Search Strategies, Inclusion and Exclusion Criteria

A systematic investigation of the literature was conducted to search for english, portuguese and spanish language articles on the online databases Scopus and ScienceDirect. These databases were chosen because together they index thousands of peer-reviewed titles and journals in the scientific fields of medical and social sciences. The search was conducted for two independent researchers in July 2021 (no discordances identified) with no date range. The search terms were used in each database search in the following format: "emotional intelligence" AND burnout; "emotional intelligence" AND "emotional exhaustion"; "emotional intelligence" AND depersonalization; "emotional intelligence" AND "personal accomplishment". The advanced search was performed in both databases. Studies were required to meet the following criteria: (1) include at least one measure of emotional intelligence and one of burnout; (2) include performance-based or self-report-based instruments, based on the original Salovey and Mayer's conception of EI and occur in more than two other studies; and (3) provide clear correlation coefficients or means and standard deviations statistical data between emotional intelligence and burnout.

The exclusion criteria were: (1) articles in press or not published in journals; (2) not published in english, portuguese, or spanish; (3) not directly related to the theme "emotional intelligence and burnout"; (4) studies that failed to fulfil the criterion of statistical clarity (specifically number of participants and correlations between EI and burnout or some specific facet of burnout); and (5) review papers, meta-analyses and case reports. This first search yielded a total of 1.393 texts. These documents were filtered according to the inclusion criteria mentioned, initially from the titles and abstracts and after from the full read of the papers. When the full texts of the articles were not identified, they were requested from their corresponding authors, with a period of 15 days waiting for the response.

Coding Procedure and Statistical Analysis

All included studies with all its correlations were coded separately in Excel spreadsheets, and no additional data beyond those published in the papers were requested from any author. Two continuous variables were considered for the calculus of the effect size, namely: emotional intelligence and burnout. In the longitudinal research cases (longitudinal cohort study) we decided to select only the values of baseline (e.g., Carvalho et al., 2018). The correlations collected were transformed into Fisher Z-values, ensuring that the variance of the effect size was based on the sample size. According to ^{Cohen (2013)}, it was considered 0.10, 0.30 and 0.50 respectively small, medium and large effect sizes. Effects lower than 0.10 were considered insignificant. The statistical analysis package Jamovi was used to carry out the analyses (^{The Jamovi Project, 2019}). To examine the variability of the sample, the parameter used was the I² test, which is low if it reaches 25%, moderate when it reaches 50%, and it is high if it exceeds 75% (^{Higgins et al., 2003}). For the analysis of the risk of bias, Egger regression was considered, which when significant (p ≤ 0.05) indicates low risk and when not significant means high risk (^{Egger et al., 1997}).

Relationships Between EI and Facets of Burnout

Regarding the facets of burnout, it was identified that emotional exhaustion has more relationships with EI, followed by depersonalization and personal accomplishment. Thus, work interventions aimed at improving EI are likely to affect emotional exhaustion more positively. The effects of EI as a work resource can buffer emotional exhaustion through emotional self-assessment and emotion regulation, for example. In the same logic, the depersonalization component can be amortized by emotion regulation, but also by the more adaptive use of emotions. The increase in EI can also provide the worker with the ability to deal with feelings of frustration that make up low personal accomplishment.

Future research can investigate cause and effect relations between EI and burnout, seeking, among other questions, to analyse whether individuals with greater EI are less prone to burnout. Furthermore, this meta-analysis identified that most studies in the literature present cross-sectional designs (95%), with very few interventional and longitudinal studies. Future research can also investigate the impacts of interventions for EI on burnout levels, performing longitudinal and even cost-benefit comparisons between types of intervention or even between the absence and presence of interventions with this objective in the organizational environment.

The Assessment Tools Used

In the analysis of the assessment tools used, it was possible to identify that EI as an ability was evaluated in only two studies through the MSCEIT. The other studies evaluated EI as a trait, with emphasis on WLEIS and TEIQue-SF instruments. There are some possible explanations for the low quantity of studies that use the model of EI as ability and for the high amount of studies that analyse EI using the trait model. It is common for the instruments that consider EI as ability, which are usually performance measures, not to foresee strongly previously predicted results by theories (^{Miao et al., 2017}). In addition, there is a shortage of instruments that evaluate EI as ability, mainly for the difficulty of creating tools capable to accurately measure the construct (^{O'Connor et al., 2019}). The most used instrument to measure EI as ability, the MSCEIT, is traded at high prices. This may be a factor that makes its use on a large scale in research unfeasible. This highlights the need to develop performance instruments, which are shorter, cheaper and with good psychometric properties to measure EI as ability.

With regard to the trait EI measures, which are usually self-report instruments, are useful for measuring emotional self-efficacy and behavioural arrangements. EI as trait is related to a large set of emotional, social, behavioural and personality variables (^{Petrides et al., 2016}; ^{O'Connor et al., 2017}). The large scope of empirical evidence with instruments that evaluate EI as a trait is one of the reasons that can explain the extensive use in the studies of this meta-analysis. In addition, trait measures are usually indicated for research that analyzes situations involving continuous stress in educational and employment contexts (^{O'Connor et al., 2019}), which characterizes a large part of the articles analysed. About the different models of EI and their evaluation, more research can be done with the ability model, perhaps evaluating separately the elements of Mayer and Salovey's model (^{Mayer, et al., 2016}). Future research can also compare the relationships between the different models of EI and burnout in order to search for the theoretical model that predicts better results in burnout. Further, future studies may examine which facets of EI exert more influence on the syndrome.

As regards burnout's assessment, it has been identified that the Maslach burnout inventories, with particular emphasis on MBI-HSS, continue to be the most widely used instruments for this purpose, as previously reported (^{Shoman et al., 2021}). A plausible explanation for the vast use of Maslach inventories is that the theoretical job burnout model that underpins the instruments, Maslach and Jackson's model (^{Maslach & Jackson, 1981}) is the most widely accepted and propagated in the literature. According to ^{Demerouti et al. (2021)}, a common failure in burnout studies is the low methodological variance and scarcity of studies that seek to analyse the contextual background of the syndrome. Future research in this context can use complementary tools to the burnout evaluation instruments, such as environmental and organizational climate assessment measures, which has already proved to be crucial variables for the development of burnout at work (^{Aronsson et al., 2017}). It is also possible to use instruments that try to evaluate the antecedents together with the symptoms of the syndrome, as was done by ^{Benevides-Pereira et al. (2017)} in his Burnout Syndrome Inventory.

Burnout in Different Occupations/Working Groups

In subgroup analysis, separated by working groups, it was identified that teachers, doctors and nurses were the most present professional categories in the studies. These three working groups were separately analysed, and it was identified that the effect sizes for the doctor's group were larger than the other two groups, but both had significant and moderate overall effect size for the relations between EI and burnout.

For doctors and nurses, studies such as ^{Holliday et al. (2017)} and ^{Vlachou et al. (2016)} presented results similar to that of this meta-analysis. The authors emphasize that these groups of workers experience stressful situations on their daily life associated with negative emotions linked to death and illness, in addition to having a low sense of emotional and social well-being, which can lead to symptoms of burnout.

However, in a meta-analysis analysing the relations between EI and burnout in teachers (^{Mérida-López & Extremera, 2017}), the findings differ somewhat, with small effect sizes for primary school teachers and effect sizes varying from small to large in middle school teachers. The professional context of teachers involves constant emotional exchange with students and their families, high demand and expectations (^{Ghanizadeh & Jahedizadeh, 2015}). There is other evidence in the literature that shows that secondary/middle school teachers have higher levels of burnout than primary school teachers (^{Ribeiro et al., 2020}). This is probably due to the greater number of students in the classroom, the amount of content to be taught to students and the greater demand from parents and guardians for these students (^{García-Carmona et al., 2019}; ^{Ribeiro et al., 2020}). In general, future studies can evaluate the context of each working group separately, analysing background factors and burnout maintainers in each profession.

Burnout at Different Ages

Regarding the age of the subjects, it was identified that the group of individuals aged 43 years or over had a greater effect on the relationship between EI and burnout. Meanwhile, younger individuals, under 20 years of age, had smaller effect sizes. These findings may indicate, for example, that older individuals may have lower EI capacity to deal with burnout. Corroborating this, studies have shown that: individuals over 40 years of age are more vulnerable to burnout, due to wear and tear due to working hours and other reasons (^{Marchand et al., 2018}; ^{Schadenhofer et al., 2018}), as opposed to younger; and that emotional intelligence increases throughout life and peaks in the intermediate age groups, between 25 and 40 years, and starts to decline thereafter (^{Sharma, 2017}). Future research can analyze the preventive effects of interventions for emotional intelligence in younger and older workers and track the burnout and EI levels of these individuals post-intervention over time. Future research can also compare burnout and EI levels in different age groups and verify the level of influence of EI in the prevention, occurrence and intensity of the syndrome over time.

Limitations and Recommendations for Future Research

Despite including a considerable number of participants, of both sexes, different professions and nationalities, this study has some limitations. First, although we have searched for literature in relevant databases without time limitation, the research undertaken was restricted to the databases used and the languages included. In addition, 37 studies had to be excluded due to the lack of clear presentation of the statistical data needed to calculate the effect size, so their results were not included in this study. Furthermore, in this meta-analysis only the relationships between IE and job burnout were analysed, not including other burnout concepts, such as the clinical and the family burnout, not related to work (^{Mikolajczak et al., 2019}).