Research in centers for people with intellectual and developmental disabilities has somewhat neglected the relationship between workers' burnout and users' service evaluations. Two independent survey studies tested this connection. In the first study (100 centers, 714 workers, and 612 family members), results confirmed that burnout has a negative relationship with workers' perceptions of service quality. In turn, these perceptions are associated with the service quality perceptions of family members and their satisfaction with the service. In a replication sample (86 centers, 601 workers, and 819 family members), we reproduced these results and added situational constraints in the model. Both social and technical constraints correlated positively and significantly with burnout. These studies offer a view of the relationships between burnout and service quality.
One of the important topics in services for people with intellectual and developmental disabilities (IDD) is burnout in workers (e.g., Lunsky, Hastings, Hensel, Arenovich, & Dewa, 2014). Understanding staff burnout in service organizations where there is direct interaction between workers and users is relevant for several reasons. First, workers' well-being is a moral and legal duty of employers (Devereux, Hastings, & Noone, 2009). Second, stress has an influence on a number of worker reactions related to health and absenteeism (Rose, 1995). Finally, burnout has been found to deteriorate staff performance, leading to negative outcomes for users of services (Hickey, 2014).
In the present study, we focus the attention on the connection between burnout and performance in services for people with IDD. Some previous research efforts have observed this relationship. Lawson and O'Brien (1994) found that burnout increases negative client contact behaviors and reduces both client contact frequency and positive client contact behaviors. Rose, Jones, and Fletcher (1998) also reported that the interaction frequency and the positive interactions with service users were higher in low-stress services than in high-stress contexts.
However, little is known about potential models describing the relationship between staff burnout and service quality reported by users in centers for people with IDD. Based on research in other contexts (Heskett, Jones, Loveman, Sasser, & Schlesinger, 1994; Yee, Yeung, & Cheng, 2011), we propose that a service chain can be defined connecting burnout and service quality. This service chain model describes the process that explains the relationship between burnout and family member service evaluations. As Hayes (2012) pointed out, the study of processes is indicative of the maturity of a science because it goes beyond the limited scope of investigating the relationship between two variables, making it possible to understand the mechanisms involved in the relationship. To do so, we carried out two independent survey studies. The first study tests a model where worker burnout in centers for people with IDD is related to the service quality perceived by workers. These worker perceptions, in turn, are associated with service quality perceptions and satisfaction with services reported by family members. In a replication sample, we extend this model by adding social and technical constraints (Figure 1) as situational obstacles in the center environment that are beyond the workers' control and can limit their performance (Dyer & Quine, 1998).
Burnout is traditionally considered “a prolonged response to chronic emotional and interpersonal stressors on the job, and it is defined by the three dimensions of exhaustion, cynicism and professional inefficacy” (Maslach, Schaufeli, & Leiter, 2001, p. 397). Emotional exhaustion refers to the depletion or draining of emotional resources, possibly caused by interpersonal demands. Cynicism reflects an indifferent or distant attitude toward one's work. Finally, lack of professional efficacy encompasses both social and nonsocial aspects of occupational accomplishment. Over time, different perspectives have emerged on the facets of burnout. However, despite the persistent debate about the dimensionality of burnout, there is an increasing consensus among researchers that the two central facets of burnout are exhaustion and cynicism (González-Morales, Peiró, Rodríguez, & Bliese, 2012; Schaufeli & Taris, 2005), whereas lack of professional efficacy is considered a personality trait (González-Romá, Schaufeli, Bakker, & Lloret, 2006). In this vein, previous studies also revealed that, rather than being a dimension of burnout, professional efficacy seems to be a leading cause of it (Salanova, Peiró, & Schaufeli, 2002). Thus, the present study focuses on the two central dimensions of burnout. More specifically, we use a measure that combines exhaustion and cynicism. It is a superordinate concept that reflects work-related experiences characterized by depletion of resources and depersonalization. This strategy allows us to use a parsimonious assessment that concentrates on the central aspects of burnout.
In services for people with IDD, burnout reduces the frequency of the contact with service users and the positive tone of these interactions (see Lawson & O'Brien, 1994; Rose et al., 1998). It is implicitly assumed that burned-out employees have fewer resources with which to reach out to others and adequately perform their tasks. In line with the Conservation of Resources theory (COR; Hobfoll, 1988), burnout describes a situation where the worker loses valuable resources, making it difficult to cope with the task demands. Although there are moderators, the consideration of performance as a consequence of burnout persists in the literature (Maslach et al., 2001, p. 404), with the COR theory playing a prominent role (Demerouti, Bakker, & Leiter, 2014). In centers for people with IDD, service quality has a critical role as an indicator of worker performance. Service quality refers to the excellence/superiority of the service (Parasuraman, Zeithaml, & Berry, 1985), including functional and relational aspects (Peiró, Martínez-Tur, & Ramos, 2005; Potočnik, Tordera, Martínez-Tur, Peiró, & Ramos, 2011). Functional service quality describes the delivery of the core service in an instrumental and efficient way. By contrast, relational service quality refers to emotional bonds in service encounters, with benefits for customers beyond the core service (Gwinner, Gremler, & Bitner, 1998), describing a service interaction characterized by respect, appreciation, and esteem (Semmer et al., 2008). In our chain (Figure 1), burnout is negatively related to workers' perceptions of the service quality they deliver to people with IDD. In other words, burned-out workers feel that they are unable to do their jobs well in their interactions with service users.
Continuing the proposed model, workers' service quality perceptions are transferred to family members' evaluations of the service. Schneider, Salvaggio, and Subirats (2002) indicated that benefits for customers improve when workers do their best work. This effort is then translated into high customer ratings of service quality. In centers for people with IDD, family members often interact with staff (e.g., meetings, participation in center activities), and they are able to assess the service quality that workers deliver.
Finally, service quality is a precursor of customer satisfaction (Martínez-Tur, Peiró, Ramos, & Moliner, 2006). Customer satisfaction has been defined as a subjective evaluation of the outcomes of service encounters, such as the quality of the service provided by the staff. Accordingly, when family members perceive high service quality, they also experience high satisfaction with the selection of the center in question.
As mentioned, in the replication sample, we added social and technical constraints into the proposed model (see Figure 1). Although there are different precursors of burnout in services for people with IDD (e.g., Ko, Lunsky, Hensel, & Dewa, 2012), we concentrate on situational constraints (Devereux et al., 2009; Dyer & Quine, 1998) for two reasons. First, research has consistently shown that situational factors related to the job environment are predominant in predicting burnout (Maslach et al., 2001). Second, because the context helps to guide and interpret the research design and findings (see Johns, 2006), we consider the specific context of our research study—centers for people with IDD located in Spain. For many years, these centers have particularly experienced the effects of the current economic crisis, with negative effects on available resources. Constraints in the workers' daily activities (e.g., lack of resources) are relevant and can be critical antecedents of burnout in this specific organizational and social context.
Situational constraints refer to environmental circumstances beyond the workers' control that limit their performance (Dyer & Quine, 1998). Situational constraints are conceptually different from service quality. Constraints refer to obstacles in the workplace environment, whereas service quality is usually associated with the behavior of service providers in their interaction with users. Research has generally classified constraints into two categories: social and technical (Brown & Mitchell, 1993; Martínez-Tur, Peiró, & Ramos, 2005). Social constraints refer to problems related to interactions with others (e.g., lack of trained staff), whereas technical constraints refer to malfunctions in technology and lack of instrumental resources (e.g., computer malfunctioning). In this study, we hypothesized that both social and technical constraints perceived by workers in their daily work are positively related to burnout (steps for the replication sample in Figure 1). It is reasonable to expect that workers' burnout will be higher in centers where both social and technical constraints exist than in centers where the existence of resources reduces the perception of constraints in the job context.
In summary, the main contribution of the current research study is to propose a model where workers' perceptions about the service quality they offer are positively related to service evaluations by family members of people with IDD. However, this relationship is hindered when workers have high levels of burnout and situational constraints. Some previous research efforts have recognized the relationship between burnout and service quality (Lawson & O'Brien, 1994; Rose et al., 1998), but there is a lack of empirical research based on a systematic model that describes the process. The main objective of the current research study is to define specific relationships connecting workers' burnout with service performance from the perspective of family members, while adding situational constraints as antecedents of workers' burnout.
A second contribution of the current research is that we consider two sources of data (workers and family members). This strategy allows us to consider these two perspectives in the service chain, reducing common source variance problems associated with the use of only one source of data (usually the worker) in empirical studies (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). The consideration of these two perspectives also allows us to check the degree to which the experience of workers (internal performance) is transferred to the reactions of family members, beyond the formal limits of centers for people with IDD.
The third main contribution involves the use of a center-level approach to burnout and service evaluations. Many scholars indicate that burnout emerges as a collective reality beyond the individual because workers share experiences and communicate with each other. This so-called “burnout contagion” (Hakanen, Perhoniemi, & Bakker, 2014; Moliner, Martínez-Tur, Peiró, Ramos, & Cropanzano, 2005) makes it possible to explore the negative influence of burnout in a more realistic way: “If burnout only affected individuals in isolation, it would be far less important and far less devastating in its impact than it is. Burnout in human services agencies is like an infection in hospitals: it gets around” (Edelwich & Brodsky, 1980, p. 25). Similarly, both workers and customers are able to develop shared evaluations of services because each actor is subjected to similar experiences in the delivery of services (workers) or the use of these services (customers; Peiró et al., 2005). Two additional processes support this collective approach: the attraction-selection-attrition (ASA) and social information processing (SIP) models. The ASA model suggests that people with similar characteristics are attracted to, selected into, and retained by the same group (Schneider, Goldstein, & Smith, 1995). Staff members who share similar values, rules, and procedures facilitate the emergence of shared climates (Schneider & Reichers, 1983), Hence, the SIP model suggests that people use information modeled from others (e.g., other family members) in developing shared perceptions of reality (Salancik & Pfeffer, 1978). This shared information causes their perceptions to become more similar. The consideration of the collective level of construct and analysis is also relevant from a practical point of view. As mentioned above, individual burnout is less devastating than a collective burnout situation. Similarly, achieving optimal collective service performance perceptions among family members is more relevant in practical terms than explaining the individual evaluation of the user. In addition, the center level makes it possible to defend collective strategies (e.g., Muhonen & Torkelson, 2008) to achieve shared optimal levels of service performance and deal with burnout in the group of workers as a whole.
Alternative Theory and Hypotheses
Our model, previously described, is based on the predominant rationale proposing that worker stress is associated with negative outcomes (e.g., Skirrow & Hatton, 2007). However, there is an alternative interpretation of this relationship. Taking into account a possible asymmetry in the role of negative emotions (Hickey, 2014), emotional exhaustion could be related to the desire to have a beneficial impact on others. Therefore, a positive relationship can exist between emotional exhaustion and service performance. By contrast, cynicism will be negatively related to this outcome. With this in mind, the current research study also tests whether this alternative proposition (asymmetry) is supported by the results. Lack of support for alternative hypotheses allows a more solid confirmation of our proposed model.
First Study Method
Sample and Procedure
The final sample of the first study, after eliminating data due to missing values, was composed of 100 healthcare organizations representing 714 workers and 612 family members of people with IDD. Data were collected from healthcare organizations specialized in attending to people with IDD and affiliated with the Confederation of Organizations for Persons With Intellectual Disabilities (FEAPS, Spain). Participation was confidential and voluntary. Workers and family members were randomly selected. Researchers trained one worker per center to prepare and implement the random selection of participants. To be eligible, workers had to be in contact with people with IDD and their families as part of their daily work. Their main professions were psychologists, occupational therapists, physiotherapists, social workers, and primary health care workers. In order to participate in the study, the family members had to have frequent contact with the center in question, its activities, and its workers. On average, family members had been using the center for about 11 years. Therefore, they had knowledge about the functioning of the centers, services delivered by workers, and activities. Two main types of centers participated in the study: occupational centers (devoted to facilitating the transition to work of people with IDD) and day care services (dedicated to organizing educational, therapeutic, and social-leisure activities). Centers had a range of 3–19 employees (M = 7.14, SD = 2.26) and 3–10 family members (M = 6.12, SD = 2.07). The average age of employees was 37.51 (SD = 9.36), their average job tenure was 6.71 years (SD = 5.96), 75.03% of them were women, and 51.68% had earned a university degree. The average age of family members was 57.33 (SD = 1.33), and 66.61% of them were women.
The central dimensions of burnout, exhaustion (5 items), and cynicism (4 items) were measured with the Spanish validated version of the Maslach Burnout Inventory General Survey (see Schaufeli, Salanova, González-Romá, & Bakker, 2002). An example of an item is “I have become less enthusiastic about my work.” All items were rated on a Likert response scale ranging from 0 (never) to 6 (everyday). As mentioned above, we used the measure of one superordinate factor of burnout by combining the items corresponding to exhaustion and cynicism.
We used the Service Quality Scale validated by Sánchez-Hernández, Martínez-Tur, Peiró, and Moliner (2010) in a cross-cultural investigation. The scale included 21 items measuring aspects such as reliability, responsiveness, assurance, personalized attention, authentic understanding, extras, and empathy. An example of an item is “We show a real interest in creating a good relationship with the users.” All items were rated on a Likert response scale ranging from 1 (strongly disagree) to 7 (strongly agree).
Family Members' Measures
We used the Service Quality Scale validated by Sánchez-Hernández et al. (2010). Items were identical to those for employees, with the only difference being that the wording was adapted to family members.
Satisfaction with the service
To assess family members' satisfaction with the service, we used the four-item scale by Burnham, Frels, and Mahajan (2003). The scale included 4 items: (a) “I am satisfied with this center” (1 = very unsatisfied to 5 = very satisfied; M = 4.61, SD = .68); (b) “Compared to what you expected, what your family member with intellectual disability obtains from this center is:” (1 = much worse than what I expected to 5 = much better than what I expected; M = 4.24, SD = .68); (c) “Imagine an ideal center with these characteristics. How does this center compare to the ideal center?” (1 = much worse than the ideal center to 5 = just like the ideal center; M = 4.06, SD = .84); and (d) “How well does this center meet the needs of your family member with intellectual disability?” (1 = extremely poorly to 5 = extremely well; M = 4.27, SD = .66). Because four different response scales were used, ratings for the four items were first standardized to account for differences in the scales and then averaged together to form a customer rating of satisfaction with the service. This procedure follows previous research efforts (e.g., Gentry, Weber, & Sadri, 2008) using standardized items measured with different response scales. As shown in Table 1, all reliability estimates were satisfactory, with values above .70.
The general data analyses plan included four steps:
Step 1. Aggregation at the center level
To connect worker and family member measures, data were aggregated at the center level. Despite the aforementioned theoretical arguments for aggregating worker and family member data at the center level, statistical justification should be also considered. This is a necessary initial step to continue with other statistical analyses at the center level.
Step 2. Measurement issues
Another initial step was the testing of the dimensionality of the measures through confirmatory factor analyses. The objective of these analyses is to test whether the original structure with different measures (e.g., burnout vs. service quality for workers) fit the data better than an alternative single-factor model where all items upload in only one factor. If findings support the original structure, discriminant validity exists, indicating that participants were able to distinguish among the different constructs involved in the study.
Step 3. Testing the hypothesized model
To test the proposed model (Figure 1), the mediation of service quality reported by workers and family members in the link from burnout to family member satisfaction should be tested.
Step 4. Testing the alternative theory and hypotheses
As mentioned above, a different hypothesis to our model is possible. Therefore, the differential links from exhaustion vs. cynicism to service performance should be discarded to provide more solid support for our model.
Although the main statistical analyses are those corresponding to testing the hypothesized model (step 3), the other analyses are needed to test the proposed model with guarantees (steps 1 and 2) and to have a more solid test of the model (step 4). All these statistical analyses are described in more detail below.
Regarding step 1, we analyzed the appropriateness of aggregating the employees' and family members' measures to the center level. To do so, we followed the procedure suggested by González-Romá, Peiró, and Tordera (2002), who used two complementary strategies: The estimation of the intraclass correlation coefficient (ICC) considered as a consistency-based strategy, and the estimation of the average deviation index (ADM(J)) as a consensus-based strategy. The ADM(J) has some advantages over the interrater agreement index (rwg; James, Demaree, & Wolf, 1984). For instance, the ADM(J) only requires a priori specification of a null response range of interrater agreement because it does not require the modeling of the random or null response distribution. Additionally, the estimates provided by this index are in the same metric of the original response scale facilitating the interpretation of the values. The recommended cut-off values for the ADM(J) are .83 for 5-point Likert scales and 1.17 for 7-point scales (Dunlap, Burke, & Smith-Crowe, 2003), whereas ICC values as small as .05 may provide prima facie evidence of a group effect (LeBreton & Senter, 2008). In addition to the ICC and ADM(J), we carried out one-way analyses of variance (ANOVA) to ascertain whether our variables presented significant between-units discrimination (Chan, 1998).
As a second step, we conducted two series of confirmatory factor analyses (CFAs)—one per informant—to examine the measurement model for the variables. More specifically, we compared the proposed structure of measures to a single-factor model where all items upload into one only factor. To this end, we used Lisrel 8.80 (Jöreskog & Sörbom, 2006).
For the third step (testing the proposed model), we used Hayes's (2012) PROCESS macro (Model 6) for SPSS to estimate the equations of the proposed model with sequential mediators and obtain bias-corrected bootstrapped confidence intervals based on 5,000 bootstrap samples for the conditional indirect effect (Hayes, 2009). The use of bootstrapped confidence intervals provided several advantages, as all mediators could be tested simultaneously, there is no need for the variables to present a normal distribution, and the number of inferential tests is minimized (reducing the risk of a type 1 error).
The PROCESS is a relatively new, freely available tool for statistical packages that is especially useful and versatile when the researcher's objective is to simultaneously examine a number of mediators that are connected serially (Hayes, 2012). In our case, we proposed and tested the existence of two mediators: service quality perceived by workers and service quality perceived by family members. In fact, our model proposed a sequence that is serially organized: burnout reduces service quality as reported by workers. Service quality perceived by workers, in turn, is linked to service quality perceived by family members. Finally, service quality reported by family members is connected to their satisfaction with the center. As scientific knowledge advances, its objective goes beyond the relationship between two variables, and more ambitious goals are pursued, including the mechanisms underlying such relationships. These mechanisms are commonly referred to as mediation processes (see Hayes, 2012). In the current research study, we examine the following mediation process: Worker burnout is transmitted to family member satisfaction through service quality reported by workers and family members. In this context, bootstrapping provides a resampling technique where statistics can be computed using a high number of samples extracted from the original sample. With this technique, the observed results are increasingly close to the reality of the investigated population, and potential errors are significantly reduced.
Finally, to examine the alternative hypotheses (Step 4), we computed separate correlations between emotional exhaustion vs. cynicism, on the one hand, and the indicators of service performance, on the other. The possible existence of positive correlations for exhaustion and negative ones for cynicism would indicate that our proposed model is questionable. By contrast, negative correlations for both exhaustion and cynicisms would be congruent with the proposed model.
First Study Results
The computed indexes are shown in Table 1. The ICC, mean ADM(J), and ANOVAs showed sufficient within-unit agreement and between-units discrimination to aggregate the employees' and family members' measures to the center level of analysis. The alternative of computing cross-level analysis, from center-level employees' measures to individual-level family members' measures, seems inappropriate in the current study. In addition to the aforementioned theoretical arguments, empirical findings indicate that the agreement among family members on service quality (ADM(J) = .52) and satisfaction (ADM(J) = .39) is greater than the agreement among workers on burnout (ADM(J) = .91) and service quality (ADM(J) = .60). The highest agreement among family members is probably based on the long-term relationship with the center (M = 11 years), which facilitates a consensual view and the emergence of collective evaluations beyond the individual. Moreover, Preacher, Zyphur, and Zhang (2010) suggested that statistical mediation can occur only at the highest level considered in the research study.
For the measurement model, and based on previous recommendations (Hall, Snell, & Foust, 1999), we created seven three-item parcels for employees' and family members' perceptions of service quality. The measurement model involving employees' measures and differentiations between burnout and service quality fits the data well: χ2(114) = 687.69, p < .01, RMSEA = .08, NNFI = .95, CFI = .96, SRMR = .08. Moreover, the fit was better than an alternative model that forced all variables to load into one general factor (Δχ2 = 2117.91, Δdf = 1, p < .01). These results provided support for the discriminant validity of employees' constructs.
The measurement model involving family members' measures and differentiations between service quality and satisfaction also fits the data well: χ2(43) = 189.57, p < .01, RMSEA = .07, NNFI = .98, CFI = .99, SRMR = .04; and was better than the alternative model that forced all variables to load into one general factor (Δχ2 = 298.25, Δdf = 1, p < .01). These results also provided support for the discriminant validity of family members' constructs.
Results of the Hypothesized Model
Means, standard deviations, intercorrelations, and internal reliability estimates are presented in Table 1. Figure 2 shows the standardized regression coefficients for the model. In this model, service quality perceived by workers and service quality perceived by family members (sequential) mediate the relationship between burnout and satisfaction. As expected, burnout presented a significant negative relationship with service quality perceived by workers (β = −.52, p < .01), service quality perceived by workers presented a significant positive relationship with service quality perceived by family members (β = .23, p < .05), and service quality perceived by family members presented a significant positive relationship with family members' satisfaction with the service (β = .73, p < .01). The total indirect relationship of staff burnout with family members' satisfaction was estimated at −.09, with a 95% bias-corrected bootstrap (5000 samples) confidence interval of −.17, −.01. Service quality perceived by workers and service quality perceived by family members thus mediated the negative relationship between burnout and family members' satisfaction with the service.
Alternative Hypothesis Testing
According to the aforementioned alternative hypotheses, a positive relationship could exist between exhaustion and service performance. By contrast, cynicism will be negatively related to this outcome. Correlations of exhaustion with service quality perceived by employees (−.39), service quality perceived by family members (−.20), and family member's satisfaction (−.21) were negative. The same pattern of correlations was observed for cynicism in relationship to service quality perceived by employees (−.57), service quality perceived by family members (−.24), and family member's satisfaction (−.27). Thus, the potential asymmetry in the impact of emotional exhaustion vs. cynicism on service performance was not confirmed in this study. This is congruent with the rationale underlying the proposed model.
Replication Sample Method
The final sample in this replication was composed of 86 centers representing 601 workers and 819 family members. Centers had a range of 3–12 employees (M = 6.99, SD = 2.16) and 3–22 family members (M = 9.52, SD = 3.30). The average age of employees was 35.76 (SD = 9.01), their average job tenure was 6.23 years (SD = 6.29), 71.30% of them were women, and 56.10% had earned a university degree. The average age of family members was 58.02 (SD = 11.66), and 60.90% of them were women. As was the case in the first study, participating centers were affiliated with the Confederation of Organizations for Persons With Intellectual Disabilities (FEAPS, Spain) and data collection procedures were the same.
In the replication sample, we used the same measures as in the first study, with the exception of situational constraints and satisfaction. We adapted the Socio-Technical Situational Constraints Scale by Martínez-Tur et al. (2005) to assess situational constraints in centers for people with IDD. Three items referred to the existence of social constraints and four items referred to the existence of technical constraints. Two example items are “(In this center there are) conflicting relations between employees and customers” (social) and “(In this center there is a) lack of financial resources” (technical). All items were rated on a Likert response scale ranging from 1 (strongly disagree) to 7 (strongly agree). To assess family members' satisfaction with the service in the replication sample, we used the four-item reduced version (Gotlieb, Grewal, & Brown, 1994; Martínez-Tur et al., 2006) of the customer satisfaction measure developed by Oliver (1980). An example of an item is “I feel happy about my decision to choose this center.” Items were scored on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree).
As indicated in Table 2, all reliability estimates were satisfactory, with values above .70.
Figure 3 represents the full, hypothesized model where situational constraints act as antecedents of burnout. This model replicates and expands the model in the first study by including constraints. For the data analyses, we followed the same four steps as in the first study.
Replication Sample Results
The computed indexes for testing the appropriateness of aggregating the employees' measures to the center level of analysis are in Table 2. Scores corresponding to ICC, mean ADM(J), and ANOVAs showed sufficient within-unit agreement and between-units discrimination to aggregate the employees' and family members' measures to the center level of analysis. Again, the alternative of computing cross-level analysis, from center-level employees' measures to individual-level family members' measures, was determined to be inappropriate. Empirical findings have indicated greater agreement among family members (ADM(J) ranging from .34 to .61) than among workers (ADM(J) ranging from .70 to .99), supporting center-level measures, especially among family members.
The measurement model involving employees' measures and differentiating between social constraints, technical constraints, burnout, and service quality fit the sample data well—χ2(244) = 1212.47, p < .01, RMSEA = .08, NNFI = .93, CFI = .94, SRMR = .08—and it was better than the alternative model that forced all variables to load into one general factor (Δχ2 = 3312.00, Δdf = 6, p < .01). These results provided support for the discriminant validity of employees' constructs.
The measurement model involving family members' measures and differentiating between service quality and satisfactions also fit the sample data well— χ2(43) = 158.54, p < .01, RMSEA = .06, NNFI = .99, CFI = .99, SRMR = .04—and better than the alternative nested model Δχ2 = 1483.98, Δdf = 1, p < .01). Therefore, results provided support for the discriminant validity of family members' constructs.
Results of the Hypothesized Model
Means, standard deviations, intercorrelations, and internal reliability estimates are presented in Table 2. Figure 3 shows the standardized regression coefficients for the model. As expected, both social and technical constraints presented a significant positive relationship with burnout (β = .32, p < .01; β = .39, p < .01, respectively). Burnout, in turn, presented a significant negative relationship with the service quality perceived by workers (β = −.47, p < .01), which in turn presented a significant positive relationship with the service quality perceived by family members (β = .45, p < .01). Finally, the service quality perceived by family members presented a significant positive relationship with their satisfaction (β = .71, p < .01). More importantly, the total indirect relationship between constraints and family members' satisfaction was estimated at −.05, with a 95% bias-corrected bootstrap (5000 samples) confidence interval of −.13, −.01 for social constraints (Hypothesis 5a), and at −.06, with a 95% bias-corrected bootstrap (5000 samples) confidence interval of −.15, −.02 for technical constraints. Therefore, burnout and service quality mediated the link from constraints to family member satisfaction with the centers.
Alternative Hypothesis Testing
Again, we explored the separate correlations between emotional exhaustion vs. cynicism, on the one hand, and the indicators of service performance, on the other. Correlations of exhaustion with service quality perceived by employees (−.50), service quality perceived by family members (−.13), and family members' satisfaction (−.01) were always negative. The same pattern of correlations was observed for cynicism in regard to service quality perceived by employees (−.55), service quality perceived by family members (-.13), and family members' satisfaction (−.01). Thus, the potential asymmetry in the impact of emotional exhaustion vs. cynicism on service performance was not confirmed in this study. This is congruent with the rationale underlying the proposed model.
Our studies describe the relationship between staff burnout and service performance reported by family members in centers for people with IDD. Specifically, we confirmed, in two separate studies, that worker burnout is related to customer satisfaction through service quality perceived by workers and service quality perceived by family members. As we hypothesized, the quality of service perceived by workers and the quality of service perceived by family members act as mediators in the indirect relationship between burnout and family members' satisfaction. Therefore, the main contribution of the current research study is that it provides empirical evidence for a model connecting workers' burnout to service performance from the perspective of family members in centers for people with IDD.
Taking into account the conclusions by Lawson and O'Brien (1994) and Rose et al. (1998), we hypothesized that employees with lower burnout levels would be expected to offer better service quality in centers for people with IDD. Our results confirmed these propositions because a negative and significant relationship was found between burnout and service quality perceived by workers. These findings are consistent with previous studies that relate positive emotions to positive results and research in the disabilities area that relates stress to negative organizational practices (Skirrow & Hatton, 2007). Thus, burnout experienced by care providers in their work environment has been shown to be connected to the service quality they provide to their users. According to the COR (Hobfoll, 1988), burnout reflects a situation where the worker loses valuable resources, hindering adequate service performance. If well-being is a valuable resource for employees, lower well-being means that employees' possibilities of performing with high standards of quality in their interactions with families also decrease.
Results also show that the way workers perceive the service quality they offer is positively and significantly related to the service quality perceived by family members. This finding is congruent with results observed in other research areas indicating the existence of a service chain that connects internal and external performance (Heskett et al., 1994). Thus, internal processes and experiences within organizations are transferred to external perceptions by service users (in our case, family members of IDD) because users (i.e., family members) frequently interact with centers and workers. Therefore, the quality with which professionals care for people with IDD has a considerable relationship with the way family members perceive service performance. Finally, family members' perceptions of service quality translate into satisfaction with centers for people with IDD.
In the replication sample, we extended the proposed model by also including situational constraints. Results showed that social and technical constraints perceived by employees in their daily work are positively and significantly related to their levels of burnout. These findings are in line with the idea that the job environment has a strong impact on burnout (Maslach et al., 2001). This is particularly relevant in the centers participating in our studies because of the economic crisis Spain has been experiencing for many years. The economic crisis has reduced available resources, and both social and technical constraints play a significant role in worker well-being.
In addition to the model we tested (Figure 1), our studies contribute to previous knowledge in at least two additional ways. First, our findings corroborated that burnout can be considered a reality beyond the individual (Hakanen et al., 2014; Moliner et al., 2005). Workers pertaining to the same center tend to share their burnout experiences, reflecting a contagion process that relates to collective evaluations of service performance by workers and family members. Second, we considered two sources of data (workers and family members), confirming that internal processes and experiences can be transferred to external evaluations of organizational life. This strategy also allowed us to reduce common source variance problems associated with the use of only one source of data (Podsakoff et al., 2003).
Our findings have relevant practical implications in at least two ways. First, there is a crossover relationship from the internal life of the center to the reactions of family members. Managers and workers can have the feeling that the consequences of their actions are restricted to the internal life of the center. However, achieving a healthy workplace is not only significant for the members of the center; it is also transferred to the evaluations of family members. Accordingly, centers can implement actions to improve burnout and self-evaluation of service performance and register positive changes in users' service performance evaluations and satisfaction.
Implications of indirect links from burnout and constraints to service evaluations of family members are especially noteworthy. The confirmation of the mediation supports the existence of a step-by-step process. The most direct and influential internal aspect of centers on family members' evaluations is the service quality as reported by workers. This is reasonable because the service quality behavior of workers is especially visible to family members during their interactions with workers at centers for people with IDD. However, the behavior of workers in terms of service quality is based on burnout. In turn, burnout is based on situational constraints. Therefore, reducing both constraints and burnout facilitates the creation of a positive, healthy, and productive service chain, from workers' experiences and behaviors to family members' service evaluations. The investigation of mediation processes (“how”) is valuable in itself for the understanding of phenomena (Hayes, 2012), but their practical implications are also relevant because it is possible to manage precursors that have an influence on variables leading to positive outcomes.
The confirmation of center-level constructs makes it possible to focus the intervention on a collective approach with more serious implications than traditional interventions based on the individual (see Muhonen & Torkelson, 2008). Because we confirmed the proposed model at the center level, collective strategies to improve burnout and service performance are reinforced. Specifically, the different stakeholders can participate in reducing the situational constraints existing in the center from a collective perspective in order to achieve optimal levels of burnout and service performance for the center as a whole.
The present studies have limitations that should be considered in future research efforts. First, all variables were assessed using self-reports. Although we reduced the possibilities of mono-method bias by implementing a multi-informant design that considers data collected from both workers and family members, other options offer a complementary path that can enrich the understanding of burnout and the service chain (e.g., using cardiovascular indicators associated with burnout). Second, we used a cross-sectional design. Although no causality is claimed, it could be interesting for future studies to use a longitudinal approach to further examine the relationships among the study variables more accurately. Based on previous empirical evidence in both the workplace in general (Demerouti et al., 2014; Maslach, Schaufeli, & Leiter, 2001) and in services for people with IDD (Lawson & O'Brien, 1994; Rose et al., 1998), and also taking into account well-established theoretical frameworks (COR; Hobfoll, 1988), we proposed that burnout is negatively related to family reactions. However, in a longitudinal approach, researchers could examine whether family reactions can also impact subsequent staff burnout over time. Also, the level of burnout was low in our studies. Other samples with higher levels of burnout can allow us to observe alternative relationships where the aforementioned asymmetrical impact of emotional exhaustion vs. cynicism on service performance might exist (Hickey, 2014). In other words, other workplaces could require high levels of emotional exhaustion to achieve optimal service performance. Finally, data on service outcomes and satisfaction should not be limited to the family perspective, but also based on the individual perspective. Future researchers should collect data regarding satisfaction with and outcomes from services directly from people with IDD.
In spite of these limitations, our studies contribute to the literature on burnout in centers for people with IDD by identifying the specific steps connecting burnout in these centers to service evaluations by family members. Our findings suggest that high-quality places to work are needed in order to produce good experiences among users (i.e., family members) in terms of service quality and satisfaction.
The authors are grateful for the financial support of the Spanish Agency of Economy and Competitiveness (PSI2013-48509-P, PSI2016-78158-R) and FEDER.