOLF-cohort of Northern Sweden is suitable for the study question. This part of the WOLF-study was launched in 1996-1998 (WOLFNorrland) and followed up both in 2000-2003 (WOLF-follow up) and 2009 (WOLF-follow up). For the particular purposes of this study, WOLF-F (T1) and WOLF-U (T2) (with a follow-up rate of 67%) were used since the outcome measures were satisfactorily represented in both studies. The sample that answered both the WOLF-F and the WOLF-U questionnaires contains 1589 men and 286 (15%) women. The sex distribution reflects the fact that only 10% to 15% of the Swedish female working population work in the manufacturing industry. The manufacturing industries included were foremost forest industries.
2.2. Procedure
The main objective of the WOLF study is to investigate psychosocial work factors’ impact on health over time in the Swedish work force. By the help of the Occupational Health Services (OHS), companies were recruited to WOLF-N to give their employees a health examinations and a questionnaire on work situation, psychosocial work factors, lifestyle, and health. The questionnaire also contained some questions on physical work environment. In WOLF-F, only a subsample was given a health examination but everyone was given a questionnaire whereas in WOLF-U only a questionnaire but no health examination was offered. In WOLF-F and U, the participants who were offered a questionnaire and invited by mail, were given two reminders unless they actively declined to participate.
2.3. Measures
All the outcome and predictive measures are presented in detail in the Appendix.
2.3.1. Outcome Measures
Work overcommitment (WOC) was assessed by the six questions included in the WOC index included in the Effort Reward Imbalance model (Siegrist, 1996). The questions were summed and dichotomized on the median. Cronbach’s alpha was 0.82. Fatigue was defined by the single question about feeling tired in the head included in the Karolinska Sleep Questionnaire (Åkerstedt, Ingre, Broman, & Kecklund, 2008). The response option scale differed somewhat between T1 and T2 as it was five-graded at T1 and six-graded at T2. The operationalization of neckand back pain also differed somewhat between T1 and T2, but in essence both scales captured the same problem.
2.3.2. Predictive Measures
The Swedish version of the demand-control questionnaire was used to assess demand, control and social support (Karasek & Theorell, 1990). These were summed for each dimension and the demand dimension was dichotomized on the median whereas the control and social support dimensions were split on the upper quartiles. Cronbach’s alphas were 0.70, 0.66, and 0.86 respectively. The questions making up the concept of role conflict (Petersen, Kristensen, Borg, & Bjorner, 2010) were summed and dichotomized on the median. Cronbach’s alpha was 0.71. The questions on work-family conflict (WFC) were summed and divided on the quartile. Cronbach’s alpha was 0.67. Work posture was assessed by five questions that were summed and dichotomized on the median.
2.3.3. Confounding Factors
A question on what education the participants had was used where university education was contrasted with all other response options reflecting lower education. Marital status was assessed by the question “What is your marital status?”. Whether the participants worked shift or not was identified by the question “Do you work shift?” and the response option no was contrasted with all the other response options that reflected different types of shift work. Age and sex were also included, as was occupation. Occupation was identified by grouping the occupational classification codes from Swedish Occupational Classification System into those who worked with people (P), things (T), and symbols (S) (Härenstam, 1999). In the manufacturing industry, no people workers were present. To describe the other categories, symbol workers can be related to typical clerical work and thing workers as the typical manual manufacturing worker.
2.4. Statistical Analyses
Chi square analyses were used for testing group differences and logistic regression analyses for studying sex differences by interaction effects. Confounding variables were selected by the change in estimate procedure meaning that the variables that correlated with both the predictor and outcome variables and changed the betavalue in the relationship between the predictor and outcome variable by at least 10%, were selected. The analyses were conducted in two steps. First interaction analyses were performed between sex and the different predictor variables (demand, control, social support, role conflict, and work family conflict) in relation to the three health outcomes (WOC, fatigue, and neck and back pain) for thing and symbol workers respectively. Thereafter, post-hoc analyses were performed where interaction effects were found, for men and women separately to investigate effect modification. These analyses were controlled for baseline outcome variables and for appropriate confounding factors selected according to the change in estimate procedure. The IBM package SPSS version 18.0 and 22.0 were used.
3. Results
There were no significant differences between men and women in the responses in thing workers. Significantly more men than women who worked with symbols, reported high demands, role conflict, and work overcommitment whereas significantly more women than men who worked with symbols had higher education, reported low control, more fatigue at T1 and more neck-shoulder pain at both T1 and T2. Over time those who reported WOC and fatigue decreased whereas those who reported neck and back pain increased in all strata (see Table 1).
3.1. Interaction Analyses
In order to have as much control over the work context that the men and women worked in as possible, the interaction analyses were performed within the thing and symbol groups respectively. The significant interaction, and main effects are presented in relation to the occupation groups below.
3.2. Thing Workers
Work Overcommitment
A two-way interaction effect was found between sex and social support in WOC at T2 (Odds ratio [OR] 3.86; 95% confidence interval [CI] 1.09 - 13.63). No main effects were found. Post hoc analyses controlled for base.
Table1 Distribution of background, predictor, and outcome variables among men and women working with things and symbols in the manufacturing industry
N = number; ns = non-significant; S = significant difference between symbol workers; p for sex differences by chi-square test with 1 df.
line WOC showed a non-significant elevated risk of WOC at T2 due to poor social support in men. However, post hoc analyses for women, controlled for baseline WOC and education, revealed a significantly elevated risk of WOC at T2 due to poor social support. Shift work also classified as a confounding variable and when entered into the analysis, the relationship between social support and WOC in women was explained (see Table 2).
A two-way interaction effect was found between sex and demand in WOC at T2 in the employees working with things (OR 4.08; 95% CI 1.04 - 16.04). No main effects were found. Post hoc analyses controlled for baseline WOC showed an elevated risk of WOC at T2 due to high demands both in men and in women (see Table 2).
When testing whether there were any sex differences between sex and WFC in WOC, two main effects were found. WFC increased the risk of WOC (OR 3.53; 95% CI 2.40 - 5.19), when controlled for baseline WOC as did sex (OR 1.92; 95% CI 1.03 - 3.57) indicating that women were at higher risk of WOC given WFC.
3.3. Symbol Workers
3.3.1. Work Overcommitment
A two-way interaction effect was found between sex and demand in WOC at T2 in the employees working with symbols (OR 0.35; 95% CI 0.15 - 0.81). Main effects were found for demand in WOC at T2 (OR 10.37; 95% CI 3.32 - 32.39). Post hoc analyses controlled for baseline WOC showed an elevated risk of WOC at T2 due to high demands in men but not in women (see Table 2).
A two-way interaction effect was also found between sex and role conflict in WOC at T2 (OR 0.36; 95% CI 0.16 - 0.82). A main effect for role conflict was also found (OR 7.47; 95% CI 2.46 - 22.65). Post hoc analyses controlled for baseline WOC showed a significantly elevated risk of WOC at T2 due to role conflict in men whereas an insignificant under risk was found in women (see Table 2).
Moreover, a two-way interaction effect was found between sex and WFC in WOC at T2 (OR 0.41; 95% CI 0.17 - 0.97). No main effects were found. Post hoc analyses controlled for baseline WOC showed an elevated risk for WOC at T2 due to WFC in men but an insignificant risk approaching unity in women (see Table 2).
3.3.2. Fatigue
A two-way interaction effect was found between sex and role conflict in fatigue at T2 (OR 0.19; 95% CI 0.05 - 0.71). A main effect was found for sex (OR 3.50; 95% CI 1.35 - 9.07). Post hoc analyses controlled for fatigue at baseline and education, showed a significantly elevated risk of fatigue at T2 due to role conflict in men but an insignificant under risk in women (see Table 2).
Table 2. Odds ratios (OR) and 95% confidence intervals (CI) from post hoc logistic regression analyses of significant interaction effects adjusted for relevant confounding factors.
Underlining marks cut-off for dichotomization.
NOTES
*Corresponding author.