Burnout in Veterans Health Administration Mental Health Providers in Posttraumatic Stress Clinics

Abstract

The purpose of this study was to conduct the first assessment of burnout among Veterans Health Administration (VHA) mental health clinicians providing evidence-based posttraumatic stress disorder (PTSD) care. This study consisted of 138 participants and the sample was mostly female (67%), Caucasian (non-Hispanic; 81%), and married (70%) with a mean age of 44.3 years (SD = 11.2). Recruitment was directed through VHA PTSD Clinical Teams (PCT) throughout the United States based on a nationwide mailing list of PCT Clinic Directors. Participants completed an electronic survey that assessed demographics, organizational work factors, absenteeism, and burnout (assessed through the Maslach Burnout Inventory-General Survey, MBI-GS). Twelve percent of the sample reported low Professional Efficacy, 50% reported high levels of Exhaustion, and 47% reported high levels of Cynicism as determined by the MBI-GS cut-off scores. Only workplace characteristics were significantly associated with provider scores on all 3 scales. Exhaustion and Cynicism were most impacted by perceptions of organizational politics/bureaucracy, increased clinical workload and control over how work is done. Organizational factors were also significantly associated with provider absenteeism and intent to leave his/her job. Findings suggest that providers in VHA specialty PTSD care settings may benefit from programs or supports aimed at preventing and/or ameliorating burnout.

The Veterans Health Administration (VHA) presides over the largest mental health care system in the United States. With wide exposure to traumatic events among American service members, treatment for Posttraumatic Stress Disorder (PTSD) is an integral component of VHA mental health care. At present, the VHA is home to 120 PTSD Clinical Teams (PCTs), outpatient care programs designed to provide specialty care to veterans with military-related PTSD.

In 2006, the VHA began a national rollout initiative to disseminate evidence-based psychotherapies (EBPs) for PTSD, including Prolonged Exposure (PE) and Cognitive Processing Therapy (CPT), with the goal of making EBPs for PTSD available at every VHA facility nationwide (Karlin et al., 2010). This initiative was a crucial step toward achieving the VHA’s mission to ensure that veterans receive high-quality care for PTSD. To date the VHA has trained more than 1500 providers in PE (Eftekhari et al., 2013), and more than 3000 in CPT (Chard, Ricksecker, Healy, Karlin, & Resick, 2012). Although efforts to assess the success of these programs are ongoing, the VHA has faced a number of challenges in establishing EBPs for PTSD as the standard of care for all appropriate patients—including negative provider attitudes toward manualized treatments, and the need to develop formalized requirements for provider participation in training and supervision (Karlin et al., 2010). While there has been significant focus on instrumental preparation for VHA providers treating trauma (e.g., training, skill maintenance, supervision, skill certification), little has been done to assess and intervene in other factors that may impact treatment effectiveness. Given evidence that burnout is linked to effectiveness and poorer quality of care in healthcare settings (Leiter, Harvie, & Frizzell, 1998; Lasalvia et al., 2009), provider burnout in the VHA is a notable example of an overlooked component of PTSD treatment effectiveness.

Research has found that mental health providers may represent a population of workers among the highest risk for burnout (Thomsen, Soares, Nolan, Dallender, & Arnetz, 1999). PCT clinicians may be at particular risk for burnout due to the emotional pressures of providing evidence-based trauma care, which vicariously exposes the clinician to traumatic material on a daily basis (Linnerooth, Mrdjenovich, & Moore, 2011). Burnout in PCT clinicians may also originate from ongoing administrative changes at the VHA designed to improve pathways for PTSD treatment (Voss-Horrell, Holohan, Didion, & Vance, 2011). While beneficial to patient care, changes such as these often place additional time and workload pressures on PCT clinicians, which may exacerbate risk of provider burnout.

Although various definitions of burnout have been posited, the most commonly referenced model of burnout was first formally proposed by Christina Maslach (1982). Maslach’s model originally characterized burnout as a syndrome affecting people who work in human resources and health care, but it is now believed that burnout can emerge in any occupation (Leiter & Schaufeli, 1996). Maslach and Leiter (1997) have described three dimensions of burnout: exhaustion, cynicism, and ineffectiveness. The exhaustion dimension pertains to overall fatigue related to carrying out work demands (Maslach, Jackson, & Leiter, 1996). The cynicism dimension concerns indifference about work, often arising as a means to gain distance from work demands. Ineffectiveness relates to whether an employee feels he/she is capable of fulfilling job responsibilities (Maslach & Goldberg, 1998).

The negative impact of burnout on healthcare delivery has been the subject of significant research. One study found that nurses experiencing high burnout were rated by their patients as providing lower quality care (Leiter, Harvie, & Frizzell, 1998). The research to date suggests that burnout in mental health providers may also result in decreased effectiveness and poorer treatment outcomes (Lasalvia et al., 2009). Burnout among providers in outreach programs has been associated with more frequent voluntary and involuntary hospital admissions among patients (Priebe, 2004), as well as more negative attitudes toward patients in ward and treatment-home settings (Holmqvist & Jeanneau, 2006). In turn, negative staff attitudes appear to be associated with poorer outcomes for patients with severe mental illness in rehabilitation programs (Gowdy, Carlson, & Rapp, 2003).

As research on burnout has expanded over time (cf. McGeary & McGeary, 2012), numerous organizational-level risk factors for burnout have been identified. The following six domains have received a great deal of attention in the research literature: 1) workload, or the amount of work needing to be accomplished and whether there are sufficient resources to enable employees to meet work demands (Maslach, Schaufeli, & Leiter, 2001); 2) control, or whether individuals feel able to influence work decisions and empowered to enact change (Maslach & Leiter, 1997); 3) reward, including whether there are sufficient rewards (e.g., praise, incentives) for the work being done (Maslach & Leiter, 2008); 4) community, or the quality of social interaction with others in the workplace, including leadership (Maslach & Leiter, 2008); 5) fairness, described as whether decisions at work are perceived as fair and equitable (Maslach & Leiter, 2008); and 6) values, which refers to whether there is alignment between the values held by the employee and the needs of the organization (Maslach & Leiter, 2008).

Highlighting the influence of organizational factors on burnout, clinicians in private practice report less burnout than clinicians employed in agency settings (Ackerley, Burnell, Holder, & Kurdek, 1988; Raquepaw & Miller, 1989; Vredenburgh, Carlozzi, & Stein, 1999). Rupert and Morgan (2005) found that clinicians in independent practice experienced less emotional exhaustion and a greater sense of personal accomplishment than their agency counterparts. They found in both settings that greater time spent completing administrative and paperwork tasks was related to emotional exhaustion and decreased feelings of personal accomplishment, and that independent practitioners spent less time on these tasks than those in agency settings. In addition, independent practitioners reported more perceived control over their work environments, which was associated with less emotional exhaustion and depersonalization of clients and more feelings of personal accomplishment.

Apart from workload and control, the interpersonal culture of an organization may also have an impact on burnout. Researchers found that a supportive work environment was negatively related to professional burnout among New York social workers providing trauma care in the aftermath of the September 11 attacks (Boscarino, Figley, & Adams, 2004). These findings suggest that an emotionally supportive work environment may actually protect against the psychological challenges of working with trauma.

It stands to reason that organizational factors in an agency as large as the Veterans Health Administration could significantly impact burnout risk in its employees. Voss-Horrell et al. (2011) proposed that specific facets of VHA organizational culture may hinder perceived control and increase risk of burnout, specifically among clinicians working with trauma survivors. This may be particularly true as the system changes to meet the growing need for effective PTSD care. As the authors also point out, VHA administrative mandates often result in increased clinical responsibilities and decreased non-clinical responsibilities (such as meetings, research, teaching, etc.), and performance measures may place increasing workload pressure on VHA staff. While intended to increase veterans’ access to care, such policies may also lead to increased burnout, staff turnover, and poor job performance. Vallen demonstrated (1993) that organizations that maintain tighter controls over work products and pace contribute to higher levels of emotional exhaustion and depersonalization. Further, changing mandates are common in the VHA, and research has found that frequent organizational changes can independently create burnout (Geuskens, Koppes, van den Bossche, & Joling, 2012). Conversely, Salyers, Rollins, Kelly, Lysaker, and Williams (2013) speculate that certain aspects of the VHA may protect against burnout, such as more job security, better pay, and focus on quality improvement.

Despite being an organization that nationwide employs more than a quarter million employees (278,565 total employees as of September 30, 2008; Department of Veterans Affairs, 2009), the empirical data assessing employee burnout in VHA settings are limited. One recent study compared community-based mental health care staff to VHA mental health workers in the same city and found that VHA staff reported lower emotional exhaustion and a greater sense of personal accomplishment, higher job satisfaction, and less likelihood of leaving their current employment (Salyers et al., 2013). However, VHA staff were significantly more likely than their community-based counterparts to report finding administrative issues challenging, such as bureaucracy, red tape, and policies (22% vs. 2%; Salyers et al., 2013).

The potential negative effects of burnout are far-reaching, impacting not only the clinician and patient but the organization as well. Research has found that stressful work environments lead to staff absenteeism and lower staff retention (cf. AlbuAlRub & Al-Zaru, 2008). Absenteeism has financial implications for the organization, as do the expenses incurred from lost productivity and employee turnover. The high financial cost of turnover in health care settings, which includes the costs of hiring and training, is well-established (Berger & Boyle, 1992; Gray, Phillips, & Normand, 1996; Mott 2000; Stoller, Orens, & Kester, 2001). One study of a major medical center found that the costs associated with turnover constituted over 5% of that center’s total annual operating budget, amounting to tens of millions of dollars (Waldman, Kelly, Arora, & Smith, 2004).

The purpose of this study was to conduct the first assessment of burnout among VHA mental health clinicians providing evidence-based PTSD care. Consistent with Maslach’s model of professional burnout, negative workplace characteristics (e.g., clinical workload, control over work) were expected to be significantly and directly associated with increased burnout based on the MBI-GS. We also hypothesized that negative workplace characteristics would be directly associated with an increased number of missed days at work (e.g., “mental health days”) and intent to leave the VHA.

Method

Participants

Anyone working as a non-prescribing VHA mental health care provider in a PCT clinic across the nation was eligible for study inclusion, including both licensed and unlicensed providers, so long as the unlicensed providers were in internship or fellowship and/or supervised by a licensed provider. Participants had to be employed at a PCT at least half-time to be eligible. There were 138 complete responses to the online survey. The sample was mostly female (67%), Caucasian (non-Hispanic; 81%), and married (70%) with a mean age of 44.3 years (SD = 11.2). Approximately 72% endorsed Cognitive-Behavioral Therapy (CBT) as their primary theoretical orientation, and over half (64%) were doctoral-level providers. Mean time since licensure was 9.9 years (SD = 8.6), and 46% reported that they had been working in a PCT clinic for 4 or more years. There were eleven PCT providers in training status (3 Psychology Interns and 8 Psychology Postdoctoral Fellows) who responded to the survey. On average, survey completers reported spending 4.7 (SD=4.9) hours a week providing Prolonged Exposure (PE), 4.0 (SD=4.2) hours a week providing Cognitive Processing Therapy (CPT), 1.3 (SD=3.3) hours a week providing the cognitive-only version of Cognitive Processing Therapy without a trauma account (CPT-C), 1.5 (SD=2.8) hours a week providing group CPT, and 14.2 (SD=9.1) hours a week providing other care (not including CPT or PE). Use of PE and CPT varied widely in this sample. For example, hours spent using PE ranged from 0 to 40 hours a week among survey respondents and hours spent using CPT (in any format) ranged from 0 to 25 hours. The sample for this research included providers in both full- and part-time status on VHA PTSD Clinic Teams. Unfortunately, we did not track these designations so we could not include this information in our analyses. This was a preliminary study, so there was no a priori power analysis to guide recruitment and planned data analyses were simplified.

Procedure

This study was approved by the institutional review boards at the South Texas Veterans Health Care System, and The University of Texas Health Science Center at San Antonio. Survey participants were recruited by e-mail. Because the research targeted VHA mental health providers working with patients diagnosed with PTSD, recruitment was directed through VHA PTSD Clinical Teams (PCT). The researchers advertised the survey to PCT clinics throughout the United States based on a nationwide mailing list of PCT clinic directors. A cover letter was sent to PCT clinic directors explaining the purpose of the research and stressing that this research project was purely voluntary. PCT clinic directors were then asked to forward the e-mail to their PCT clinic providers. To minimize any appearance of institutional coercion, the introductory letter was generated and signed by our non-VA investigator. The letter was accompanied by a link to an electronic survey on Survey Monkey. Survey Monkey has an option to decline the record of IP addresses per completed survey; it therefore was chosen in order to ensure respondent confidentiality. Two iterations of the invitation letter were sent at a two-week interval to maximize participation; each invitation letter was accompanied by a statement describing how recipients could opt out of subsequent e-mails. The survey also included a list of resources for managing distress due to burnout and basic referral options for more information or help.

PCT clinic providers were informed that their participation was completely voluntary and confidential, that no identifying information would be collected, and that information gathered from the survey would be reported only in aggregate. They were also informed that there was no way for PCT clinic directors to know who did or did not respond to the survey (to avoid any potential for perceived coercion from the clinic director). Unfortunately, the need to ensure participant anonymity—due to the potentially sensitive nature of the responses and possible ramifications for respondents’ employment—precluded our ability to track participant response rates. However, current data from the Northeast Program Evaluation Center (2013) indicate that 671 PCT providers would have met our eligibility criteria during the study period. With 138 survey responses, we are able to calculate a conservative response rate of 20.6%. Given the unlikelihood that all PCT directors forwarded the survey, the response rate was likely higher. Even so, the actual response rate cannot be known, as it was impossible to verify how many PCT directors forwarded the survey and to how many of their providers.

Measures

Survey items assessed burnout through a digitized version of the Maslach Burnout Inventory-General Survey (MBI-GS). Basic demographics, training background, absenteeism due to physical or emotional exhaustion, organizational work factors, intent to leave, and novel items were developed by the research team. Demographic and training questions included gender, ethnicity, age, marital status, number of children, theoretical orientation, degree, training, licensure, and years of VHA service. The self-report survey contained items asking the PCT clinician to recall the number of work days missed in the past year due to physical problems or emotional distress (i.e., “mental health days”). Workplace characteristics were also examined by asking respondents to rate their level of agreement with statements describing workplace concerns: having too much clinical workload, insufficient staffing, being part of a coherent team, having supportive coworkers, having control over work, being treated fairly by supervisors, and being rewarded for good performance (see Table 4 for complete items). Responses were rated on a on a 5-point Likert-type scale ranging from “Strongly Disagree” to “Strongly Agree”. Respondents were also asked to rate the likelihood that they would leave their current position within the next two years on a 5-point Likert-type scale ranging from “Not Likely” to “Very Likely.”

Table 4. Coefficient beta weights and significance for predictors analyzed for all three MBI subscales and mental health days regression models.

Item	EXH Beta	CYN Beta	PE Beta	Mental Health Days Beta
DEMOGRAPHICS
Ethnicity	N/A	N/A	N/A	.716
Age	N/A	N/A	N/A	.051
WORKPLACE CHARACTERISTICS
I feel there is more clinical work than I am practically able to do	1.925**	1.550	−.563	.140
My clinic is not sufficiently staffed for the number of patients that we are expected to see	N/A	−.576	N/A	.256
I feel that I am part of a coherent team	N/A	.506	.190	−.017
I feel there is more administrative work than I am practically able to do	.497	−.421	−.084	N/A
I feel that organizational politics or bureaucracy negatively impact the ability to perform my job well	1.202	2.712**	−.833	.199
I feel I can count on the emotional support of my co-workers	−.394	−1.400	.978	−.895*
I feel I have some control over the manner in which I conduct my work	−2.556**	−1.274	1.345*	N/A
I feel I am treated fairly by my superiors	.780	−.690	−.174	N/A
I feel that my accomplishments are rewarded/acknowledged	.366	.332	.111	.390

^*significant at p=.05;

^**significant at p=.01;

N/A – Not Applicable – this variable was not entered into the associated regression analysis.

Studies examining burnout in a VHA or DoD setting have found that mental health providers report less burnout than providers in other settings, but they express organizational dissatisfaction within these agencies (Ballenger-Browning, et al., 2011; Salyers, et al., 2013). Thus, this study utilized the MBI-General Survey (MBI-GS), which measures burnout related to job stressors as well as patient care (as compared with the Human Services Survey, which places greater emphasis on patient care encounters and less on organizational factors). The 16-item MBI-GS is self-administered, takes approximately 10-15 minutes to complete, and is divided into three subscales—Exhaustion, Cynicism, and Professional Efficacy—that have been confirmed in factor analyses (Schaufeli, Leiter, Maslach, & Jackson, 1996). Statements are written in the form of personal feelings or attitudes. The measure queries various timeframes to establish the respondents’ current level of burnout. Participants use a 0-6 Likert-type scale to indicate frequency of experiencing a particular feeling regarding work from “Never” to “Daily.” Cynicism and Exhaustion (score range 0-30) indicate higher levels of burnout. High Cynicism is measured on the MBI-GS as a score of 11 or above. Moderate levels of Cynicism are reflected in scores ranging from 6-10, and low Cynicism is measured as a subscale score of 0-5 on the MBI-GS. A high level of Exhaustion is denoted by scores of 16 or higher. Moderate levels of Exhaustion on the MBI-GS range from 11-15, and low levels of Exhaustion range from 0-10. The Professional Efficacy (score range 0-36) subscale is scored in the opposite direction, with lower scores indicating higher levels of burnout. For Professional Efficacy, the cut-off scores are as follows: High Personal Efficacy is determined as a score of 30 and above. Moderate Personal Efficacy is determined by scores ranging from 11-15, and low Personal Efficacy is denoted by scores ranging from 0-10. This instrument has been shown to have sufficient validity, reliability, and internal consistency (Maslach, 1982; Maslach & Jackson, 1981; Pines & Maslach, 1978); cut-off scores distinguishing “high” vs. “low” levels for each of the three subscales have been previously established (Schaufeli, et al., 1996). The Cronbach alpha for the MBI in this sample was good (alpha=.783).

Analyses

This study was developed to test the contribution of demographics and workplace factors to burnout as measured by the MBI-GS. Demographic characteristics of the sample were summarized using descriptive statistics. Contribution of variables to MBI-GS score and missed work days were tested using a linear regression model based on a hierarchical forward enter method. We only included variables in the model that demonstrated a significant correlation with MBI-GS subscale scores or missed work days. In keeping with our hypotheses, we assessed the candidacy of variables from two different domains: demographics and workplace factors, with each factor to be entered as a different step in the hierarchical regression (allowing logits to vary freely between domains). Secondary analyses were conducted to assess the contribution of the burnout subscales to rated likelihood of leaving one’s current PCT position within two years. The Likert-type scale for this item represents an ordinal scale of increasing intensity. As such, we hypothesized systematic linear trends in burnout data across the five levels of this item (i.e., burnout scores would increase systematically as intent to leave categorically increased from “Not Likely” to “Very Likely”). Because of uneven distribution across ratings that we believe to be representative of the population as a whole, linear trends were tested using weight means univariate ANOVA. Missing data were not imputed at this stage of the research and were allowed to remain missing.

Burnout, as assessed using MBI-GS subscale scores, was the primary endpoint of this research. Most burnout research using the MBI treats the concept as multi-component, so all three subscales were tested independently without attempting to aggregate the scores into a “total burnout” data point (as was done by Lasalvia and colleagues, 2009). Although it was anticipated that the scale scores would be intercorrelated, they are theoretically distinct and were included separately in analyses. All analyses were conducted in IBM SPSS 19.

Results

Provider Burnout

Overall, the sample reported high levels of Exhaustion and Cynicism and high Professional Efficacy. Twelve percent of the sample reported low Professional Efficacy, 50% reported high levels of Exhaustion, and 47% reported high levels of Cynicism. Demographics (see Table 1) and workplace characteristics (see Table 2) were individually tested against the three MBI-GS subscales to explore their candidacy for inclusion in linear regression analyses. Only workplace characteristics were significantly associated with provider scores on all three scales. No demographics variables had any significant relationship with any of the MBI-GS subscales.

Table 1. Differences in MBI subscales based on demographic characteristics of PCTproviders who completed the online survey.

Demographic	n	Exhaustion Mean (SD)	Cynicism Mean (SD)	Professional Efficacy Mean (SD)	Number of days missed in last year due to emotional distress Mean (SD)
Gender
Male	n = 45	14.0 (8.1)	11.6 (7.5)	29.3 (4.3)	2.67 (4.1)
Female	n = 93	15.4 (7.8)	11.2 (8.3)	29.2 (6.0)	2.12 (3.3)
Marital Status
Married	n = 96	15.0 (7.8)	11.2 (7.8)	29.6 (4.8)	1.88 (3.1)
Divorced	n = 11	14.8 (8.4)	12.1 (9.8)	30.5 (3.9)	4.27 (5.2)
Separated	n = 5	18.8 (9.9)	10.8 (11.4)	31.4 (4.2)	1.8 (2.4)
Never Married	n = 26	14.2 (7.9)	11.6 (8.1)	26.9 (7.8)	3.12 (4.3)
Ethnicity
African-   American	n = 7	16.4 (7.3)	13.3 (9.3)	30.4 (3.6)	2.43 (2.4)
Asian	n = 5	12.2 (8.1)	9.4 (4.8)	33.0 (1.2)	0.60 (1.3)
Caucasian	n = 112	14.5 (7.7)	11.1 (8.0)	28.9 (5.5)	2.1 (3.4)
Hispanic	n = 7	17.6 (9.9)	13.7 (8.7)	31.0 (3.1)	3.6 (3.5)
Native American	n = 2	12.5 (10.6)	2.0 (0.0)	34.5 (0.7)	1.5 (0.7)
Other	n = 5	24.8 (3.4)	17.0 (8.3)	26.4 (8.8)	7.6 (7.0)*
Age	n = 138	−.040	−.114	.146	.211†

^*The omnibus significant difference between genders for missed emotional days was significant at p = .05 based on weighted means univariate ANOVA. Posthoc analyses using Tukey HSD show that the significant F value is best attributed to differences in missed days between respondents identifying as “Other” ethnicity compared to Caucasian and Asian respondents.

^†significant at p = .05 based on Pearson product-moment correlation coefficient, two-tailed

Table 2. Correlation between MBI subscales/number of days missed in the last year due to emotional distress (i.e.., mental health days) and PCT workplace characteristics (Spearman’s Rho, 2-tailed).

Burnout Criterion Variable	Too much clinical work	Insufficient staffing	Coherent team	Too much admin work	Organizational Bureaucracy	Supportive co- workers	Control over work	Fair treatment	Good job rewarded
Exhaustion	.372**	.157	−.146	.345**	.371**	−.200*	−.334**	−.205*	−.199*
Cynicism	.316**	.191*	−.234*	.284**	.467**	−.270**	−.331**	−.325**	−.244*
Professional Efficacy	−.244**	−.137	.207*	−.248**	−.336**	.229*	.423**	.294**	.209*
Mental Health Days	.196*	.243**	−.209*	.080	.185*	−.185*	−.118	−.104	−.171*

^*significant at p = .05 based on Spearman’s Rho, two-tailed;

^**significant at p = .01 based on Spearman’s Rho, two-tailed

Planned hierarchical regression had only one step for predicting each MBI-GS subscale (see Tables 3 and 4). Professional Efficacy was significantly predicted by workplace characteristics (p < .001) accounting for 17% of the variance. Examination of individual item beta weights revealed that control over work was the only significant predictor of Professional Efficacy (B = 1.345, p = .016). The linear regression model for Exhaustion was also significant (p < .001), accounting for 22% of the variance. Exhaustion was significantly predicted by feeling like there was more clinical work than providers felt able to accomplish (B = 1.925, p = .002) and having some perceived control over how work is done (B = −2.556, p = .001). Cynicism was significantly predicted (p < .001) by having more clinical work than providers could accomplish (B = 1.550, p = .021) and organizational bureaucracy/politics (B = 2.712, p = .002). These workplace characteristics accounted for 23% of the variance in Cynicism.

Table 3. Hierarchical regression models for MBI subscales/number of days missed in the last year due to emotional distress (e.g., mental health days).

Dependent Variable	Model	Adjusted R^2	SE of the Estimate	R^2 Change	df 1	df 2	p-value
PE	1	.167	4.98	.216	8	129	<.001
EXH	1	.222	6.95	.262	7	130	<.001
CYN	1	.231	7.06	.281	9	128	<.001
Mental Health Days	1	.071	3.45	.085	2	134	.003
	2	.101	3.39	.069	6	128	.005

Absenteeism

Demographic and workplace characteristic variables were also assessed for their suitability as predictors of the number of days PCT clinicians reported missing work over the past year due to emotional distress (i.e., mental health days). There were significant differences between ethnic groups based on number of mental health days reported in the past year (p = .016; see Table 1), and age was significantly correlated with mental health days (r = .211, p = .014; see Table 1). Post-hoc analyses (using Tukey’s HSD due to a reasonable expectation of equal variance among samples drawn from the same population) revealed that the significant between-groups difference in mental health days was accounted for by higher mean days off reported by participants self-identifying as “Other” ethnicity compared to those self-identifying as Asian or Caucasian (non-Hispanic). Numerous workplace characteristics demonstrated a significant correlation with mental health days taken (based on Spearman’s Rho, see Table 2) including feeling like there is more clinical work than one can do (r = .196, r = .021), feeling like the clinic is not sufficiently staffed for the patients needing to be seen (r = .243, p = .004), feeling like one is part of a coherent team at work (r = −.209, p = .014), feeling like organizational bureaucracy is impacting work ability (r = .185, p = .030), feeling like co-workers are supportive (r = −.185, p = .030), and feeling like good work is acknowledged by superiors (r = −.171, p = .045).

A two-step hierarchical regression model was tested to determine the contribution of significant demographic and workplace variables in predicting the number of reported “mental health days” missed from work last year (see Tables 3 and 4). The final regression model was significant (p = .005), accounting for approximately 10% of the variance in mental health days. An examination of the individual items by Beta weight revealed that feeling like one can count on the support of co-workers was the only significant predictor (B = −0.895, p = .045), though ethnicity (B = 0.716; p = .052) and age (B = 0.051, p = 060) approached significance.

Intent to Leave VHA

PCT clinicians were asked to identify the likelihood that they would leave their present position within the next two years. In analyzing this variable, respondents identified as being in training status (3 psychology interns, 8 postdoctoral fellows) were removed from the analyses because these individuals are likely to leave their current position upon completion of their training (and not due to burnout). Even so, approximately 32% of the sample (n = 41) indicated that it was either “Somewhat Likely” or “Very Likely” that they would leave their current position, while 58% (n = 74) indicated that it is “Not Likely” or “Not Very Likely” that they would leave. Potential demographic differences were analyzed using the Mantel-Haenzel linear-by-linear association to detect linear changes in categorical demographics frequencies (gender, ethnicity, marital status) as the categorical likelihood of leaving one’s position increased from “Not Likely” to “Very Likely” (see Table 5). No differences were found for any categorical demographic variable and there was no difference in age between the five ratings for likelihood to leave (univariate ANOVA, weighted linear term p = .388; see Table 5).

Table 5. Differences in demographic characteristics of between PCTproviders who are planning to leave their position within two years and those who do not.

Demographic	n	How likely is it that you will leave this position within two years (%)?					p-value *
		Not Likely n = 56	Not Very Likely n = 18	Neutral n = 12	Somewhat Likely n = 20	Very Likely n = 21
Gender
Male	n = 45	52.4	7.1	4.8	11.9	23.8	.953
Female	n = 93	40.0	17.6	11.8	17.6	12.9
Marital Status
Married	n = 96	46.0	13.8	10.3	14.9	14.9	.607
Divorced	n = 11	45.5	9.1	0.0	9.1	36.4
Separated	n = 5	40.0	20.0	0.0	20.0	20.0
Never Married	n = 26	37.5	16.7	12.5	20.8	12.5
Ethnicity
African- American	n = 7	16.7	16.7	0.0	50.0	16.7	.175
Asian	n = 5	25.0	25.0	25.0	25.0	0.0
Caucasian	n = 112	49.1	15.1	8.5	13.2	14.2
Hispanic	n = 7	20.0	0.0	20.0	40.0	20.0
Native- American	n = 2	0.0	0.0	0.0	0.0	100.0
Other	n = 5	20.0	0.0	20.0	0.0	60.0
Age; Mean (SD)	n = 126	44.7 (10.8)	44.4 (10.3)	41.2 (10.5)	43.9 (10.2)	48.6 (13.0)	388†

^*Mantel-Haenzel statistic (linear trend);

^†Univariate ANOVA, weighted linear term

The MBI-GS burnout scales were tested for differences based on rating of likelihood to leave one’s current position (see Table 6). Differences were assessed using the weighted linear means because of our assumption that unequal distribution of participants on intent to leave is fairly reflective of the population of interest (all PCT Clinicians). Significant linear trends were detected for all three MBI-GS subscales as intent to leave progressed from “Not Likely” to “Very Likely.” There was a significant decrease (p < .001) in Professional Efficacy with higher levels of intent to leave the PCT position r anging from a mean score of 31.0 (SD = 3.6) for those “Not Likely” to 27.9 (SD = 7.0) for “Very Likely” to leave. Exhaustion score rose significantly (p < .001) from 11.2 (7.1) for “Not Likely” up to 18.5 (SD = 9.2) for “Very Likely.” A similar pattern was noted for Cynicism, which climbed from a mean score of 7.8 (SD = 6.9) for those “Not Likely” to leave up to a mean of 16.1 (SD = 9.3) for those “Very Likely” to leave.

Table 6. Examination of mean MBI-GS subscale scores across ratings of likelihood to leave one’s current PCT position within two years.

How likely is it that you will leave this position within two years?	n	PE Mean (SD)	EXH Mean (SD)	CYN Mean (SD)
Not Likely	56	31.0 (3.6)	11.2 (7.1)	7.8 (6.9)
Not Very Likely	18	30.0 (3.8)	13.1 (5.8)	10.0 (4.9)
Neutral	12	26.3 (5.9)	19.8 (5.1)	15.3 (8.3)
Somewhat Likely	20	26.3 (7.6)	19.6 (6.4)	16.9 (7.3)
Very Likely	21	27.9 (7.0)	18.5 (9.2)	16.1 (9.3)
Weighted linear p-value		<.001	<.001	<.001

Discussion

The present study investigated the extent to which organizational factors impact burnout among PCT providers in VHA PTSD outpatient clinics. Overall, we found high levels of burnout in the study population for two dimensions of burnout as measured by the MBI-GS: 50.0% reported feeling exhausted, and 47.1% reported feeling cynical. Only a relatively small percentage (12.3%) reported low professional efficacy. High professional efficacy is common among health care providers working with war-injured patient populations (cf. Hagerty, Williams, Bingham, & Richard, 2011), so it is not surprising that the majority of the PCT clinician sample would report high professional efficacy. Findings of high exhaustion and cynicism, however, are more concerning.

While we found that organizational factors had a significant impact on the experience of burnout, demographics made no contribution to the regression models. These findings are consistent with prior work demonstrating significant contributions of workplace and job characteristics to burnout across various occupational contexts (Fernet, Austin, Trepanier, & Dussault, 2013), and recent meta-analyses finding modest to non-significant contributions of demographics to burnout (Purvanova & Muros, 2010; Smith & Silva, 2011).

The results indicate that PCT providers may be unlikely to feel effective at their job if they lack control over the way in which they conduct their work. Providers who felt they lacked control over their work, as well as those who felt they had too much clinical work, were also more likely to feel exhausted on the MBI-GS and may be at greater risk of exhaustion when patient care loads are large (especially if they feel as though they have little control over how and when patient care can be scheduled and completed). In an organization as large as the VHA, policies controlling the manner in which work is conducted are often implemented with the goal of ensuring productivity and quality of care. Prior research suggests that top-down management structures may leave little room for feedback from frontline employees, making it difficult for leadership to identify negative effects, such as burnout, associated with a given policy (Leiter, & Harvie, 1998). Although these findings significantly implicated control over work as an important factor in PCT clinician burnout, more research is needed to shed light on the specific mechanisms underlying this relationship.

Voss-Horrell et al. (2011) have previously suggested that VHA administrative policies may impact providers’ perceived sense of control, which may in turn increase risk for burnout. For instance, they cite what is known as the “8/14 measure”—a performance measure intended to ensure that all Veterans of Operations Enduring Freedom (Afghanistan), Iraqi Freedom and New Dawn (Iraq) receive eight EBP appointments within 14 weeks after receiving the second diagnosis of PTSD (Department of Veterans Affairs, 2012)—as having unintended effects such as increasing already full provider caseloads, which they argue may diminish the provider’s perception of control, increase risk for burnout, and reduce job performance and satisfaction. Indeed, prior research indicates that a perceived sense of control in mental health workers protects against burnout (Abu-Bader, 2000). Our findings indicate that this population of mental health providers may be at particular risk for burnout when they feel less control over their work; however, additional research will be required to understand factors that specifically predict levels of perceived control among PCT providers. Qualitative research in this area would be particularly useful in elucidating how providers make sense of and manage changing policies, mandates, and performance measures within VHA, as well as factors affecting their sense of control.

It is worth noting that providers in our sample reported feeling that they were professionally efficacious. While these findings do not identify specific reasons for a sense of professional efficacy in PCT providers, Finley (2011) found that PCT providers often reported significant satisfaction associated with the use of PE and CPT to treat Veterans with PTSD, because of the treatments’ effectiveness (Karlin et al. 2010). This may contribute to the overall sense of efficacy in this population, although professional efficacy among VHA providers remains relatively understudied and more research will be required.

The remaining dimension of burnout, Cynicism, was most impacted by perceptions of organizational politics/bureaucracy and increased clinical workload. This is consistent with existing research, as workload is a well-established predictor of provider burnout across a variety of settings (Maslach, Schaufeli, & Leiter, 2001). Moreover, in prior qualitative work, PCT providers have reported feeling that their clinical judgment is sometimes at odds with VHA organizational policy (Finley, 2011), which may be associated with increased cynicism. In addition, it is possible that organizational politics and bureaucracy may contribute to providers’ perception of their work environment as more or less supportive, and supportive work environment has previously been shown to be associated with lower levels of burnout among trauma professionals (Boscarino, Figley & Adams, 2004). Additional research will be required to clarify which aspects of organizational politics/bureaucracy are most troubling to providers and most strongly associated with burnout.

Workplace absenteeism is an extraordinarily costly phenomenon, accounting for billions of dollars in lost productivity annually (Harrison & Martocchio, 1998). In specialty care environments, like PCT Clinics, absenteeism can impact both the organization (through lost revenue from care visits) as well as the veterans whose suffering is prolonged in the absence of treatment resources. In the present research, both demographic and workplace variables predicted mental health days. The resulting hierarchical regression model accounted for 10% of the variance, but the only significant predictor of mental health days was emotional support from coworkers. Those who reported higher levels of collegial support reported taking fewer mental health days. This is consistent with the broader literature on work absenteeism, in which the importance of social support has been well-documented (Vaananen et al., 2003), and suggests that emotional support may be an important resource for sustaining provider resilience in PCT settings.

All three dimensions of burnout were significantly associated with a provider’s intent to leave a PCT position within two years. Of the sample studied, 32% of PCT respondents expressed their intent to leave their current position (i.e., rated their departure as either “Somewhat Likely” or “Likely”). As one would expect, those who intended to leave reported lower levels of Professional Efficacy and higher levels of Exhaustion and Cynicism compared to those who did not intend to leave (i.e., who rated their intent to leave as “Not Likely” or “Not Very Likely”). Approximately 9% of the survey respondents rated their intent to leave as “Neutral.” Interestingly, these “Neutral” providers had mean burnout scale scores that were very similar to those recorded by providers who expressed a “Likely” or “Very Likely” intent to leave. As expected, those who were unlikely to depart were less cynical, less exhausted, and felt more efficacious at work. There were no demographic differences associated with level of intent to leave a PCT position.

The results of this study should be qualified by several methodological limitations. There may have been sampling bias in that providers responding to the survey were disproportionately experiencing burnout. Conversely, providers handling the busiest clinical workload, and therefore at greater risk for burnout, may have been less likely to complete the survey due to competing time demands. In addition, PCT Directors may have been less inclined to forward the survey to staff members perceived as being overworked or at greater risk of burnout. It is therefore difficult to determine whether sampling bias may over- or underestimate burnout in this population. Although every effort was made to circulate the survey to all PCT clinicians, there was no way to track respondents vs. non-respondents because anonymity was vital to completing research on this sensitive topic, and potential implications for respondents’ employment was a primary concern. Thus we were not able to confirm a response rate beyond our estimate of 20.6%, which likely underestimates our actual response, as noted above. For greater response rates, future research may benefit from inviting participation from providers directly, rather than via leadership. One way to support this research, as well as burnout education, would be to develop a listserv of PCT providers. Also, some outcome measures were developed by the investigators and not previously validated. Although the use of these items may suggest a threat to assessment validity, there is evidence suggesting that single items like those used here can be as valid and reliable as multi-item measures for constructs related to job satisfaction (Nagy, 2002). PCT provider burnout was measured using the MBI-GS rather than the MBI-HSS. The MBI-GS has been shown to be valid and reliable across all occupations (to include the human service occupations; Leiter, & Schaufeli, 1996), so there is no reason to believe using the more general burnout inventory reduced the validity of the burnout assessment. The MBI-GS was chosen for this work due to an expectation that organizational factors would most strongly impact burnout in this population, which proved to be the case. Finally, the sample included respondents who were full-time PCT clinicians and some who were only part time in the PCT. As a result, differences in hours spent with EBTs may be accounted for by this difference in appointment status. Unfortunately this was not assessed or tracked in this sample so we could not further explore the contribution of full- versus part-time status on PCT provider burnout. However, we did systematically track a number of variables associated with VHA clinical experience and believe that our findings translate well into the VHA as a whole. More granular findings specific to the PCTs are certainly targets for future research.

As the single largest mental health care system in the US, the impact of VHA organizational culture on provider burnout can be seen as a public health concern—not only in terms of the potential impact on quality of care for millions of veterans, but also on the health of the largest mental health care workforce in the nation. Our findings suggest that many providers in VHA specialty PTSD care settings may benefit from programs or supports aimed at preventing and/or reducing burnout, to include those designed to foster social support among colleagues. Few standards exist for developing burnout prevention programs among providers facing the unique, combined challenges of providing intensive trauma specialty care and working within a large organization like the VHA (Linnerooth, Mrdjenovich, & Moore, 2011). The critical next step in this effort will likely be a broad-scale needs assessment aimed at identifying the prevalence and severity of burnout among PCT specialty care providers, with particular attention to key issues highlighted in this paper.

For example, more research is needed on specific VHA organizational factors influencing burnout among PCT care providers. Further, future research should include comparison groups outside PCTs, in both VHA and non-VHA settings, in order to more precisely identify the contribution of organizational factors versus trauma work to provider burnout. Additional research examining the impact of specific types of treatment offered in PCTs (e.g., PE, CPT, supportive care) on provider burnout is also needed. Moreover, research is needed to examine factors protecting against burnout more directly. Based on the findings of this research, we can potentially build upon existing VHA infrastructure already designed to support workforce development and satisfaction, such as the National Center for Organizational Development, or to develop new programs or interventions as necessary. There may also be some value in bringing in outside expertise in industrial/organizational psychology to support research and intervention development efforts, as a third party may be better able to foster a sense of psychological safety in addressing work-related burnout. Given the potential costs involved in loss of productivity and staff turnover, resources used to reduce workplace stressors, lower burnout, improve staff morale, and encourage staff retention are likely to represent an excellent investment.

Finally, as the VHA continues to face a growing number of veterans requiring postdeployment mental health care, it will be important to determine whether burnout measurably impacts treatment outcome in mental health care settings. Because EBPs for PTSD have been shown to be largely effective (cf. Karlin et al., 2010), it is essential that potential impediments to the continued implementation and sustainment of EBPs—including the health and professional satisfaction of the clinicians implementing them—are well-understood. Such understanding will help to assure the professional longevity of the VHA’s highly-trained PCT clinicians, with benefits for not only clinicians themselves but for the organization and, ultimately, our nation’s veterans.