==== Front BMJ OpenBMJ OpenbmjopenbmjopenBMJ Open2044-6055BMJ Publishing Group BMA House, Tavistock Square, London, WC1H 9JR bmjopen-2017-02077010.1136/bmjopen-2017-020770Occupational and Environmental MedicineResearch15061716Which work-related characteristics are most strongly associated with common mental disorders? A cross-sectional study Rivière Mathieu 12Leroyer Ariane 3Ferreira Carreira Lionel 3Blanchon Thierry 1Plancke Laurent 4Melchior Maria 1Younès Nadia 56 1 Sorbonne Université, INSERM, Institut Pierre Louis d'épidémiologie et de Santé Publique (IPLESP UMRS 1136), F75013, Paris, France 2 Department of Infectious Diseases, Centre Hospitalier Régional, Orléans, France 3 University of Lille, France 4 Regional Federation of Research in Psychiatry and Mental Health Hauts-de-France, Lille, France 5 EA 40-47 University of Versailles Saint-Quentin, Versailles, France 6 Academic Unit of psychiatry for adults, Versailles Hospital, Versailles, France Correspondence to Mr Mathieu Rivière; mathieu.riviere@iplesp.upmc.fr2018 5 8 2018 8 8 e02077028 11 2017 18 5 2018 26 6 2018 © Author(s) (or their employer(s)) 2018. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.2018This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.Objectives Studies exploring work-related risk factors of common mental disorders (CMDs), such as major depressive disorder (MDD), generalised anxiety disorder (GAD) or alcohol abuse, have generally focused on a limited set of work characteristics. For the first time in a primary care setting, we examine simultaneously multiple work-related risk factors in relation to CMDs. Method We use data from a study of working individuals recruited among 2027 patients of 121 general practitioners (GPs) representative of the Nord-Pas-de-Calais region in the North of France (April–August 2014). CMDs (MDD; GAD; alcohol abuse) were assessed using the Mini-International Neuropsychiatric Interview. Six worked-related factors were examined (work intensity, emotional demands, autonomy, social relations at work, conflict in values and job insecurity). Several covariates were considered (patient, GP and contextual characteristics). To study the association between workplace risk factors and CMDs, we used multilevel Poisson regression models adjusted for covariates. Results Among study participants, 389 (19.1%) met criteria for MDD, 522 (25.8%) for GAD and 196 (9.7%) for alcohol abuse. In multivariable analyses adjusted for covariates, MDD/GAD was significantly associated with work intensity (RR 1.16, 95% CI 1.06 to 1.27) (absolute risk=52.8%), emotional demands (RR 1.24, 95% CI 1.13 to 1.35) (absolute risk=54.9%) and social relations at work (RR 0.78, 95% CI 0.70 to 0.87) (absolute risk=15.0%); alcohol abuse was associated with social relations at work (RR 1.25, 95% CI 1.01 to 1.53) (absolute risk=7.6%) and autonomy (OR 0.82, 95% CI 0.67 to 0.99) (absolute risk=8.9%). Conclusions Several workplace factors are associated with CMDs among working individuals seen by a GP. These findings confirm the role of organisational characteristics of work as a correlate of psychological difficulties above and beyond other sources of risk. primary carepsychiatryIle-de-France region – DIM GestesNord – Pas-de-Calais regional health agency (ARS)special-featureunlocked ==== Body Strengths and limitations of this study Cross-sectional study design. Study of occupational factors in relation to common mental disorders among working adults in primary care evaluated with a standardised diagnostic tool in a large sample. The inclusion of participants living in the Nord-Pas-de-Calais region—one of the poorest in France—and the selective participation of general practitioners who took part in the study, may have led to an over-representation of patients with psychological disorders. Introduction Individuals who are part of the labour force are generally in better health than the unemployed,1 however, work can also have negative effects on somatic and psychosocial health.2 A study conducted among general practitioners (GPs) trained in occupational medicine found that mental health issues are frequently attributed to work.3 They are responsible for most of sickness absence and long-term work incapacity.4 In France, data from the national health insurance show that 20% of sickness absences are caused by mental disorders, and this proportion is even higher for long-term sickness absences (on average 111 days).5 The most frequent mental health difficulties among working individuals include mood, anxiety and substance use disorders (particularly alcohol-related problems), which can be grouped as ‘common mental disorders’ (CMDs).6 A systematic review of the literature in European countries shows that there is great diversity in the ascertainment of mental disorders and thus the prevalence estimates vary between countries. The authors suggest that the study of a larger range of diagnoses and the standardisation of methods can help the comparability across countries.7 The association between work and CMDs is bidirectional: work has been shown to be a risk factor of poor mental health,8 but the presence of a CMD can also influence job performance and well-being.9 10 Other risk factors of CMDs include individuals’ sociodemographic characteristics including being divorced or widowed, having a low educational level, older age, female gender,11–13 certain genetic factors14 and a history of chronic somatic or psychiatric disorders.15 Environmental factors (eg, social and material deprivation, etc) were described and show that low socioeconomic status was associated with higher rates of depression.11 12 Psychosocial factors related to the work environment are of particular interest because they may be more easily prevented than those which result from life events and are often unavoidable. Three main theoretical models have been proposed to explain relations between work characteristics and mental health. First, Karasek and Theorell16 argued that psychological demands, decision latitude and social support are especially important. Second, Siegrist17 proposed that what matters most is the subjectively ascertained effort–reward balance. A third model, developed by Elovainio et al, put an emphasis on the role of organisational justice including interpersonal comparison, that is, to say comparison of the response of the company in the same situation for different employees.18 Several studies evaluate the impact of work on mental health using these theoretical models.8 19 20 Overall, the risk of mental disorders is higher when individuals experience high job demands, low job control, high effort–reward imbalance or low organisational justice. As work organisation is evolving, other psychosocial factors described as ‘emergent’ have appeared in recent studies (eg, job insecurity, conflicts in values)21–24: Workers experiencing high job insecurity or role conflicts also seem to have a higher levels of CMDs.21 22 A recent systematic meta-review identified three overlapping categories of work-related risk factors that may contribute to the development of common mental health problems: imbalanced job design, occupational uncertainty and a lack of values and respect in the workplace.8 This review did not precisely describe different CMDs (major depressive disorder (MDD) was the most frequent outcome, generalised anxiety disorder (GAD) and alcohol abuse being less explored8 25 26). Additionally, most studies were based on self-reported questionnaires and not validated diagnostic interviews. Work-related risk factors are also influenced by changes in society and work environments (globalisation, demographic change, job specialisation, communication load, new forms of work organisation, industry 4.0,27 etc). A French study assessed changes in psychosocial work factors between 2006 and 2011 and reported that some worsened (decision latitude, social support, reward, role conflict and work–life imbalance) over that period. These changes have been shown to vary with age, occupation, sector activity and type of contract.28 The objective of this study is to assess the association between GAD, MDD and alcohol abuse in a primary care setting, testing different psychosocial work-related risk factors. Combining emergent and classical factors is important in order to identify which are most strongly related to workers’ mental health, as outlined in the meta-review conducted by Harvey et al.8 Since GPs usually are the first contact point for employees in the healthcare process, the evaluation of primary care patients is of paramount importance.29 30 In primary care, the prevalence of CMDs is high, ranging from 3%22 to 25% for anxiety disorders,13 29–32 6%13 to 25% for depression11 29–32 and 2%30 to 11% for alcohol abuse.29 30 Two studies conducted in the UK show that one-third of patients seeing a GP for work-related reasons have a mental health issue.3 33 Yet GPs often have difficulties managing their patients’ work-related mental health problems, as they often lack negotiation strategies regarding sick leave, communication skills and cooperation with occupational physicians.34 GPs encounter a variety of workers with systematic, unsystematic or non-existing occupational health services at their workplace. A better understanding of work-related factors associated with individuals’ mental health is important to help GPs consider specific actions. Methods Design and study population Héraclès is a cross-sectional exploratory study conducted between April and August 2014 among working individuals consulting a primary care physician in the Nord-Pas-de-Calais region in the North of France. Patient and public involvement The number of subjects needed and the set-up of the study have previously been described.35 Briefly, with an estimated prevalence of 20%, to have a precision of 10%, we aimed to include 2000 patients via their GP. Participating GPs gave an oral consent to participate and were asked to randomly include a maximum of 24 patients who met the following criteria: being (1) actively employed and (2) aged 18–65 years, regardless of the reason of their medical appointment. GPs were asked to include the first two patients who met study inclusion criteria in each randomly selected time slot which had previously been defined with the GP. Approximately, one-fourth of the GPs in the region, selected to be representative of those practising in 15 areas of Nord-Pas-de-Calais region, were contacted to participate in the study. Participating GPs gave written information to their patients regarding the study and asked them to sign an informed consent. This study was conducted by the Sentinelles network,36 part of the INSERM-Paris Sorbonne University research unit UMR-S 1136. This research group has a standing authorisation from the French independent administrative authority protecting privacy and personal data to conduct research among GPs and their patients (CNIL no 471 393). Data collection Participating GPs received a 15 min phone training regarding the study protocol and questionnaire. After their regular appointment, GPs interviewed participating patients for the purposes of the study. Study questionnaires included information on: Measurement of CMDs CMDs were measured using a standardised diagnostic interview: the Mini-International Neuropsychiatric Interview (MINI) that was used as a screening tool. The MINI is, a structured clinical interview that enables the diagnosis of mental disorders based on the Diagnosis and Statistical Manual of Mental Disorders, fourth edition.37 Specifically, three different diagnoses were ascertained: MDD (in the preceding 2 weeks), GAD (in the preceding 6 months) and alcohol abuse (in the preceding 12 months). The sensibility of the MINI varied between 83% and 94% (MDD: 94%; GAD: 88%; Alcohol: 83%), the specificity between 72% and 97% (MDD: 79%; GAD: 72%; Alcohol: 97%) and the Kappa concordance coefficient between 0.36 and 0.82 (MDD: 0.73; GAD: 0.36; Alcohol: 0.82). The inter-rater and test–retest reliability measured by Kappa coefficient were good, respectively, 0.88–1 and 0.76–0.93.38 Work characteristics Work characteristics were self-reported by the patient to their GP. We used a national French questionnaire proposed by experts in the field based on the international scientific literature and after auditioning Karasek and Siegrist.23 It combines (1) questions measuring psychological demands—work control—social support developed in Karasek’s model16 (two questions about decision latitude, four questions about psychological demands and two questions about social support); (2) questions measuring effort/reward balance based on Siegrist’s model17 (three questions about rewards and one question about overinvestment); (3) questions about organisational justice from Moorman’s questionnaire39; (4) questions from the Copenhagen Psychosocial Questionnaire40 and from the General Nordic Questionnaire for Psychological and Social Factors at Work41 or from WOrking Conditions and Control Questionnaire.42 Overall, the questionnaire included 20 work-related items exploring 6 different areas (online supplementary appendix 1): (1) five related to work intensity and duration (contradictory orders, excessive amount of work, too much to think about at work, difficulties in balancing work and family life, time needed for work), (2) six concerning emotional demands (contacts with customers/beneficiaries, contact with people in distress, conflicts with customers/beneficiaries, the need to hide emotions, fear, exposure to aggressions), (3) two regarding autonomy (limited decision-making possibility, full use of skills), (4) three on the quality of social work relations (full recognition of the work performed, support from colleagues, support from superiors), (5) two concerning conflicts in values (possibility to perform quality work, doing disapproved things), (6) two about job insecurity (ability to work until retirement, fear of job loss). For four of these items (contacts with the public at work, contacts with people in distress, contradictory orders, ability to work until retirement) the response was either ‘yes’ or ‘no’, and for other factors the responses were ‘always’/‘often’/‘sometimes’/‘never’ numbered from 1 to 4. The reliability of questions pertaining to work characteristics was assessed by computing an omega coefficient.43 This coefficient varied between 0.35 and 0.79. The reliability was higher for social relations at work (ω=0.72), emotional demands (ω=0.75) and work intensity (ω=0.79) than for autonomy (ω=0.66), job insecurity (ω=0.50)or conflicts in values (ω=0.35). 10.1136/bmjopen-2017-020770.supp1Supplementary file 1 Covariates Patient’s characteristics We considered already described previously risk factors of CMD11:Past somatic problems. Previous mental health problems/disorders. Sociodemographic (age, gender, family status, family income, level of education). Occupational grade44: blue-collar (farmer/manual worker), pink-collar (technician/associate professional/clerk/service worker) or white-collar (manager/professional).45 Company size. Job instability assessed based on the type of contract (temporary vs permanent). Healthcare characteristics46: Reason for medical appointment (somatic, psychological, chronic disease management). GP’s sociodemographic characteristics (age, gender). Practice characteristics (size; comfort with psychological distress issues; opportunity to collaborate with mental health specialists). Contextual characteristics (by the 15 proximity area of the region) Contextual characteristics shown to be associated with CMDs in primary care11 12:Density of psychiatrists, psychologists and GPs. Social deprivation (loneliness, single parenthood, widowhood/divorce) and material deprivation (unemployment, income, level of not graduated).47 48 Geographical area: 15 proximity areas defined by the regional health agency of the Nord-Pas-de-Calais region. Statistical analyses Some of the covariates were recoded to use fewer categories. For family status, participants living alone or living with parents were grouped into one category. For family income, participants were grouped in two categories: (€0–€3000) (which corresponds to approximately two times the minimum wage in France) and >€3000. For educational level, we created two categories: less than a high school degree (no degree, degree below high school) or a degree higher or equivalent to a high school degree. For age, our continuous variable was studied in three categories based on the distribution 18–35, 36–50, 51–65. Associations between sociodemographic characteristics and GAD, MDD and alcohol abuse were studied using the χ2 test. Covariates associated with the outcomes with p<0.2 were included in the multivariate analysis. Work-related factors were regrouped according to six previously suggested dimensions transformed each into a Z-score to be comparable to each other.23 A correlation matrix of different work characteristics was computed and presented in a supplementary file (online supplementary appendix 2). Each dimension was dichotomised based on the third quartile or studied as continuous variable in the multivariable models. At first, statistical analyses were conducted separately for each outcome, but factors associated with MDD and GAD were very similar, therefore, to gain statistical power we merged these two disorders into one outcome. To study the association between occupational factors and GAD/MDD and alcohol, we used multilevel Poisson regression models using a robust error variance procedure (sandwich estimation)49 with patient as level 1 and geographical area as level 2. Given the high prevalence of these problems, Poisson regression was preferred to logistic regression to avoid the overestimation of risk ratios.50 GAD/MDD or alcohol abuse were the dependent variables and the six dimensions of work-related factors were the exposure variables. Statistical models were adjusted for each exposure variable and for other covariates that were associated with GAD/MDD (previous mental health problems/disorders, alcohol abuse, material deprivation and GP’s gender) or alcohol abuse (family status, company size, previous mental health problems/disorders, job instability, education level, past unemployment, GAD and MDD) (p<0.05) in a multivariable Poisson regression model excluding occupational factors. Age, gender and occupational grade were included directly in the adjustment variable. Absolute risks among persons who were exposed were computed for each of the studied work dimensions. 10.1136/bmjopen-2017-020770.supp2Supplementary file 2 All analyses were performed using GNU R software V.3.1.1. (lme4 package).51 52 Results Participation and description of the population Of the 1000 GPs contacted by mail, 185 accepted to participate (response rate=18.5%) and 121 completed the study (figure 1). Participating GPs were more likely to be male (sex ratio=1.82), and to be 50 years or older; they were disseminated throughout the Nord-Pas-de-Calais region (table 1). Participating GPs were representative of those practising in the region in terms of geography, age, type and years of practice. Figure 1 Flow chart of participation in the Héraclès study, France, 2014. GPs, general practitioners. Table 1 Description of the study population, Héraclès study, France, 2014 N % Work characteristics Work intensity  High 437 21.6  Low 1588 78.3 Emotional demands  High 476 23.5  Low 1549 76.4 Autonomy  High 598 29.5  Low 1427 70.4 Conflict in values  High 685 33.8  Low 1340 66.1 Social relations at work   High 688 33.9   Low 1337 66 Job insecurity   High 565 27.9   Low 1460 72 Covariates Patient characteristics Gender   Male 46.4   Female 53.6 Age group   (18–35) 597 29.5   (36–50) 872 43.1   (51–65) 552 27.3 Occupational grade   Blue collar 273 13.9   Pink collar 1185 60.1   White collar 513 26 Educational level