The construct of burnout syndrome appeared for the first time around the early 1970s, aimed at explaining the process of physical and mental deterioration in professionals working in areas such as teaching, health care, social work or emergency legal services (Freudenberger, 1974). Subsequently, burnout syndrome was defined as a sustained response to chronic work stress comprising three dimensions: the experience of being emotionally exhausted (emotional exhaustion), negative attitudes and feelings toward the recipients of the service (depersonalization), and feelings of low accomplishment and professional failure (lack of personal accomplishment). This definition was proposed on the basis of studies carried out with what is possibly the most commonly used self-report instrument in research on this phenomenon, the Maslach Burnout Inventory (MBI; Maslach & Jackson, 1981, 1986).

The sequential order of appearance and development of the three dimensions of burnout syndrome, and its etiology as a response to occupational stress in human service professionals, have been considered important questions for designing adequate preventive interventions, and various explanatory models have been developed. Gil-Monte and Peiró (1997, 1999) made a detailed examination of several models from these perspectives, proposing, in turn, new interpretative models. As regards sequential development, these authors have suggested that burnout syndrome begins in combination with feelings of low personal accomplishment and emotional exhaustion, considering attitudes of depersonalization as a coping strategy. From a consideration of the causes of burnout at work, they have postulated a model integrating personal, interpersonal and organizational variables, conceptualizing burnout as "...a response to perceived work stress that emerges after a process of cognitive reevaluation, when the coping strategies used by professionals are not efficient for reducing this perceived work stress" (Gil-Monte & Peiró, 1997, pp.44). Thus, coping thus takes on great relevance for preventive intervention, in the sense that adequate coping strategies can be of great help in avoiding the appearance and development of burnout syndrome at work.

From the transactional approach to stress, Lazarus and Folkman (1986) have defined coping as "those changing cognitive and behavioural efforts developed for managing the specific external and/or internal demands judged as exceeding or surpassing the individual’s own resources" (p. 164). Coping strategies have customarily been classified as specific methods, or according to the precise objectives towards which they are directed. Billings and Moos (1981) for example, identified three methods of coping: a) active-cognitive, understood as the management of assessing potentially stressful events; b) active-behavioural, as the observable efforts aimed at managing a stressful situation; and c) avoidance, as refusal to face a problematic or stressful situation. On the other hand, and in accordance with the objectives of coping, authors have made an essential distinction between coping oriented to the problem and coping oriented to the emotion (Lazarus & Folkman, 1986; Edwards, 1988; Begley, 1998). Coping oriented to the problem would represent an attempt to respond directly to the stressful situation; coping oriented to the emotion would consist in attempts to moderate the emotional response to stressful events.

Various studies have related coping strategies with burnout and other consequences of occupational stress in professionals working in the caring professions and other human services. Thornton (1992), for example, found a statistically significant association between coping of the avoidance type with burnout in a sample of workers at a psychiatric clinic. In the longitudinal study by Koeske (1993), carried out with social workers, it was found that coping strategies oriented to control provided greater capacity for coping with difficult situations at work. Chan and Hui (1995) found that coping strategies addressed to avoidance were positively related to the three components of burnout in a group of secondary school teachers.

Similarly, and with a population from the same professional context, Yela (1996) reported that the greater the feelings of emotional exhaustion, the more likely these professionals were to use strategies coinciding with a passive form of coping, including strategies based on behavioural and mental disconnection from the situation, concentration on one’s emotions and venting one’s feelings when faced with difficult or stressful events. Gil-Monte, Peiró and Valcárcel (1995) reported that coping strategies of avoidance increased emotional exhaustion, while "control" coping maintained personal accomplishment at work in a group of nursing professionals. Analogous results were reported by Hart, Wearing and Headey (1995) with a group of police officers. Finally, and in the exhaustive review on occupational stress in special education teachers by Wisniewski and Gargiulo (1997), it was shown that although stress cannot be avoided in this environment in absolute terms, it is potentially manageable, through systems of social support in the workplace by administrative management, superiors and colleagues.

Although it seems evident that strategies oriented to the problem are much more effective for coping with stressful situations than those oriented to the emotion and to avoidance (Roger, Jarvis & Najarian, 1993; Hart et al., 1995), there is evidence that the effectiveness of strategies oriented to the problem would depend on effective control of the potential stressors of the environment and individual emotions (Folkman, 1984; Dewe, 1987; Edwards, 1988; Peiró & Salvador, 1993; Labrador, 1995; Long, 1998; Peñacoba, Díaz, Goiri &Vega, 2000; Ito & Brotheridge, 2001). On the other hand, persistent use of strategies oriented to the problem when there are few possibilities of controlling and/or changing the stressors in the environment may greatly exacerbate the undesirable effects of work stress (Schaubroek & Merritt, 1997; de Rijk, Le Blanc, Schaufeli & de Jonge, 1998). At the same time, it has been pointed out that in less controllable circumstances, strategies oriented to the problem in combination with strategies oriented to avoidance may be useful for improving adaptation and wellbeing, so that flexible coping would be adaptive rather than maladaptive, that is, coping oriented to the problem would be adaptive in controllable situations, whilst coping oriented to avoidance would be adaptive in situations difficult to control (Latack, 1986; Koeske, 1993). In this line, Cheng (2001) carried out a study on flexible coping, concluding that both perception of control and objective controllability of the stressors would play a key role in the achievement of effective coping.

According to the cybernetic theory of stress, coping and wellbeing in organizations postulated by Edwards (1992), stressful stimuli, understood as the discrepancies between a perceived state and a desired state, and which at the same time are considered important by the individual, can activate coping in two different ways, directly or indirectly. Stress would activate coping directly, anticipating potential threats to psychological wellbeing, whilst coping would be activated indirectly when psychological wellbeing had already been damaged in some way. In this line, it seems plausible to assume that different coping strategies will precede the process of burnout syndrome, whilst others will be found after it. Various authors have examined the development of burnout, considering different relationships between the dimensions of the syndrome evaluated by means of the MBI. For example, Golembiewski, Munzenrider and Stevenson (1986) proposed that burnout starts with feelings of depersonalization, which leads to low personal accomplishment and, subsequently, to emotional exhaustion, whilst Leiter and Maslach (1988) suggested that emotional exhaustion develops first, giving rise to attitudes of depersonalization and, thirdly, to low personal accomplishment in the workplace. In any case, the longitudinal study by Lee and Ashforth (1993), which compared the two models, provided greater support for the Leiter and Maslach (1988) model. On the other hand, the study by Gil-Monte et al. (1995) proposes a model in which work overload and self-confidence, as well as coping strategies oriented to control and avoidance, are specified as causal antecedents of emotional exhaustion and low personal accomplishment at work, whilst depersonalization, intention to leave one’s job and the propensity for absenteeism are proposed as consequences of emotional exhaustion and low personal accomplishment. According to these authors, depersonalization constitutes a coping strategy that emerges subsequent to the feelings of emotional exhaustion and low personal accomplishment.

In accordance with these premises, this study examines an analogous model in which coping strategies based on control and lack of control over work stressors and individual emotions influence personal accomplishment and emotional exhaustion, though it incorporates a reciprocal relationship between these two dimensions of the syndrome. Considering this system of relationships between different patterns of coping, it is hypothesized that if personal accomplishment and emotional exhaustion appear jointly, personal accomplishment will have greater impact on emotional exhaustion.

The study was carried out with social educators in various intervention contexts. A total of 127 people, 86 women and 41 men (mean age 31.28 years; SD=7.63), completed self-report measures on coping strategies and burnout. At the time of data-collection, the participants had 5.9 years of professional experience in their jobs (SD=5.80), carrying out their work in public and private centres in an urban environment, and providing socio-educational services to the disabled, children and adolescents at social risk, and the elderly.

Burnout syndrome- We used the Spanish adaptation of the Maslach Burnout Inventory (MBI, Maslach & Jackson, 1986; Maslach & Jackson, 1997). This questionnaire includes 22 items to be responded to on a 7-point Likert-type frequency scale: 0, never; 1, a few times a year or less; 2, once a month or less; 3, a few times a month; 4, once a week; 5, a few times a week; 6, every day. The items were distributed in three dimensions: emotional exhaustion (EE; feelings of being emotionally exhausted), depersonalization (DP; feelings of impersonal response towards recipients of the service, care, treatment or instruction), and personal accomplishment (PA; feelings of competence and success in working with people). Construct validity of the MBI has been assessed in different sociocultural contexts with quite acceptable results (Schutte, Topinnen, Kalimo & Schaufeli, 2000). with regard to its reliability, coefficients of internal consistency of the three scales varied between .82 and .90 for EE, .48 and .79 for DP, and .57 and .71 for PA (Maslach & Jackson, 1986; Aluja, 1997; Gil-Monte & Peiró, 1997). In the present study, alpha coefficients of internal consistency were .81, .80 and .56 for EE, PA and PA, respectively.

Coping strategies- We used a Spanish translation of the PsychNurse Methods of Coping Scale (PNMCS; McElfatrick, Carson, Annett, Cooper, Holloway & Kuipers, 2000).

This instrument has 35 items, with a response format on a Likert-type frequency scale with 5 points: 1, never; 2, rarely; 3, occasionally; 4, often; and 5, always. Five types of coping strategies are measured: diverting attention from work, self-regulation and self-attitude, social support, positive attitude, and emotional comfort. The authors report an acceptable convergent validity of this instrument, in reference to the coping scale of the Occupational Stress Indicator (Cooper, Sloan & Williams, 1988), with alpha reliability coefficients of between .67 and .78. This questionnaire was initially designed for use with mental nurses, and it was chosen taking into account that the constructs it defines are appropriate for the aim of this study. In the translation of this instrument into Spanish, in item 8 the term "educational team" was introduced in place of "team", whilst in item 9 the term "users" was employed instead of "others," in an attempt to reflect as far as possible the professional context of social educators. None of the other changes arising from the translation from Spanish to English substantially affected the content of the instrument’s items.

Various centres were considered in collecting data for this study in a non-random way, taking into account the criterion that their staff included professionals in social education, since these carry out their work in a variety of intervention areas. Each centre received a personal visit, in which the directors were told about the general objectives of the study and permission was requested for the questionnaires to be filled in voluntarily and anonymously by the social educators employed there. Approximately 80% of the centres visited agreed to participate in the study.

Our study hypothesized a system of relations between different coping methods and the dimensions of burnout syndrome, so that a suitable way of analyzing it is by means of the technique of structural equations modelling. The main advantages of this procedure are that it provides a direct method for dealing simultaneously with multiple relations between variables, it is capable of assessing the relationships exhaustively, and it is statistically efficient (Bollen, 1989; MacCallum & Austin, 2000). The strategy followed is a comparative one of nested models, through which it is attempted to assess the best of a set of models, allowing a more accurate assessment and with a wider perspective of the model that best fits the data observed (Mulaik, James, Van Alstine, Bennett, Lind & Stilwell, 1989; Bentler, 1990). Given that a reciprocal relationship is established between personal accomplishment (PA) and emotional exhaustion (EE), a nonrecursive model is stipulated, in which one variable is defined as the cause of another, and vice-versa. The suitability of nonrecursive models in cross-sectional studies has been considered theoretically by Edwards (1992), whilst Wong and Law (1999) have studied their aptness for cross-sectional data from an applied point of view. In sum, the bi-directional relationship between personal accomplishment (PA) and emotional exhaustion (EE), as a function of different methods of coping, is assessed by comparing four possible alternatives: a) no relationship between PA and EE; b) EE as cause of PA, c) PA as cause of EE, and d) a reciprocal relationship between PA and EE. More specifically, we propose a model consisting of two endogenous latent variables, Personal accomplishment (PA) and Emotional exhaustion (EE), each of which is measured by means of four indicators, Pa1-Pa4 and Ee1-Ee4, respectively. These indicators were obtained from the grouping into parcels of the items of the respective MBI subscales, by means of the correlations method (for an exhaustive review on methods of grouping items into parcels, see Landis, Beal & Tesluk, 2000). The variable Depersonalization (DP) is also specified as latent measured by means of the five items making up this dimension. Additionally, four observed variables are specified, corresponding to the different coping strategies: control (CO) and social support (SS), as coping strategies based on control, and emotional support (ES) and avoidance (AV), based on lack of control. The coping strategies are considered exogenous variables or instrumental variables, insofar as the model is nonrecursive (Wong & Law, 1999). Coping strategies that imply control (CO and SS) are established as determinant of PA, and coping strategies implying lack of control (ES and AV) as determinant of EE. Correlations are specified between all the coping strategies. Between PA and EE, we define a bi-directional relationship and the consequent correlation between their respective error terms, since there may be factors not included in the model that affect PA and EE similarly when there is a reciprocal relationship between the two (Anderson & Williams, 1992; Loehlin, 1992).

A principal components analysis with direct oblique rotation and criterion of extraction of equivalent values >1 yielded a five-factor structure. Fabrigar, Wegener, MacCallum and Strahan (1999) have suggested that oblique rotations provide more information than orthogonal rotations, as well as more accurate and realistic factorial solutions; it is for these reasons that this rotation method was chosen for the present study. Six items obtained secondary loadings >.30 in other factors, indicating an overlap between factors and making their interpretation difficult, so that the analysis was repeated using the same procedure but eliminating the six items with secondary loadings. In this secondary analysis we obtained a four-factor solution, which explained 46.3% of the variance: control (CO, 24.3%), emotional support (ES, 8.5%), social support (SS, 7.1%) and avoidance (AV, 6.4%). In the subsequent analyses we used these four dimensions, considering that, according to the content of the items making up each factor, control (CO) and social support (SS) imply greater capacity for control, while emotional support (ES) and avoidance (AV) imply less capacity for control over stressors in the environment and the emotions. The alpha coefficients of internal consistency of each one of these dimensions are .85 and .74 for CO and SS, and .58 and .77 for ES and AV, respectively.

Table 1 shows descriptive statistics, values of normality in skewness and kurtosis, alpha coefficients of internal consistency and correlation coefficients for the subscales of burnout syndrome and coping. The burnout syndrome scales meet normality assumptions, except for DP. The correlations are significant and with negative sign for EE and DP with PA (-.37 y -.24), and with positive sign for EE with DP (.36). The four coping dimensions meet normality assumptions. The coefficients of internal consistency are acceptable, except for the ES scale, with an alpha of .58, though this is also the scale with the fewest items. As far as ES is concerned, it shows a statistically significant correlation of .25 with AV, and smaller, non-significant correlations with CO (.14) and SS (-.01), which may indicate a degree of independence of the two coping methods, control and lack of control. Nevertheless, AV also shows significant correlations of .53 and .30 with CO and SS, respectively, which indicates a flexible type of coping. All the coping strategies obtain significant correlations for EE, with values of -.32 with CO, -.29 with SS, .26 with ES, and -.19 with AV. With PA, the correlation coefficients are slightly higher: .53 with CO, .30 with SS and .33 with AV, though it is practically non-existent with ES (.03). The correlations are not significant between the coping strategies and DP. It is noteworthy that the only coping strategy that shows a positive correlation with EE is ES, whilst the sign of the correlation between AV and EE is negative, given that AV seems to imply, in principle, lack of control, since it is the only strategy that correlates positively and significantly with ES (.25).

Assessment of the model was carried out in a sequence of four nested models with the aim of identifying the alternative that provides the best fit to the data, so that the two parameters in the bi-directional relationship between PA and EE are fixed at zero according to the possible combinations: a) both parameters in model 1 (PA EE); b) the relationship of PA to EE in model 2 (PA EE); c) the relationship of EE to PA in model 3 (PA EE); and finally, d) no parameter in model 4 (PA EE). The null model is the baseline that defines the simplest set of relationships theoretically justifiable, in this case, the absence of any relationship between the variables included in the model. The model is identified correctly when fixing at unity the regression coefficients of all the error terms, and of Pa1, Ee1 and Dp5 at PA, EE and DP, respectively (see Figure 1). The matrix of variances-covariances among the variables observed is analyzed as raw data, and is available from the authors on request. The Amos 3.6 program (Arbuckle, 1997) was used, in addition to the method of estimation of maximum likelihood.

Table 2 shows the results of each model obtained in different goodness of fit indices. Significant values of c2 (p< .05) indicate that the proposed model is significantly different from the observed data, thus determining a poor fit of the model in question. Values of >.90 for the Goodness of Fit Index (GFI), the Non-Normed Fit Index (NNFI), and the Comparative Fit Index (CFI), and values <.05 for the Root Mean Square Error of Approximation (RMSEA), suggest an acceptable fit (Browne & Cudeck, 1993). According to these criteria, models 1 (PA EE) and 2 (PA EE) show a poor fit, whilst models 3 (PA Æ EE) and 4 (PA ´ EE) are those that offer the best fit to the data: c2 (109) =131.98, p= .066, c2 (108)= 130.80, p=.067, respectively, with adequate and equivalent values of GFI= .89, NNFI= .94, CFI= .96 and RMSEA= .04. Table 2 also shows the differences between the c2 values of the four models; when the difference is significant the model with the fewest degrees of freedom is preferred. As it can be seen, there are significant differences between the c2 values for models 1 and 3, Dc2= 12.58 (1 df), p<.001, models 1 and 4, Dc2= 13.76 (2 df), p< .001, and models 2 and 4, Dc2= 11.28 (1 df), p< .001, whilst the differences in c2 values are not significant between models 1 and 2, Dc2= 2.48 (1 df), p<.10, and models 3 and 4, Dc2= 1.18 (1 df), p< .30. In addition, Table 3 shows the regression coefficients with associated standard errors, as well as the multiple correlation coefficients for the latent variables obtained in the four models. Figure 1 shows model 4 with the corresponding regression coefficients. The coping strategies with a significant impact on PA and EE are CO (.45, p< .01) and ES (.30, p< .01), respectively. The regression coefficients in the reciprocal relationship between PA and EE are of negative sign, and of considerable and significant size from PA to EE (-.58), in comparison with the value from EE to PA (-.27), while the multiple correlation coefficients for PA and EE in model 4 are the highest in comparison with the other three models (.48 and .37, respectively). In sum, and in accordance with the theory and hypothesis proposed, models 3 and 4 are those that best represent the relationships between the variables analyzed in this study.

This study has analyzed a nonrecursive model through which there is established a reciprocal relationship between two central components of burnout, personal accomplishment and emotional exhaustion, in situations involving coping patterns determined by the degree of control over work stress and individual emotions. The use of nonrecursive models in the field of organizations in cross-sectional research has been considered by Edwards (1992) and Wong and Law (1999) from theoretical and applied perspectives, respectively. In the examination of such dynamic and multifactorial processes as those involved in the development of occupational stress, nonrecursive models can provide an adequate analysis of the relationships between the variables studied in a given context. In the case of the dimensions of burnout assessed by means of the MBI, establishing a temporal priority may depend on a variety of factors that are difficult to control in both cross-sectional and longitudinal research. The results obtained are consistent with our hypothesis, in the sense that in the assumed simultaneous relationship between feelings of personal accomplishment and emotional exhaustion in the workplace mediated by different coping patterns, personal accomplishment would tend to reduce emotional exhaustion, whilst emotional exhaustion would have a non-significant impact on personal accomplishment. These results are also in line with those obtained in the study by Gil-Monte et al. (1995), and with later approaches to the process and sequential order of development of burnout dimensions assessed with the MBI (Gil-Monte & Peiró, 1997, 1999), in which it is considered that emotional exhaustion and low personal accomplishment emerge when coping strategies are inadequate, giving rise to feelings of depersonalization. Gil-Monte et al. (1995), stressed the need to assess the fit of their model with different occupational samples and with different groups of variables. Although no other variables relevant to the burnout process have been incorporated, except for different coping methods, the present study provides empirical support for the model proposed by these authors with a sample of social educators workers through the assessment of latent variables. Furthermore, we consider this perspective to be in line with the idea of direct and indirect coping postulated by Edwards (1992); that is, the coping strategies specified as antecedents of personal accomplishment and emotional exhaustion could be considered as direct coping methods, since they would be attempting to anticipate potential threats to psychological wellbeing, whilst depersonalization would constitute, in our view, an indirect coping method, since psychological wellbeing would already have been damaged in some way as a result of the feelings of emotional exhaustion and low personal accomplishment that precede it.

The correlations observed for the strategy of avoidance with emotional support (lack of controllability), and at the same time with control and social support (controllability), appear to point in the direction of a type of flexible coping understood as the fit between the nature of the coping and the perceived and objective controllability of the stressful situation (Cheng, 2001; Zellars & Perrewé, 2001). In the social services context, in which we find numerous situations with little or null capacity for control on the part of professionals, the combined use of coping strategies oriented to the problem and to avoidance may help to achieve greater adaptation and psychological wellbeing. It is likely that people working in social education contexts and other human services utilize coping strategies oriented to avoidance as they lack sufficient possibilities for control in their workplace. The use of coping strategies oriented to control, or in terms of proactive personality, as proposed by Parker and Sprigg (1999), depends to a large extent on the opportunities offered by the work context, and on certain personality characteristics. Given low perceived and objective controllability, it is preferable to use strategies oriented to avoidance, since the elements necessary for modifying certain situations are unavailable, and, as has been pointed out in other studies (Schaubroek & Merritt, 1997; de Rijk et al, 1998), coping strategies oriented to the problem in situations of low control are counterproductive in that they produce more stress. In any case, and according to our results, coping strategies based on control are preferable to those based on lack of control, since the impact of personal accomplishment on emotional exhaustion is greater than that of emotional exhaustion on personal accomplishment when both types of coping act on each dimension of burnout.

The importance of effective control by workers on the resources and abilities necessary for working in the way they feel is most convenient and comfortable has been emphasized from various relevant models in the study of stress at work, as an essential element for reducing the undesirable effects of occupational stress (Edwards, 1988; Karasek & Theorell, 1990; Maslach, Schaufeli & Leiter, 2001). To some extent, our results corroborate this point of view, given that the control coping strategy (CO), has a notable impact on personal accomplishment, which is seen, in turn, to be an important element in reducing emotional exhaustion. On the other hand, the strategy of emotional support (ES), which presumably implies lack of control, has a positive and significant impact on emotional exhaustion, even though emotional exhaustion does not have a relevant effect on personal accomplishment (though it does have a considerable effect on depersonalisation).

From an approach oriented to the prevention of burnout and work stress, Maslach et al, (2001) propose an interesting change with regard to the conceptualization of the burnout syndrome as a negative psychological state: devoting more attention to the study of the potentialities and optimum functioning of human beings than to deficits and pathologies, which has traditionally constituted the predominant approach in psychology. In our opinion, the results obtained in this study also support this idea, since the impact of personal accomplishment at work on emotional exhaustion is much greater than that of emotional exhaustion on personal accomplishment. In sum, rather than trying to avoid emotional exhaustion from a merely palliative approach, it seems much more preferable and profitable to attempt to provide the individual and organizational elements for achieving greater personal accomplishment at work – such as autonomy and power of decision for utilizing individual resources and skills.

We should like to thank an anonymous reviewer for the suggestions made about this article. The work was produced with the support of the Departament d’Universitats, Recerca i Societat de la Informació of the Generalitat de Catalunya.

References