Gender inequalities in occupational health related to the unequal distribution of working and employment conditions: a systematic review

Electronic databases searched included MEDLINE (through PUBMED) EMBASE, Sociological Abstracts, LILACS, EconLit and CINAHL. The search was limited to publication dates from 01/01/1999 to 31/12/2010. This period was selected because it was during this time that the gender perspective was largely incorporated in the occupational health literature [18]. Keywords used were different terms reflecting gender and occupational health. The terms reflecting gender included: sex, gender, women, men, woman, man; the terms reflecting occupational health included: occupational health, industrial health, occupational safety, employee health, work, health, and workplace. The terms reflecting inequality (inequity, inequality and difference), were not used because otherwise they reduced significantly the results of the search conducted. The Boolean operators AND and OR were combined in a common search strategy in order to achieve the most sensitive, but not the most specific results in the search (Figure 1). Although the search strategy was not limited by the language of the publications, only articles in English or Spanish were selected, due to the researchers’ fluency in these languages.

This common search strategy identified 5,498 references, 1,308 in MEDLINE, 4,190 in EMBASE and none in the remaining databases (Figure 1). Of these, 1,153 duplicated articles were excluded. In the first phase, after applying the inclusion and exclusion criteria to the titles and abstracts (Table 1), the principal reviewer (JC) excluded 4,057 additional articles, for the most part because they did not focus on occupational health. In the second phase, after applying the selection criteria to the complete text of the remaining 288 articles, 104 were found to match the inclusion criteria. From these 104 articles, 28 were left out because they were reviews. Finally, a quality check was conducted during the second phase of the selection process. After going through the title and abstracts, 5% of the 288 included references were randomly selected. The two principal reviewers (JC and ER) independently applied the criteria for inclusion and exclusion throughout the complete text of the articles. The Kappa statistic for agreement between them was quite acceptable (k=0.85). All disagreements between JC and ER were due to different interpretations of one of the exclusion criteria, which were resolved after a short discussion. When this process was finished, the selection criteria were clarified and rewritten. Thus, it was not necessary to ask any other reviewer’s opinion since JC and ER reached agreement on all studies for which there was initial disagreement.

In the third phase, the remaining 76 articles were critically and independently appraised by two reviewers (JC and ER) using two different specific standardised evaluation guidelines appropriate to the type of the epidemiological design of the study [21, 22], both based on the STROBE statements [23].

The specific tool used to assess the quality of the 65 cross-sectional studiesl [21] comprised 27 items distributed in 8 domains with 6 categories of answer (poor, fair, good, very good, no information available, and not applicable). The domains were: a) research question, one item mainly evaluating whether the study is based on a clearly defined research question; b) participants and internal validity, five items mainly evaluating the sample adequacy and similarity to the base population and the control of selection bias; c) comparability between groups, four items mainly evaluating the study groups’ comparability and the control of selection bias; d) study variables, four items mainly evaluating the adequacy of the measurements of the main variables and the control of information bias; e) statistical analysis and control of confounders, four items mainly evaluating the adequacy of the analysis in measuring the control of confounding variables; f) results, four items mainly evaluating to which extent the results are well described, useful and precise; g) conclusions, four items mainly evaluating whether the results can be generalized to the population and to the context in which it aims to apply; and h) conflict of interest, one item evaluating whether the conflict of interests do not prejudice either the results or the conclusions of the study. The total quality score was determined as high-quality, if the majority (50% or over) of the 8 domains were classified as very good or good, unless the internal validity (evaluated through domains b to e) was classified as fair or poor; medium-quality, if the internal validity was classified as fair, or if the majority of the domains were classified as fair; and low-quality, if the internal validity was classified as poor, or if the majority of the domains were classified as poor. The internal validity was classified as fair or poor, when at least two of the four domains from b to e were scored as fair or poor, respectively.

The tool [22] used to assessed the quality of the five case–control studies included a checklist of 37 items distributed in 6 domains with 4 categories of answer evaluating if the domain was optimal or not (yes, no, partially and not applicable). The domains were: a) research question using three items; b) methods and internal validity, evaluating the participants with three items, selecting case and controls with 11, the groups’ comparability with two, the exposure with four, and the statistical analysis with five; c) results using five items; d) conclusions, using one; e) conflict of interest, using two; and f) external validity, using one.

The tool [22] to assess the quality of the six cohort studies used a checklist of 49 items distributed in 6 domains with 4 categories of answers evaluating if the domain was optimal or not (yes, no, partially and not applicable). The domains were: a) research question, using four items; b) methods and internal validity, mainly evaluating the participants with six items, the groups’ comparability with three, the exposure with nine, the effects with four, the groups’ monitoring with seven, and the statistical analysis with six; c) results, using six items; d) conclusions using one; e) conflict of interest using two; and f) external validity using one.

In both case–control and cohort studies, all six domains were taken equally into account to classify them as high, medium or low-quality studies. They were classified as high-quality, when five or more of any of the six domains were assessed as optimal; medium-quality, when three or four domains were assessed as optimal; and low-quality, when only one or two domains were assessed as optimal.

After this assessment, 41 medium and 5 low-quality studies were rejected, because the review team decided to limit the focus to articles with the highest standards of quality. Twenty-nine high-quality studies were finally included in our review.

We obtained a set of descriptive variables to characterize each of the 30 studies included in the review: first author, year of publication, country where the study was conducted, study design, study population, main dimension assessed, sample size, main objective, main findings and the relevance for the review. We also calculated the frequency of the dimension and subject mainly examined. Finally, we conducted a qualitative analysis of the descriptive results of each of the 30 studies. We focused on the differences observed between women and men in the prevalence of exposure to different working and employment conditions as determinants of gender inequalities in occupational health. In addition, we focused on the differences between women and men in the prevalence of work-related health problems as outcomes in each of the 30 studies. The statistical significance of each of the differences observed in the prevalence (p-value) was considered in the results and discussion section of this manuscript, but not in the selection process of the differences observed in the studies. Gender differences in the prevalence of exposure to the working and employment conditions were selected when the same gender difference appeared in two or more of the 30 studies included in the review, regardless of whether the difference in prevalence was statistically significant. We considered that a difference between women and men in the prevalence of exposure to the working and employment condition or in the work-related health problems was an inequality when it was avoidable, unfair, and systematically affected a higher proportion of women than men or vice-versa [24]. We grouped gender inequalities identified in the three dimensions analysed in this review: working conditions and employment conditions as determinants of gender inequalities in occupational health and inequalities related to work-related health problems as outcomes; including physical and mental health.

Most of the 30 studies included in the review [25‐54] were conducted in Europe (n = 19) and the United States (n = 5). The most frequent design was cross-sectional (n = 24), followed by cohort (n = 4) and case–control (n = 2). Most of the studies (73.3%) used samples of more than 1,000 people and were based on some type of working population (Table 2).

Table 3 shows main objective, findings, relevancy and quality score of each study included in the review. Most of the studies (n = 24) were focused mainly on how differences between women and men in the exposure to any kind of occupational hazards impact on their physical and mental health; another four [26, 27, 29, 54] out of the 30 studies examined whether women and men with similar work tasks exhibit differences in their health impact; finally, only two studies [33, 52] looked at injuries related to occupational accidents. Only one study [42] introduced the occupational social class as the main factor explaining gender differences in the exposure to working conditions and their impact on health. The quality score of most of the studies (n = 28) was over 83%.

In general terms, the studies included in the review covered a wide range of dimensions and subjects (Table 4). Most were related to the dimension of working conditions as a determinant of gender inequalities in occupational health (n = 20). Of these 20 studies, eight [25, 30, 34, 36, 37, 47, 48, 51] focused mainly on the differences between women and men in the exposure to work-related psychosocial hazards and how that fact impacts on their health. Five [25, 34, 47, 48, 51] of these 20 studies were based on the demand-control-support model, two of which followed a cohort design [34, 48]. Another two studies [36, 37] of these 20 were based on the effort-reward imbalance model, one of which followed a case–control design [37]. A smaller number of studies [32, 40, 41, 45, 46] (n = 5) were related to the dimensions of employment conditions as a determinant of gender inequalities in occupational health and another 5 studies [31, 33, 50, 52, 54] examined how work-related health problems are differentially reported by employed women and men. Only one of the 30 studies included in the review focused on social class position [42], time spent in paid work [35] and work organization [45] as the main explanatory variables for the gender inequalities observed.

Table 5 shows gender differences in the prevalence of exposure to a series of occupational hazards related to working and employment conditions and differences between women in men in reporting work-related health problems. Differences between women and men in the exposure to the demand-control-support model were analysed in 12 out of the 30 studies included in the review. In addition, differences in reporting musculoskeletal symptoms were analysed in 7 out of the 30 studies. More men than women were exposed to low support in two studies with a cohort design [34, 48], high physically demanding work in a study with a cohort design [48] and effort-reward imbalance in a study with a case–control design [37]. Conversely, more women than men were found to work with a temporary contract and in a shift-work in a study with a cohort design [48]. In addition, more women than men were found to report musculoskeletal symptoms in a study with a cohort design [28].

Whereas the studies that analyzed the differences between women and men in the prevalence of exposure to work-related psychosocial hazards were contradictory, the results were more consistent in those studies that analyzed the differences between women and men in the exposure to long work hours, high physically demanding work, and noise, with more men than women exposed to these hazards. Conversely more women than men were found to be exposed to high job insecurity. Three of the 30 studies [48, 49, 51], showed that a higher proportion of women than men were exposed to high demand and low control, but two other studies [43, 45] determined that there was a higher proportion of men who worked exposed to these conditions. None of these five studies had information about the statistical significance of differences observed in the prevalence. In addition, when high demand and low control were analysed separately, three studies [25, 34, 49] showed a higher proportion of women and four others [36, 42, 46, 47] a higher proportion of men who worked exposed to high demand. However, most studies (six [34, 36, 42, 46, 47, 49] out of seven) showed a higher proportion of women working exposed to low control and a higher proportion of men working exposed to low support (six [25, 34, 46‐49] out of six). In addition, in two studies [36, 37], one of which had a case–control design [37], a higher proportion of men than women were exposed to effort-reward imbalance. Three studies [43, 48, 51] found that a higher proportion of men than women were exposed to high physically demanding work and another three [42, 44, 49] found more men exposed to a high noise. However, no study was found showing a higher proportion of women than men exposed to these two hazards. Two studies [42, 51] showed a larger proportion of women than men with exposure to high job insecurity.

Overall, employment conditions were less favourable among women. Two studies [41, 46] showed a higher proportion of women than men working with no contract. However, none of them had available information about the statistical significance. While a higher proportion of women worked part-time (in three [32, 46, 51] out of three studies), or with a temporary contract (four [32, 42, 46, 48] out of four studies) or a temporary fixed-term contract (one [41] out of one study); a higher proportion of men worked full-time (in two [32, 46] out of two studies), or with a temporary non-fixed term (one [41] out of one) or permanent contract (two [32, 41] out of two). In addition, a higher proportion of men than women worked as supervisors (two [45, 51] out of two) or as upper managers (two [25, 47] out of two).

Women had worse physical and mental health than men. Three studies [25, 32, 51] identified a higher prevalence of self-perceived poor physical health in women than men (only one of which was statistically significant) [25]; and five [32, 40, 41, 46, 54] found poorer self-perceived mental health in women than men (only one of which statistically significant) [40]. While seven studies [28, 31, 32, 47, 49, 50, 54] observed a higher proportion of women than men reporting any kind of musculoskeletal symptoms, five of which were statistically significant [28, 31, 49, 50, 54], no studies were found showing more men than women suffering these types of symptoms. A higher prevalence of poor mental health status in women than men was found in three studies [41, 46, 54], and a higher prevalence of mental health disorders [32, 40], psychosomatic complaints [25, 54] and self-reported occupational stress [30, 51] was observed in two studies. No study found a higher prevalence in men than in women.

The majority of the results found on working and employment conditions could be explained by the powerful influence that employment, social class, and family exert on people’s everyday experiences, as well as the sexual-based division of labour that assigns different positions to men and women in these spheres of life. Moreover, these work-related gender inequalities in working and employment conditions are also linked to gender inequalities in power and resources. One of the most significant elements is the persistence of the ideology of domesticity, in which domestic work and childcare are normatively assigned to women [55].

The sexual-based division of work explains gender differences in time spent in paid work and other differences in working conditions. In this regard, the fact that we found a higher proportion of men than women who were exposed to a high level of noise at work could be explained by horizontal segregation. Horizontal segregation puts men in sectors of activity such as industrial environments and agriculture, as well as in mines, shipyards, and forges, where workers are exposed to this hazard to a much higher degree [11, 12].

According to our findings, several studies have indicated that employed women experience worse psychosocial working conditions than employed men, and that a higher health burden might result from these exposures [19, 20, 56]. In addition, previous studies have found that men experience higher job demands, effort, and overcommitment; and lower social support at work; whereas women exhibit lower job control, higher emotional job demands and higher job reward [57‐59]. In addition, women’s jobs are characterized by a greater level of monotony, with lower participation in planning, higher demands, more psychological and sexual harassment, higher exposure to the public, lower salaries, fewer prospects for promotion, and more precariousness than those of men [60]. The unequal gender distribution of work-related psychosocial hazards between women and men is mainly related to the horizontal segregation of the labour market, which concentrates women in occupations and economic activities (e.g. services) with higher exposure to work-related psychosocial hazards [6, 61]. In addition, the unequal distribution of working tasks by gender within the same job title [11, 13, 14] may expose women to even higher levels of work-related psychosocial hazards [15]. Furthermore, vertical segregation, which places women in the lowest positions of the decision making scale, reinforces this effect [6]. It has been suggested that these inequalities put women at a higher risk of physical [62] and mental disorders [63], sickness absence [5], disability [64], and mortality [65] from work-related psychosocial hazards.

Moreover, gender differences in power that place men in a better situation than women to bargain their employment conditions, could explain the gender inequalities identified in type of contract and job status, which show more men than women working with a permanent contract and occupying the higher job status positions. Vertical segregation and the “glass ceiling” phenomeno [6, 7], a metaphor for the invisible barriers that prevent women from reaching positions of power that are occupied by men reinforce these two gender inequalities. These two gender inequalities could also be caused and maintained by the so-called “sticky floor” phenomenon [6], which prevents women from loosening the emotional ties that bind them to the rest of the members of the family unit. Furthermore, the modified male breadwinner model [55], in which males are engaged in paid work and work full-time, and their female partners are engaged in unpaid and paid work but work part-time, is very well represented in the results of our systematic review. These results in terms of the higher proportion of men working full-time and the higher proportion of women working part-time could also be partly explained by the gender division of social expectations among women and men in these two spheres of work and personal or family life [66]. In developing a flexible employment regime – which generally harms women more than men – employers could also exploit these differences in expectations towards work and employment [3, 4]. This phenomenon could also explain why women perceive more job insecurity than men, and why men work longer hours in paid work than women. Women, more than men, may assume part-time work in an attempt to resolve their conflicts in balancing work and family life [55, 67]. Although women’s conflicts in balancing work and family life could be improved by working part-time hours, part-time jobs are segregated into a narrower range of occupations than full-time jobs [67, 68]. These narrow ranges of occupations are typically lower-paid, lower status, more monotonous, with fewer opportunities for advancement and related to job insecurity. In addition, part-timers have fewer social-work-benefits, less professional promotion, fewer opportunities to occupy managerial position in the company and are exposed to worse psychosocial work environments than full-timers [68‐70].

Previous studies have found that more women than men work without a contract or with a temporary contract. This could be related to possible mechanisms of discrimination, which more women than men suffer in the workplace, specifically, the discrimination associated with gender [71, 72]. This higher level of discrimination could be explained by horizontal and reinforced by the vertical segregation, which place women in job positions with less power compared to men. However, further analysis would be required to clarify this hypothesis.

Gender inequalities in health may result from the poorer working and living conditions of female workers. Although men experienced more physically demanding work than women, women experienced more musculoskeletal symptoms. This might be related to differences between women and men in the exposure to work-related hazards even working under the same job title [13, 14]. For instance, the exposure to awkward working positions and repetitive movement with low loads is more common among women than men [67, 73]. But, we have no information on these exposures in the studies included in our review. However, the gender differences related to the self-perceived physical health and musculoskeletal symptoms may also be influenced by a certain biological differences between women and men, which make women more susceptible than men to suffering musculoskeletal problems [74, 75]. For example, there are anthropometric differences between men and women in muscle, fatty tissue and bone mass. Men have greater muscle mass and women have more fatty tissue [76]. In this regard, many workplaces and tools required by workers have been designed with men in mind, but without taking into account the anthropometric differences of women [50]. It is likely that both explanations – social and biological differences between women and men – complement each other. Moreover, the “double burden” phenomenon, which refers to the double exposure to the same occupational hazards at work and at home, such as housecleaning and caretaker tasks that mostly affect women [77, 78], could explain these facts. For example, domestic work also implies exposure to ergonomic and psychosocial hazards, such as those related to informal care in families with disabled people that, besides physical and mental effort, often pose high emotional demands [67]. Women have a lower level than men not only of self-reported physical health, but also of self-reported mental health. This finding could perhaps be explained by the greater work-related psychosocial risk factors under which women work in the workplace and at home. Despite the dramatic increase of women in the labour market in recent decades, there have been no significant changes in the distribution of domestic work, even when both partners are working full time. Domestic tasks are still unequally distributed, with most of them (for example, caring for children, the elderly and disabled people) remaining women’s responsibility [67]. This fact reduces women’s recovery time after a day of work compared to men, a situation previous studies have associated with increased musculoskeletal symptoms [50].

Several previous studies show the influence of occupational social class on unequal gender distribution in the exposure to occupational hazards related to working and employment conditions [67, 72, 79‐81]. For example, women in the most advantaged occupational social class – but not those in the most disadvantaged occupational class – are more likely to work in a worse psychosocial work environment than their male counterparts [72, 81]. However, occupational social class as a determinant of gender inequalities in occupational social class only appear in our review in one study [42]. Thus, researchers should invest more efforts to incorporate the dimension of occupational social class in the analysis of gender inequalities in occupational health.

Some well-known gender inequalities such as discrimination [82], sexual harassment [83], bullying [84], and the gender-wage gap ratio in terms of median hourly wage for comparable work [85], which have also been cited as important work-related gender inequalities in occupational health in previous studies, were not identified in our review. This lack of findings could be due to either an artefact of the literature search strategy or the keywords used in our review, which was more focused on finding studies that analyze the different distribution among women and men in the exposure to occupational hazards in terms of working and employment conditions as determinants of gender inequalities in occupational health. Another explanation for this lack of findings could be that these aspects of discrimination, sexual harassment and bullying have been insufficiently investigated in the field of occupational health from a gender perspective, and more research is needed on these subjects. Although we found some studies in these fields, they were conducted only in women and thus were excluded from the review, since the gender perspective means that men must also be included in the analysis. Another limitation in this review is that some of the differences identified as existing between women and men in some studies were not statistically significant, or that a statistical analysis of significance was simply not conducted. However, the differences shown are present in the highest-quality studies available. Although MEDLINE and EMBASE include both biomedical and sociological references, the indexed scientific literature may not cover all investigations of the impact of gender inequalities as a determinant of occupational health. Therefore, many other studies that reflect gender inequalities may be also published in other documents that are less easily identifiable, the so-called grey literature. However, we applied a highly sensitive search strategy, which produced an optimal result in both databases. Another limitation could come from the possibility of missing articles published in languages other than English or Spanish. However, the search strategy was not initially limited by the language of the article, thus articles in a different language also followed the steps in the selection process. In addition, provision was made for translating and including any foreign-language article considered to be key in this field.