In all the 25 studies exposure was based on self-report in standardized and validated questionnaires. For job demand and control the most commonly applied instruments were the Job Content Questionnaire (JCQ)  and Copenhagen Psychosocial Questionnaire . About ten different scales were used for the exposure assessments.
With regard to burnout, a handful of well validated instruments were used in the included articles. In 18 of the 25 studies, MBI, MBI-General Survey [4, 5] or the Dutch adaption of MBI, the Utrecht Burnout Questionnaire  were used. The other screening scales were the CBI (Copenhagen Burnout Inventory) , Oldenburg Burnout Questionnaire , Shirom-Melamed Burnout Questionnaire  and the Proactive Coping Inventory . CBI was used in three studies and the other instruments in only one study each. In general there are high correlations between these measures. These were reasons for non-differential use of the measures.
Work environment factors
Table 1 shows the results of the evidence grading process. The overall picture is that most of the studies were focused on emotional exhaustion and somewhat fewer on the dimension of depersonalisation. For (reduced) personal accomplishment and for the burnout measure there was often only one relevant study of sufficient quality in relation to a defined work factor. For the work factor low reward in relation to depersonalisation and to reduced personal accomplishment there were two studies of sufficient quality (Maslach & Leiter 2008; van der Ploeg & Kleber 2003), but the GRADE-assessment was “limited scientific evidence” because of inconsistent results in the two studies.
In the following, we present weighted odds ratios for the associations between the work factor and the burnout measure. For several of the exposure measures, such as job control, demands and workplace support, we also present forest plots.
Associations between job control and emotional exhaustion were analysed in a total of 9 studies (De Lange, Taris, Kompier, Houtman & Bongers, 2004; Geuskens, Koppes, van den Bossche & Joling, 2012; Janssen & Nijhuis 2004; Koponen, Laamanen, Simonsen-Rehn, Sundell, Brommels &Suominen, 2010; Langballe, Innstrand, Aasland & Falkum 2011; Le Blanc, Hox, Schaufeli, Taris & Peeters, 2007; Magnusson-Hansson, Theorell, Oxenstierna, Hyde & Westerlund, 2008; Maslach & Leiter 2008; Van der Ploeg & Kleber, 2003). The results are very consistent, eight of the nine point estimates were significantly above 1.0 (Fig. 2, note that some data were re-calculated from correlations to odds-ratios). The weighed odds ratio for these studies was 1.63, (95% CI 1.53 to 1.75). The results supported a moderate scientific evidence (grade 3) for a relationship between low job control and increased emotional exhaustion.
Two studies (Borritz, Bultman, Rugulies, Christensen, Villadsen & Kristensen 2005; Sundin, Soares, Grossi & Macassa, 2011b) used the burnout measure and provided support for limited scientific evidence for a relationship between low job control and increased burnout (grade 2).
Job demands were the most frequently studied occupational exposure in the included studies. Unspecified psychological demands were studied in relation to emotional exhaustion in 13 studies (Bakker, Schaufeli, Sixma, Bosweld & van Dierendonck, 2000: Chrisopoulos, Dollard, Winefield & Dormann, 2010; De Lange et al., 2004; Demerouti, Le Blanc, Bakker, Schaufeli & Hox, 2009; Geuskens et al., 2012; Janssen et al., 2004; Koponen et al., 2010; Le Blanc et al.,2007; Prieto, Salanova, Martinez & Schaufeli, 2008; Magnusson-Hansson et al., 2008; Sundin, Hochwalder & Lisspers, 2011a; Van de Ven, Van den Tooren & Vlerick., 2013; Van der Ploeg et al., 2003). For 11 of them (not Koponen et al., 2010 and Prieto et al., 2008) weighted odds ratios could be calculated (Fig. 3). The results were consistent, ten of eleven point estimates were significantly above 1.0 (weighted OR of 2.53, 95% CI 2.36 to 2.71).
Unspecified psychological demands were also studied in relation to depersonalisation and cynicism in 4 studies (Bakker et al., 2000; Demerouti et al., 2009; Le Blanc et al., 2007; Sundin et al. 2011a). For three of the four studies the point estimate was above 1.0 with a weighted odds ratio of 2.37, 95% CI 1.86 to 3.03).
Emotional demands had been studied in relation to emotional exhaustion in five studies (study Chrisopoulos et al., 2010; Le Blanc et al., 2007; Prieto et al., 2008; Van de Ven et al., 2013; Van der Ploeg et al., 2003). The point estimate was above 1.0 in the four studies where it could be calculated with a weighted odds ratio of 2.95, 95% CI 2.40 to 3.62.
Finally, demands from patients were studied in relation emotional exhaustion in three studies (Bakker et al., 2000; Demerouti et al., 2009; Sundin et al. 2011a). Two of the studies had 95% CI just below 1.0 (0.99-2.43 and 0.97 to 3.34) but the weighted odds ratio ended at 2.02 (1.50 to 2.72).
In summary, for all three aspects of demands (unspecified, emotional and from patients) the results supported limited scientific evidence in relation to emotional exhaustion and depersonalisation (grade 2). For demands and reduced personal accomplishment there were too few studies for any conclusion.
Job strain, i.e. the combination of high psychological demands and low decision latitude, was investigated in only one study (Ahola & Hakanen, 2007). The study indicated a strong association, but the scientific evidence was judged insufficient since the association with burnout was investigated only in one relevant study of sufficient quality.
Different sources of social support have been studied. The strongest evidence (grade 3) was found for low workplace support and emotional exhaustion without specified source (Burke & Greenglass 1995a;, De Lange et al., 2004; Geuskens et al., 2012; Janssen et al., 2004; Koponen et al., 2010; Le Blanc et al., 2007; Magnusson-Hansson et al., 2008; Sundin et al. 2011a; Van der Ploeg et al., 2003). The weighted odds ratio based on seven of these studies was 1.81 (95% CI 1.68 to 1.95) (Fig. 4). There was also evidence (grade 2) for an association between low supervisor support (Geuskens et al., 2012, Magnusson-Hansson et al., 2008; Sundin et al. 2011a; Theorell, Nyberg, Leineweber, Magnusson-Hansson & Westerlund 2012; Van der Ploeg et al., 2003), as well as for low co-worker support (Geuskens et al., 2012; Magnusson-Hansson et al., 2008; Sundin et al., 2011a; Van der Ploeg 2003) on one hand and emotional exhaustion on the other hand.
There were three studies on low workplace support and depersonalization (Burke et al. 1995a; Le Blanc et al. 2007; Sundin et al. 2011a) with a weighted ratio 1.59, (95% CI 1.11 to 2.26) and four studies (Borritz et al. 2005; Burke et al. 1995a,b; Sundin et al., 2011b) on low workplace support and burnout. The scientific evidence was rated limited (grade 2).
The high work load category refers to a mixed group of stressors all of which result in a large volume of work. In several of these studies the author refers to events leading to a high work volume. It differs from demands, which refer to rushed tempo (unspecified) or – and specified – emotional demands and demands from patients. There were 7 studies with focus on emotional exhaustion (Burke et al., 1995b; Demerouti et al., 2009; Langballe et al., 2011; Le Blanc et al., 2007; Prieto et al., 2008; Maslach et al., 2008; Van der Ploeg et al., 2003). A weighted odds ratio could be calculated for 5 of these which resulted in an odds ratio of 4.22 (CI % 3.50 to 5.11). There was also evidence for an association between high work load and the dimension of depersonalisation/cynicism. In six studies the outcome was described as depersonalisation. It was only possible to calculate weighted odds ratio for two of them (Maslach et al., 2008; Van der Ploeg et al., 2003) with an OR of 2.52 CI 95% 1.85 to 3.44). For two of the three studies on high work load and the outcome described as cynicism (Demerouti et al., 2009; Le Blanc et al., 2007) weighted odds ratios were calculated (3.03, CI 95% 3.03 (2.21 to 4.16). The analyses supported limited scientific evidence (grade 2) in relation to the dimension of depersonalisation/cynicism. There were three studies on work load and reduced personal accomplishment. A weighted odds ratio could not be calculated, but based on a narrative analysis the evidence was rated limited.
Low reward had a weighted odds ratio of 1.86 (1.37 to 2.52) in relation to emotional exhaustion (Maslach et al., 2008; Van der Ploeg et al., 2003).
Job insecurity had a weighted odds ratio of 1.39 (1.22 to 1.57) based on all three studies in relation to emotional exhaustion(Geuskens et al., 2012; Koponen et al., 2010; Magnusson-Hansson et al., 2008). The result for low reward and job insecurity supported limited scientific evidence (grade 2).
There was an association between workplace justice and decrease in emotional exhaustion (Koponen et al., 2010; Maslach et al., 2008; Ramarajan, Barsade & Burack, 2008) with an evidence grade of 2 (a weighted odds ratio of 0.35, 95% CI 0.27 to 0.45 based on the 3 studies).
Insufficient evidence (grade 1) was found for several factors, which are listed in the lower part of Table 1. Mostly there are no studies of sufficient quality. One study investigated the association between physical environmental factors (in a school environment, Burke et al., 1995a) and burnout and showed a significant association but one study is not enough for evidence. Finally, there were no relevant studies with sufficient quality of associations between exposures to vibrations, chemical and biological factors on one hand and burnout on the other hand.