A Test of the Job Demands-Resources Model with HIV/AIDS Volunteers

Abstract

Burnout has been identified as a significant factor in HIV/AIDS volunteering. It has been associated with depression, anxiety, and the loss of volunteers from the health care delivery system. The aim of this study was to test the independence of the health and motivational processes hypothesized within the Job Demands-Resources model of burnout in HIV/AIDS volunteers. Participants were 307 HIV/AIDS volunteers from state AIDS Councils throughout Australia who completed self-report measures pertaining to role ambiguity and role conflict, social support, burnout, intrinsic and organizational satisfaction, and depression. Findings suggested that the independence of the dual processes hypothesized by the model was only partially supported. These findings provide a model for burnout that gives a framework for interventions at both the individual and organizational levels, which would contribute to the prevention of burnout, depression, and job dissatisfaction in HIV/AIDS volunteers.

KEYWORDS:

• burnout
• Job Demands-Resources model
• volunteers

Many social services, including HIV/AIDS services, rely on volunteers to undertake a significant amount of work. Despite this, most of the research on burnout that has taken place has focused on paid workers, usually service personnel (Grube, Beckerman, & Strug, 2002; Halbesleben & Buckley, 2004; Lee & Ashford, 1996; Maslach, Schaufeli, & Leiter, 2001). There are fewer studies that have examined the role of burnout in volunteer work (e.g., Lewig, Xanthopoulou, Bakker, Dollard, & Metzer, 2007). The current study is situated within the context of HIV/AIDS volunteering due to the prominence of burnout problems among these volunteers. For example, Claxton, Catalan, and Burgess (1998) found that 75% of HIV/AIDS volunteers reported moderate or high levels of burnout. Burnout within HIV/AIDS volunteers has been associated with mental health problems. In a single-cohort cross-sectional study, Claxton et al. (1998) found that 41% of volunteers were classified as suffering from anxiety and/or depression. Similarly, Guinan, McCallum, Painter, Dykes, and Gold (1991) found that over one third of HIV/AIDS volunteers showed evidence of psychological morbidity. Additionally, Raphael, Kelly, Dunne, and Greig (1990) found that 37% of volunteer AIDS counselors had a high likelihood of psychological morbidity and 14% severe psychiatric morbidity.

In addition to mental health problems, burnout has also been associated with a loss of volunteers from the HIV/AIDS health care delivery system. In a longitudinal study of HIV/AIDS volunteers conducted by Ross, Greenfield, and Bennett (1999), the dropout rate for HIV/AIDS volunteers was 53% over 2 years.

Despite this research within the context of HIV/AIDS volunteers, little of the previous research has been placed within a useful and easily comparable theoretical framework such as the Job Demands-Resources Model (JD-R model) (Demerouti, Bakker, Nachreiner, & Schaufeli, 2001). This article examines burnout in HIV/AIDS volunteers from the perspective of the JD-R model. By doing so, burnout within this context becomes more theoretically informative, allowing better integration of previous findings beyond individual variables influencing burnout into an integrated coherent process of burnout.

The Job Demands-Resources Model of Burnout

The JD-R model of burnout (Demerouti et al., 2001) consists of two independent psychological processes, each associated with a different component of burnout. The first process, the health impairment process, is activated by excessive job demands that lead to exhaustion, one component of burnout, and ultimately to physical and psychological health problems. The second process is motivational in nature, hypothesizing that resources, when depleted, lead to a decrease in engagement, a second, motivational component of burnout. This motivational decrement then causes a reduction in job-related outcomes such as performance and satisfaction. These two processes are described in more detail later.

THE HEALTH IMPAIRMENT PROCESS

According to the JD-R model, the impairment of physical and psychological health is driven by excessive job demands. Bakker, Demerouti, and Euwema (2005) define job demands as physical, social, and organizational aspects of a job that require sustained mental or physical effort and therefore are associated with psychological or physical costs. These can include job characteristics such as role conflict, role ambiguity, and excessive work load. According to the JD-R model, chronic emotional exhaustion, a component of burnout (Maslach & Jackson, 1981), is characterized by a lack of energy, depleted emotional resources, and feelings of being emotionally unavailable. The central premise of the health impairment process is that job demands do not affect health outcomes directly, but the effect is mediated through increased exhaustion and particularly emotional exhaustion. It is increased exhaustion brought on by demands that is the direct cause of negative health outcomes.

There are many possible demands of relevance to volunteer work. The meta-analysis by Lee and Ashforth (1996) examined the role stress literature within the working context and found that work demands such as role ambiguity, role conflict, and role overload were associated with emotional exhaustion. As these demands are also of relevance to volunteer work, it is useful to focus on them so as to allow increased comparability between the different contexts. In the current study, all three demands are used to test the health impairment process in volunteers as posited by the JD-R model.

Role ambiguity is defined as the degree to which clear information is lacking about the expectations associated with a particular role (Singh, Goolsby, & Rhoads, 1994). Ross et al. (1999) identified role ambiguity as a longitudinal predictor of burnout and dropout in HIV/AIDS volunteers. Held and Brann (2007) also noted high levels of uncertainty as a work-related stressor in HIV/AIDS volunteers.

Role conflict is defined as the degree of congruency among the set of expectations associated with a role (Rizzo, House, & Lirtzman, 1970) and has been associated with role stress (Singh et al., 1994). To date, this construct has not been examined in the HIV/AIDS volunteer context; however, role conflict has been found to contribute to the development of burnout, and especially emotional exhaustion, in paid workers (Lee & Ashforth, 1990). For example, higher role conflict has been associated with burnout in information systems professionals (Sethi, Barrier, & King, 1999) and with increased job dissatisfaction, exhaustion, and turnover intent in nurses (Zellars, Hochwarter, Perrewe, Miles, & Kiewitz, 2001).

In addition to role ambiguity and role conflict, role overload has been identified as a role stressor. Crook, Weir, Willms, and Egdorf (2006) identified role overload as a contributor to stress and burnout in HIV/AIDS volunteers. In particular, they suggest that a lack of knowledge or skills may affect the ability of volunteers to perform their role and that limited organizational resources may affect workload, role effectiveness, and satisfaction. The present study examines role overload in terms of the number of hours per week each volunteer spends caring for clients.

THE MOTIVATION PROCESS

The second process within the JD-R model is the motivational process, in which the absence of job resources undermines motivation leading to disengagement and cynicism, and ultimately reduced performance and other job-related outcomes. Job resources refer to physical, social, and organizational aspects of a job that serve to promote positive job-related outcomes. Job resources can operate in three ways: to reduce the impact of job demands and their costs, to stimulate personal growth, and to be functional in achieving work goals. Similar to the health impairment process, a lack of resources does not affect job-related outcomes directly but rather indirectly through the decreased levels of engagement they produce. Engagement is a component of burnout that reflects psychological motivation (Salanova, Agut, & Peiro, 2005) and involves the development of negative feelings toward others, resulting in cognitive and emotional distance from other people. Originally this component of burnout was referred to as cynicism or depersonalization, and it is measured using the depersonalization subscale of the Maslach Burnout Scale (Maslach & Jackson, 1981). The lack of engagement then results in reduced performance and other work-related outcomes.

The major resource that has been investigated in research, particularly within the stress-buffering literature (Sargent & Terry, 2000), is social support. Within the JD-R model, social support is a job resource that is theorized to activate the motivation process. High levels of social support operate to increase a sense of engagement with the organization and increased levels of dedication to the work and vigor in its completion (Lewig et al., 2007). Reduced social support from supervisors and co-volunteers will promote lower motivation, decreasing levels of engagement.

The JD-R model thus theorizes that job demands lead to an increase in the emotional exhaustion component of burnout, and then to negative health outcomes, and that a decrement in resources leads to a decrease in the engagement component of burnout and ultimately to negative job outcomes. The original theorizing of burnout (Maslach & Jackson, 1981) included a third component, personal accomplishment, which has not been explicitly included in the JD-R model. A reduced sense of personal accomplishment is characterized by increased dissatisfaction with task accomplishments, heightened perception of failure to make task-related progress, and perceptions of self as being ineffective and incompetent. It was dropped as emotional exhaustion and disengagement were viewed as forming the core of the burnout experience (Bakker, Demerouti, & Verbeke, 2004).

An argument can be made, however, for revisiting the role of personal accomplishment, particularly within the context of volunteer work. With regard to the motivational process, the JD-R model focuses on the consequences of reduced levels of job resources and suggests that driving disengagement is a (weakened) motivation process due to lowering of resources. Personal accomplishment, however, could also be a result of a positive motivational process, a “gain” effect, as suggested by the Conservation of Resources model (Hobfoll, 1989). With greater job resources, higher motivation might ensue, leading to higher feelings of personal accomplishment. The weaker empirical results found in relation to the correlations between job demand and resources, and personal accomplishment may also be somewhat contextual (Lee & Ashforth, 1996). Most empirical work has been conducted within the paid work force. For volunteers, the experience of personal accomplishment may be more important because of fewer other formal rewards compared to paid workers. Some evidence for such a situation has been found for those who are engaged with HIV/AIDS volunteering, which has been found to be associated with intrinsic rewards including heightened self-esteem and evidence of an increased sense of role significance (Crook et al., 2006). The current study thus reintroduces personal accomplishment to test the degree to which it conforms to a motivational hypothesis within volunteers, as evidenced by a relationship with resources rather than demands.

Testing the JD-R Model

The JD-R model theorizes two distinct psychological processes linking the different components of burnout to different types of outcomes and as having different types of causes. The evidence for the distinctiveness of these dual processes hypothesized by the model has been mixed (Halbesleben & Buckley, 2004), with some research supporting the two separate psychological processes, while other empirical research suggests that demands and resources exhibit cross-process effects. For example, several studies have reported that while disengagement was related to resources as hypothesized, exhaustion was related to both demands and resources (e.g., Bakker et al., 2004; Schaufeli & Bakker, 2004), thereby suggesting the separation of the two psychological processes is not as complete as theorized. As concluded by Halbesleben and Buckley (2004), the specific paths within the model require refinement, with evidence for the existence of two distinct processes not well supported. It appears that there may well be cross-process paths whereby demands affect motivational processes and resources affect health impairment processes. This pattern of results is probably not all that surprising. The lack of some job resources can increase the demands experienced. For example, training is viewed as a resource, but inadequate levels of training can make a job more demanding, thus potentially influencing the health impairment process. Likewise, excessive demands may lead to demotivational effects. This circularity between demands and resources has been identified in earlier writing about stress (see, e.g., Lazarus & Folkman, 1984). Indeed, the meta-analysis by Lee and Ashforth (1996) demonstrated that job demands exhibited substantial correlations with depersonalization. Further, the meta-analysis showed that some resources, especially some aspects of support and a number of job enhancement opportunities, were also significantly negatively related to exhaustion.

In sum, while two general processes appear to be operating, their effects are not manifested as two independent processes. Rather, the empirical research suggests that there are cross-process effects.

Despite a moderate number of studies having been conducted to test the dual-process model, the manner in which statistical analysis of the data has been conducted has also lead to ambiguity when attempting to evaluate that research. Some researchers have tested the two processes correctly, using separate measures of exhaustion and engagement, while other studies (e.g., Bakker et al., 2005; Hakanen, Bakker, & Schaufeli, 2006; Lewig et al., 2007; Llorens, Bakker, Schaufeli, & Salanova, 2006; Singh et al., 1994) have combined these two constructs into one latent factor in modeling, thus confounding the two processes. The current study maintains these two constructs as separate measures, allowing a stricter test of the original model.

Within a volunteer context, performance outcomes are difficult to capture due to there being few indicators of volunteer performance. Rather, outcomes such as turnover and satisfaction are better indicators of motivation outcomes. In the current study, we use two measures of satisfaction: satisfaction with the volunteer job and satisfaction with the organization in which volunteering takes place. Health outcomes were operationalized in terms of psychological health, specifically depression scores on the Beck Depression Inventory (BDI). This was chosen as depression has been found to be related to burnout in previous research (e.g., Claxton et al., 1998) and is an issue among volunteers within an AIDS care context. We believe it to be a sensitive outcome measure in this context to test for psychological health outcomes.

In sum, this study tests the JD-R model within the context of HIV/AIDS volunteers. It examines the distinctiveness of the theorized dual-process model, allowing a theoretical integration of previous work in the field. Further, this study reintroduces personal accomplishment to test the degree to which it conforms to a motivational process within a volunteer context.

METHOD

Recruitment and Participants

Prior to recruitment, the project received ethical clearance through the University of Queensland's ethical review processes covering consent to participate, anonymity, and confidentiality. AIDS Council branches in each state and territory in Australia (excluding the Northern Territory and Tasmania) mailed to their volunteers currently engaged in either “buddy” or “home-care” work a study invitation package containing an information sheet, a participation consent form, and a reply-paid envelope. On receipt of signed consent forms from participating volunteers, the researchers posted a questionnaire package that contained an instruction sheet that outlined key terms used in the study, such as the distinction between the terms “buddy” and “home-care volunteers,” with buddies being defined as volunteers who supported clients who were in relatively good health, while home-care volunteers were considered as supporting those clients who were more unwell.

The respondents were asked to complete the questionnaires regarding the client they spent most time with if they were caring for more than one client with HIV/AIDS. Buddies were asked to complete the questionnaires only after their provision of care lasted at least 1 month. Home-care volunteers were asked to complete questionnaires after the completion of at least two shifts with the client. Those volunteers not currently providing care were asked to delay their responses until they fulfilled the above criteria.

A total of 830 questionnaire packages were distributed nationally to each of the AIDS Councils involved in the study. There were 307 completed packages returned, yielding a 37% response rate. Just over half were males (53%) and the mean age was 43.9 years. A summary of volunteer characteristics is outlined in Table 1.

TABLE 1 Summary of Volunteer Characteristics

	N	Percentage
Gender
Male	143	46.7
Female	163	53.3
Marital status
Never married	105	34.2
Married	65	21.2
Separated/not divorced	18	5.9
Divorced	43	14
Partner/lover	59	19.2
Widowed	15	4.9
Highest education level completed
Year 10 or lower	54	17.6
Year 12	47	15.3
Tertiary	176	57.3
Trade/other	28	9.1
Employment status
Employer/self-employed	26	8.5
Part-time/casual	63	20.5
Full-time	113	36.8
Retired	49	16
Unemployed	23	7.5
Other	31	10.1
Income
$0–$30,000	163	54.2
$30,000–$60,000	112	37.2
Over $60,000	26	8.7
State of AIDS Council
Queensland	74	24.1
New South Wales	115	37.5
Victoria	28	4.1
South Australia	42	13.7
A.C.T.	20	6.5
Western Australia	28	9.1
Sexual orientation of volunteers
Heterosexual	151	49.2
Gay	114	37.1
Lesbian	22	7.2
Bisexual	19	6.2
HIV status of volunteer at time
HIV positive	14	4.6
HIV negative	172	56
Awaiting results	7	2.3
Not tested/no reason for test	111	36.2
Types of volunteer
Buddy	160	52.1
Home-care	121	39.4
Both	13	4.2
Other/transport	6	1.9
	Mean	Standard deviation	Range
Age	43.9 years	12.6 years	20–76 years
Duration volunteer has been caring for current client	39.8 weeks	48.2 weeks	1–280 weeks
Hours per week spent in direct client contact	5.1 hours	3.9 hours	0.5–24 hours

Measures

Three sets of measures are required to test the JD-R model. First, a set of variables that operationalize job demands and resources is required. The current study operationalized job demands as role conflict, role ambiguity, and job overload. Resources were operationalized as social support from two sources: supervisors and other volunteers. The second set of variables required operationalized health- and motivation-related outcomes. The current study used depression to operationalize psychological health and volunteers' satisfaction with the job itself and with the organization to operationalize motivational outcomes. The third set of variables operationalized the mediating health and motivation processes of the model. Following many other studies, we used the Maslach Burnout Inventory to assess the two processes. One component, emotional exhaustion, operationalized health impairment, and two components, depersonalization (often referred to as engagement) and personal accomplishment, operationalized the motivation process. All variables are discussed next.

DEMANDS

Three variables were selected as job demands relevant to volunteers. The first two, role conflict and role ambiguity, were assessed with the Role Conflict and Ambiguity Scale (Rizzo et al., 1970). The scale, originally composed of 14 items, was designed to measure role conflict and role ambiguity in complex organizations. One item was removed from the ambiguity scale due to the specific work-related nature of one question that was not relevant to volunteers (“I feel certain how I will be evaluated for a raise or promotion”). This resulted in an 8-item role conflict subscale (e.g., “I have to do things that should be done differently”) and a 5-item role ambiguity subscale (e.g.,“I feel certain about how much authority I have”) that were applicable to volunteers. Responses are made on a 7-point scale (1 = “completely false” to 7 = “completely true”). Mean scores were calculated for each scale. Role ambiguity items were reverse scored to maintain a consistent scoring procedure such that higher scores indicated greater role conflict and ambiguity. Rizzo et al. (1970) reported reliabilities of .82 for the conflict subscale and .81 for the ambiguity subscale. The third job demand assessed was job overload, measured as the number of hours spent volunteering per week.

RESOURCES

Based on the format of Sargent and Terry (2000), resources were operationalized as the social support volunteers perceived to be available from two sources: supervisor social support and co-worker social support. Seven items were used to measure support from each source that tapped into perceived availability of support for work-related problems. An example item was, “How much can you count on these people to help you feel better when you experience problems with your volunteer work?” Responses were obtained using a 5-point scale, ranging from very much (1) to no such person (5). Alpha reliabilities for the subscales have been reported above .90 (Sargent & Terry, 2000).

JOB SATISFACTION

Job satisfaction was used to assess motivational outcomes and was measured using the Job Satisfaction Scale (JSS: Koeske, Kirk, Koeske, & Rauktis, 1994). The JSS is a self-report measure originally devised to assess job satisfaction in the workplace. It consists of three subscales, two of which were chosen for the present study. The intrinsic job satisfaction (e.g., “Opportunities for really helping people”) and organizational job satisfaction (e.g., “Opportunity for involvement in decision making”) subscales were used in the present study. One item pertaining to satisfaction with the field of specialization in which the volunteer works was deleted from the intrinsic subscale as it was thought that it was not relevant to the more generalist volunteer work in the study sample. Removal of the item resulted in a 6-item intrinsic subscale and a 5-item organizational subscale. Items were rated on a 7-point scale (1 = “very dissatisfied” to 7 = “very satisfied”). Alpha reliabilities for the intrinsic satisfaction subscale have been reported between .85 and .90, and between .78 and .90 for the organizational satisfaction subscale (Koeske et al., 1994).

DEPRESSION

The BDI (Beck, Steer, & Garbin, 1988) was used to assess psychological health outcomes. The BDI is a widely used 21-item self-report measure of depressive symptomatology. It uses a 4-point scale with responses ranging from 0 (e.g., “I do not feel sad”) to 3 (e.g., “I am so sad or unhappy that I can't stand it”). A meta-analysis of the BDI's internal consistency has yielded a mean coefficient alpha of .81 for nonpsychiatric patients (Beck et al., 1988).

BURNOUT COMPONENTS

Burnout was measured using the Maslach Burnout Inventory (MBI: Maslach & Jackson, 1981). The MBI is a widely used 22-item measure of burnout. It has three subscales: emotional exhaustion, depersonalization, and personal accomplishment. In the present study, respondents completed the scale with respect to their volunteer work. Responses were made on a 7-point scale (1 = “never” to 7 = “every day”). Sample items include, “Working with people directly puts too much stress on me” (emotional exhaustion), “I have accomplished many worthwhile things in my volunteer work” (personal accomplishment), and “I have become more callous towards people since beginning volunteer work” (engagement). High internal reliability coefficients for emotional exhaustion, engagement, and personal accomplishment have been reported: .86, .82, and .82, respectively (Cheung & Tang, 2007). The MBI has been used in previous studies of burnout in HIV/AIDS volunteers (Bennett, Ross, & Sunderland, 1996; Claxton et al., 1998; Ross et al., 1999).

DEMOGRAPHICS AND VOLUNTEERING CONTEXTUAL VARIABLES

A range of demographic and volunteering contextual variables have been identified as potential predictors of burnout (e.g., Acker, 1999; Cordes, Dougherty, & Blum, 1997). In the present study, the following demographics were assessed: age, gender, marital status, education, employment, income, sexual orientation, and HIV status. In addition, the following HIV/AIDS volunteering context variables were assessed: duration of volunteering for the agency, amount of time caring for clients per month, total number of clients cared for, type of volunteer (e.g., buddy, home-care), and client characteristics including age, gender, sexual orientation, and severity of symptoms and disability. With respect to the latter, volunteers rated their client on a 28-item HIV/AIDS symptoms checklist (Burgess, Irving, & Riccio, 1993) with each symptom rated on a 5-point scale (0 = “not present” to 4 = “very severe”). Clients' levels of independence in six areas of Activities of Daily Living (ADLs) (bathing, dressing, toileting, transfer, continence, and feeding) were also rated by volunteers on the Index of ADL (Katz & Akpom, 1976).

RESULTS

Preliminary Statistical Analyses

Descriptive statistics, internal consistency, and correlations for each of the measures are presented in Table 2. All measures with the exception of depersonalization exhibited satisfactory internal reliability. An attempt was made to use a subset of items with higher reliability, but no such subset was found that was adequate. The mean score for depersonalization was very low, with most respondents reporting very low scores, and across the sample, there was very low variation. Personal accomplishment exhibited marginally adequate reliability.

TABLE 2 Means, Standard Deviations, Pearson Correlations, and Reliabilities for All Variables

Measure	1	2	3	4	5	6	7	8	9	10	11	12	13	M	SD
Age	—													43.85	12.6
Client symptoms	0.05	—												53.24	18.87
Hours per week	.16* *	.31* *	—											5.11	3.85
Role conflict	–.02	.20* *	.17* *	(.86)										32.32	13.38
Role ambiguity	–.13*	–.01	–.03	.48* *	(.74)									21.34	7.33
Co-worker social support	0.01	0.01	0.06	–.24* *	–.19* *	(.95)								22.63	5.55
Supervisor social support	0.04	–.07	–.01	–.41* *	–.31* *	.45* *	(.97)							24.49	4.43
Emotional exhaustion	–.24* *	0.06	–.09*	.24* *	.37* *	–.23* *	–.26* *	(.76)						23.87	8.05
Depersonalization	–.10	–.11	–.10	.20* *	.22* *	–.17* *	–.12*	.46* *	(.46)					7.78	2.84
Personal accomplishment	–.03	0.14	.20* *	–.09	–.35* *	.27* *	.14*	–.58* *	–.33* *	(.63)				36.85	6.4
Depression	–.09	–.05	0.01	.16* *	.21* *	–.07	–.05	.30* *	0.08	–.22* *	(.83)			5.63	5.37
Organizational satisfaction	0.07	0.03	–.03	–.45* *	–.51* *	.41* *	.61* *	–.29* *	–.10	.32* *	–.17* *	(.86)		5.69	1.06
Intrinsic satisfaction	.16* *	.15* *	.25* *	–.07	–.32* *	.25* *	.22* *	–.35* *	–.13* *	.49* *	–.16* *	.51* *	(.87)	6.08	0.95

To determine whether the focal variables (demands, resources, the three burnout components, satisfaction, and depression) varied as a function of demographic or contextual variables, correlations were performed on continuous data and analyses of variance (ANOVAs) were performed on categorical data. An adjusted alpha of p < .01 was used due to the number of tests performed. Age of the volunteer was significantly correlated with intrinsic satisfaction (.13), and the severity of HIV/AIDS symptoms measured by the HIV Symptom Checklist was also correlated with intrinsic satisfaction (.18). Age of the volunteer and HIV symptoms were therefore included in the model as covariates.

Model Specification and Evaluation

ANALYTIC STRATEGY

Similar to the strategy of Bakker, Demerouti, and Schaufeli (2003), we undertook a series of analyses to tease out the various paths that effects could follow. We began by testing direct effects models (Models 1 to 3), followed by the more theoretically focused tests involving mediation (Models 4 to 9). Model fit results are presented in Table 3.

TABLE 3 Model Fit Indices for Path Models

Model number and name	Chi square	df	p	CFI	RMSEA	RMSEA 90% CI–	RMSEA 90% CI+	SRMR
Direct effects models
Model 1	101.00	8	.00	.75	.20	.16	.23	.06
Model 2	100.43	6	.00	.75	.23	.19	.27	.06
Model 3	.69	2	.71	1.00	.00	.00	.08	.01
Mediated models
Model 4	76.47	14	.00	.92	.12	.10	.15	.07
Model 5	75.53	12	.00	.92	.13	.10	.16	.07
Model 6	74.56	12	.00	.92	.13	.10	.16	.07
Model 7	56.91	12	.00	.94	.11	.08	.14	.05
Model 8	15.84	8	.04	.99	.06	.01	.10	.03
Model 9	4.63	6	.59	1.00	.00	.00	.05	.01

Direct Effects Models

Three direct effects models were tested so as to examine the pattern of relationships among the demands and resources on the one hand and the outcome variables on the other. In the first model (Model 1), a strict bifurcation was maintained, with demands predicting only BDI scores, and resources predicting only satisfaction scores. The model exhibited a poor fit to the data. Two subsequent models were then tested. In Model 2, the crossed effect from resources to health outcomes was included, such that resources were allowed to predict not only satisfaction but also BDI scores. Model fit was no better than Model 1, suggesting that resources exhibited no relationship with health outcomes. In Model 3, demands were permitted to predict both health and motivation outcomes. The model exhibited an excellent fit to the data, suggesting that in this sample, demands affect motivational outcomes as well as health outcomes but that resources only affect motivational outcomes. Model 3 thus becomes the baseline model for tests of the dual process mediation process.

Mediation Models

The JD-R model hypothesizes that the effects of demands will be mediated only through exhaustion and that resources will be mediated only via depersonalization and accomplishment. This strict dual-process model (Model 4) exhibited a poor fit to the data. There are a number of possibilities that might explain the poor fit. One set of possibilities is that the “output” side of the JD-R model is flawed and that the mediation effects are not separated out as the model suggests. Rather, exhaustion might mediate not only between demands and health outcomes but also between demands and motivational outcomes.

To test these “output” hypotheses, several models were tested. In Model 5, emotional exhaustion, the health component of burnout, was allowed to predict not only depression but also the motivation outcomes―that is, satisfaction. Model fit indicated that this crossed effect did not improve fit over Model 4 (the strict dual-process model). A second crossed model (Model 6) was also tested in which depersonalization and accomplishment, the motivational components of burnout, were allowed to predict depression, the health outcome. Again, this model fit the data no better than Model 4. Together, these three models suggest that the dual processes do bifurcate in terms of their mediation processes. Emotional exhaustion, the health component of burnout, does not appear to affect motivational outcomes, and motivation components of burnout do not appear to affect health outcomes.

The third set of models tested the “input” side of the JD-R model―that is, the effects of resources and demands on the burnout components. Two models were initially tested. In the first (Model 7), resources were allowed to predict exhaustion (i.e., to examine the effect that motivation inputs have on the health impairment process). This model exhibited a significant improvement in fit over Model 4, the strict dual-process model; however, the model fit indices suggested this model exhibited still only a marginal fit to the data. A second cross-process input model was tested (Model 8) in which demands, the health inputs, were allowed to be mediated through depersonalization and accomplishment (i.e., via the motivational process). Compared to Model 4, Model 8 also exhibited a significant improvement in fit. Taken together, these two models suggest that both crossed-input processes are operating.

A final model was therefore tested (Model 9) in which both cross processes were included, allowing both resources and demands to be mediated via all three burnout components. The model exhibited an excellent fit to the data and was a significant improvement over Models 7 and 8, both of which included only one set of crossed processes. These results suggest that the dual processes on the input side of the JD-R model are not distinct, as theorized within the JD-R model. The output side of the JD-R model was supported, with the motivation and health components of burnout significantly predicting only the motivation and health outcomes, respectively.

The path results from Model 9 are reported in Table 4 and significant paths shown in Figure 1.

TABLE 4 Model Coefficients for Each Path in Model 9

Outcome variable	Predictor	R^2	B	β
Depression	Age		–0.01	–.02
	Client symptoms		–0.03	–.10
	Hours per week		0.08	0.06
	Role ambiguity		0.06	0.09
	Role conflict		0.03	0.06
	Emotional exhaustion		0.17	.26***
		0.11
Intrinsic satisfaction	Age		0.01	.14***
	Client symptoms		0	0.03
	Personal accomplishment		0.06	.40***
	Depersonalization		0.02	0.07
	Supervisor social support		0.03	.12*
	Co-worker social support		0.01	0.08
	Hours per week		0.03	.13**
	Role ambiguity		–0.02	–.16**
	Role conflict		0.01	0.08
		0.34
Organizational satis	Age		0	0.04
	Client symptoms		0	0.07
	Personal accomplishment		0.03	.16***
	Depersonalization		0.04	.10*
	Supervisor social support		0.1	.42***
	Co-worker social support		0.02	.11**
	Hours per week		–0.02	–.06
	Role ambiguity		–0.04	–.25***
	Role conflict		–0.01	–.14*
		0.54
Emotional exhaustion	Age		–0.12	–.19***
	Client symptoms		0.04	0.08
	Hours per week		–0.18	–.09
	Role ambiguity		0.3	.28***
	Role conflict		0.02	0.03
	Supervisor social support		–0.19	–.11†
	Co-worker social support		–0.16	–.11*
		0.22
Depersonalization	Age		–0.02	–.07
	Client symptoms		0.02	.14**
	Supervisor social support		0.02	0.03
	Co-worker social support		–0.07	–.13*
	Hours per week		–0.10	–.14*
	Role ambiguity		0.05	.14*
	Role conflict		0.02	0.11
		0.11
Personal accomplishment	Age		–0.05	–.10*
	Client symptoms		0.03	0.08
	Supervisor social support		–0.04	–.03
	Co-worker social support		0.25	.22***
	Hours per week		0.26	.16**
	Role ambiguity		–0.31	–.36***
	Role conflict		0.04	0.07
		0.22

FIGURE 1 Significant standardized direct effects from final model (Model 9). †p < .08, *p < .05, **p < .01, ***p < .001.

Specific Mediation Effects Through Burnout Factors

To address the focal issue of mediation, both total and specific indirect effects were estimated using Mplus V5.1, and their standard errors and confidence intervals were estimated using 5000 bias-corrected bootstrap samples as recommended by Shrout and Bolger (2002). In total, seven significant indirect effects were identified. A summary of the effects are shown in Table 5.

TABLE 5 Unstandardized and Standardized Indirect Effects

Indirect effect	Indirect B	Indirect β	p
Coworker support–PA–intrinsic satisfaction	0.02	.09	.00
Ambiguity–PA–intrinsic satisfaction	–0.18	–.14	.00
Total hours–PA–intrinsic satisfaction	0.02	.06	.01
Coworker support–PA–organizational satisfaction	0.01	.04	.01
Ambiguity–PA–organizational satisfaction	–0.01	–.06	.00
Total hours–PA–organizational satisfaction	0.01	.03	.03
Ambiguity–EE–depression	0.05	.07	.01

Only one indirect effect was indicative of the health impairment process: ambiguity demonstrated a significant indirect effect through emotional exhaustion to depression. The remaining six indirect effects were all mediated through personal accomplishment, and all related to the motivational outcomes. One set of indirect paths mediated from co-worker support to both motivational outcomes, consistent with the motivation process of the JD-R model. The remaining indirect paths were cross effects paths in which demands were mediated via personal accomplishment to motivational outcomes. Specifically, role ambiguity was indirectly related to both satisfaction outcomes, as was hours of work.

For only three relationships was there evidence of total mediation. The direct effect from ambiguity to depression was no longer significant with the inclusion of the mediators, nor was the relationship from co-worker support to intrinsic satisfaction or that from hours of work to satisfaction. The other four significant indirect effects exhibited only partial mediation due to the presence of significant direct effects from those demands and resources to the outcome variable in question.

DISCUSSION

The aim of the present study was to examine the JD-R model of burnout in HIV/AIDS volunteers. The model hypothesizes that there are two independent processes linking demands and resources to work- and worker-related outcomes: a health impairment process and a motivational process. We found some support for the JD-R model, but the strict bifurcation of the two processes was not supported. We raise three theoretical issues confronting the JD-R model in light of our results: the relationship between demands and resources and the types of outcomes with which they are related; the relationships of demands and resources to burnout components; and the mediating effects as a test of the proposed mechanism that is central to the JD-R model. We also make some comments about the application of the model to volunteering, and specifically to HIV/AIDS volunteering.

Direct Effects of Demands and Resources on Job Satisfaction and Depression

For the JD-R model to work, arguably a primary condition is for demands to relate only to health implicative outcomes and for resources to be associated only with motivational outcomes. In the current study, resources, in the form of social support, exhibited only small and nonsignificant relationships with depression, lending support to the prediction from the model. In contrast, demands were related to both health and motivational outcomes. This was most pronounced with respect to role ambiguity, which demonstrated moderate to strong relationships with all three outcomes.

While some previous studies have found support for separate outcomes, other researchers have found resources and, more often, demands, related to both health and motivational outcomes. This mixed pattern of results is probably not surprising. As discussed previously, a reduction in resources might ultimately result in health implicative outcomes, and increasing demands could reasonably be expected to affect motivation outcomes, such as satisfaction. Selecting measures that reflect purely either health or motivational outcomes is also difficult. While some health outcomes are primarily health related (e.g., workplace injury), other outcomes, such as absenteeism, are not. Rather, arguably they have both health and motivational components. The same applies to motivational outcomes.

Relationships Between Demands and Resources and Burnout Components

The lack of independence of effects becomes more centrally problematic for the JD-R model when the relationships among resources and demands and the burnout components are examined. Both of the resources assessed and one of the demands, ambiguity, exhibited significant correlations with all three components of burnout. The remaining demands exhibited mixed relationships. The two components argued to be most central to burnout, exhaustion and depersonalization, did not demonstrate the strict differential relationships predicted by the model. This pattern of results suggests that the effects of demands and resources are not as cleanly defined and delineated as suggested by the model. As has been noted elsewhere (Halbesleben & Buckley, 2004), it is probably unrealistic to expect demands and resources to operate in distinct ways as they are entwined together to provide a psychological experience of volunteering.

Meditational Aspects of the Model

The mediation paths of the JD-R model showed that while resources were mediated through the motivational processes as hypothesized by the JD-R model, demands were mediated by the health impairment process, as theorized, but also by the motivational pathway. This pattern of results is in concert with previous research (Halbesleben & Buckley, 2004) that has shown that the dual health impairment and motivational pathways are not distinct, in contrast to the predictions of the JD-R model.

The strongest support was for the health impairment process, in which emotional exhaustion fully mediated the relationship between ambiguity and depression. Depression was not implicated in any other significant relationships in the final model. Resources were not related to depression, although support exhibited weak relationships with exhaustion. These findings indicate that negative health outcomes are clearly linked with an increase in exhaustion, which arises from increased job demands experienced by volunteers.

Motivational outcomes, however, while positively affected by increased resources, as expected, were negatively affected by increased demands. This pattern suggests that not only is the mechanism through which demands affect outcomes a health impairment process but that also demands can activate a (negative) motivational process.

Within this study, the strongest motivational effects found were related to the role of personal accomplishment. This component was implicated in three indirect effects, two of them full mediation. All three indirect effects were associated with intrinsic satisfaction, with which a strong positive relationship was found. Ambiguity was strongly negatively related with accomplishment, suggesting unclear roles have the greatest potential for affecting a sense of accomplishment among volunteers. These results support and expand upon the findings of Ross et al. (1999), who found that role ambiguity was associated with burnout in HIV/AIDS volunteers. The finding that intrinsic satisfaction was negatively affected by a lack of co-worker support also suggested that it is volunteers' reduced sense of accomplishment that explains this relationship. Hours of client contact was weakly but positively related via accomplishment to intrinsic satisfaction.

Unexpectedly, volunteer hours exhibited a negative relationship with depersonalization, such that with increasing hours of client contact, depersonalization decreased. More hours volunteering were also associated with increased personal accomplishment and decreased depersonalization. Several explanations may account for these unexpected findings. First, the majority of volunteers may not have felt “overloaded” by their workload with the mean number of hours spent volunteering in the current study being 5.1 hours per week. Second, this group of volunteers may not have perceived their workload as stressful regardless of the number of hours spent volunteering. This is supported by the findings of Gueritault-Chalvin, Kalichman, Demi, and Patterson (2000), who found that perceived workload was associated with burnout, rather than the actual number of hours spent volunteering. Third, the volunteers may have received intrinsic or extrinsic rewards, which buffered against feelings of stress associated with workload. Crook et al. (2006) found that intrinsic rewards such as heightened self-esteem and self-improvement and extrinsic rewards including constructive feedback, recognition, and appreciation of role significance are associated with HIV/AIDS volunteering. It is possible that these factors may have protected against the development of burnout and job dissatisfaction under conditions of increased workload.

Overall, these results suggest that emotional exhaustion is of concern for these volunteers with regard to possible mental health outcomes but that depersonalization has not occurred. Rather, a sense of accomplishment was found to be most potent as a mechanism for linking between demand and resources and intrinsic satisfaction.

Volunteering and AIDS

These findings add to the current understanding of burnout in HIV/AIDS volunteers as burnout has not been examined as a mediator within this population. There also appears to be some contrast between the mediating role of burnout in paid workers compared with its role in HIV/AIDS volunteers. In paid workers, all three phases of burnout have been found to mediate the relationship between role stressors and satisfaction (Bacharach, Bamberger, & Conley, 1991; Singh et al., 1994). However, personal accomplishment was found to be the only significant mediator in these relationships for HIV/AIDS volunteers. It is possible that personal accomplishment mediates in these relationships for HIV/AIDS volunteers as it acts as a stronger reward for volunteers compared with those in the paid workforce who receive other rewards including monetary and promotional rewards. Bennett et al. (1996) highlighted the significance of rewards (including personal effectiveness and empathy/self-knowing) in buffering against burnout in HIV/AIDS volunteers.

While role stressors did not directly predict depression, it was found that where volunteers became emotionally exhausted in the presence of role conflict, depression was also present. This expands on the findings of Leiter and Maslach (1988), who found that role conflict predicted emotional exhaustion. These findings suggest that even in the presence of significant role stressors, emotional exhaustion is a prerequisite for depression in HIV/AIDS volunteers. This is not surprising given the similarity between emotional exhaustion and the symptoms of depression, which include depressed mood, diminished pleasure, fatigue, and diminished concentration. The prevalence of depression in the present study was 8.5%. This is somewhat higher than the findings of Claxton et al. (1998), who used the Hospital Anxiety and Depression Scale to conclude that less than 2% of HIV/AIDS volunteers were depressed.

Summary

The findings of the present study highlight the relevance of burnout in HIV/AIDS volunteers. It is possible that volunteers experiencing burnout may benefit from relaxation or mindfulness strategies that could particularly limit the harmful effects of role stressors such as role ambiguity and role conflict. Additionally, cognitive restructuring where there are unrealistic self-expectations, perceived threats or uncertainty could assist the HIV/AIDS volunteer in coping with role stressors, thus preventing burnout. By addressing the imbalance between individual resources and job demands, role stressors may be reduced.

There are also important implications for volunteering organizations that would assist HIV/AIDS volunteers in their role by contributing to increased satisfaction and less depression. While role stressors such as role ambiguity and role conflict contribute to job dissatisfaction for HIV/AIDS volunteers, the protective role of personal accomplishment is significant in relation to role ambiguity. Clarification of role descriptions and interventions that would address role conflict through preventative conflict management practices are important, but more important is the need for volunteers to feel a sense of personal accomplishment through their volunteering role. Assessing personal values, setting personal goals, and achieving such goals, as well as discussing these with management personnel in the volunteering organization, may help volunteers feel more satisfied with their work. Additionally, it appears that feelings of personal accomplishment may enhance intrinsic satisfaction when long hours are spent in the volunteering role.

The central role of emotional exhaustion is also important for HIV/AIDS volunteers, specifically in the prevention of depression. Education initiatives by the volunteering organization outlining the symptoms of emotional exhaustion, including increased fatigue, depleted emotional resources, feeling emotionally unavailable to others, and the need to be vigilant for such symptoms, may prevent emotional exhaustion becoming severe enough to lead to depression.

The current study has a number of strengths, including the relatively large sample size used, the assessment of a wide range of demographic and contextual variables, and the inclusion of covariates as controls. The findings highlight specific interventions for HIV/AIDS volunteers and initiatives that volunteering organizations could implement in order to support HIV/AIDS volunteers. This would in turn enhance job satisfaction and prevent depression and burnout.

Study Limitations

Limitations of the present study include the use of a nonrandom sample and selection bias as volunteers were all recruited via state AIDS Council branches. The use of self-report measures as the only source of data as well as the use of non–volunteer-specific questionnaires presents limitations in this study. The low alpha coefficient obtained for depersonalization is also problematic; however, a range of factor analysis methods were used to explore other dimensional structures of the MBI. As none of these produced an improved three factor structure, the original three factors were retained. These problems are certainly not unique to this study (e.g., Bryne, 1994). For comparability reasons, most studies retain the original factor structure, as we did.

Future Recommendations

Future research should focus on examining the volunteers' perception of their workload to ascertain whether they perceive their current workload to be a role stressor. It could also be of interest to examine specific areas of role conflict (e.g., conflict with supervisors or peers) and how these have an impact on burnout. Additionally, the use of a longitudinal design to examine the accumulative capacity of burnout would contribute to the understanding of the development of burnout over time. A risk of focusing too exclusively on the individual volunteer is that important organizational policy levers such as training, including training for paid staff who manage volunteers, and policies about funding levels for actions that promote appropriate levels of resources and minimize demands are not given sufficient attention. A multiorganizational study that explicitly examines the policy environments in which volunteers work would allow these organizational factors to be explored more fully.

The authors would like to thank the participating AIDS Councils for assisting with recruiting participants for this research and, in particular, the volunteers who gave so generously of even more of their time. This research was supported by funding from the Public Health Research and Development Committee (PHRDC) of the National Health and Medical Research Council (NHMRC), Project Grant 964164.

Notes

Note. Due to missing data, not all percentages add to 100%.

N = 307.

Note. Parenthetical numbers on the diagonal are the coefficient alpha values.

*p < .05, **p < .01.

Note. df =degrees of freedom; p =chi-square probability; CFI = comparative fit index; RMSEA = root mean square error of approximation; RMSEA 90% CI– = lower bound of 95% confidence interval of the RMSEA; RMSEA 90% CI+ = upper bound of 95% confidence interval of the RMSEA; SRMR = squared root mean residual.

N = 307.

Note. B = unstandardized coefficient; β = standardized coefficient; R ² = variance explained.

†p < .08, *p < .05, **p < .01, ***p < .001.

N = 307.

Note. PA = Personal accomplishment; EE = emotional exhaustion; indirect B = unstandardized indirect effect; indirect β = standardized indirect effect.