Burnout, Compassion Fatigue, and Vicarious Traumatization Paper

Measurement and Evaluation in Counseling and Development 2015, Vol. 48(3) 177 –191 © The Author(s) 2015 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/0748175615578758 mec.sagepub.com

Assessment, Development, and Validation

School counselors are in a unique position to assist students with academic, career, and per- sonal/social concerns (American School Counselor Association [ASCA], 2012). Addi- tionally, approximately 70% of mental health care accessed by children and adolescents is completed only at school (Farmer, Burns, Phillips, Angold, & Costello, 2003). In addi- tion to the stress that school counselors face on a daily basis (Kesler, 1990), many other factors have been identified that contribute to burnout in school counselors including low collective self-esteem (Butler & Constantine, 2005), lower levels of ego development (Lam- bie, 2007), increased levels of organizational stressors (e.g., professional relationships and financial security; Wilkerson & Bellini, 2006), longer time on the job (Wilkerson, 2009), a counselor’s Adlerian lifestyle (Wachter, Cle- mens, & Lewis, 2008), type of school district (i.e., rural, suburban, and urban; Butler & Constantine, 2005), lack of supervision (Moyer, 2011), high student-to-counselor ratios (Moyer, 2011), and a large amount of nonguidance responsibilities (Coll & Free- man, 1997; Moyer, 2011). Additionally, with the increase of school violence nationwide, school counselors are called upon in times of

crisis, which can lead to excessive stress and burnout (Paine, 2009). Given the multiple and increasingly complex demands faced by school counselors on a daily basis, counselor burnout, defined as a combination of multiple physical and emotional symptoms, could compromise their well-being and their ability to serve students (Kesler, 1990).

Within the human service professions (i.e., social, medical, and legal services), burnout has been a well-researched topic for several decades. Freudenberger (1974) first conceptu- alized burnout with volunteer staff workers at a free medical clinic. Shortly after, Warnath and Shelton (1976) advocated within the counseling profession for greater understand- ing of burnout through ongoing education of the concept. Maslach and Jackson (1981) fol- lowed with researching burnout within the human services professions and developed the

578758MECXXX10.1177/0748175615578758Measurement and Evaluation in Counseling and DevelopmentGnilka et al. research-article2015

1DePaul University, Chicago, IL, USA 2Kent State University, Kent, OH, USA

Corresponding Author: Philip B. Gnilka, Department of Counseling and Special Education, DePaul University, 2247 N. Halsted St., Chicago, IL 60614, USA. Email: [email protected]

Factor Structure of the Counselor Burnout Inventory in a Sample of Professional School Counselors

Philip B. Gnilka1, Aryn C. Karpinski2, and Heather J. Smith2

Abstract This study tested a five-factor model structure with a sample of professional school counselors. Confirmatory factor analysis indicated that a modified five-factor model was the most appropriate. In addition, the levels of professional school counselors’ burnout differed from other diverse samples of counselors. Implications for professional school counselors are discussed.

Keywords confirmatory factor analysis, school counseling, counseling, counselor appraisal, counseling

178 Measurement and Evaluation in Counseling and Development 48(3)

Maslach Burnout Inventory–Human Service Survey (MBI-HSS), yielding three dimen- sions experienced within the individual: emo- tional exhaustion, depersonalization, and personal accomplishment.

While the majority of research on burnout used the MBI-HSS and conceptualized the phenomena as an individual syndrome within the social services professions, more recent research suggested taking into account domain specific environmental and organizational factors that may contribute to burnout. Indeed, Maslach (2005) asserted that burnout mea- sures should measure both domain specific factors (i.e., school counseling) and environ- mental/organizational factors (i.e., workplace environment). To address this concern, S. Lee et al. (2007) designed the Counselor Burnout Inventory (CBI). S. Lee, Cho, Kissinger, and Ogle (2010) defined counselor burnout as “a combination of multiple emotional and physi- cal ailments manifesting cognitively or within the workplace” (p. 131).

S. Lee et al. (2007) developed the CBI out of an initial pool of 296 items from several sources (i.e., focus groups, reviews of litera- ture, expert reviews) and reduced it to 40 items. An exploratory factor analysis was completed with 258 participants and found five factors: Exhaustion, Incompetence, Neg- ative Work Environment, Devaluing Client, and Deterioration in Personal Life that accounted for approximately 55% of the total variance. Next, a confirmatory factor analysis (CFA) was completed with 132 participants and found the five-factor structure was a good model fit for the data accounting for approxi- mately 67% of the variance. Rasch analysis was used to examine the average measure and threshold of each scale category and con- cluded that the five-category scoring method was appropriate for all items. Finally, conver- gent and criterion-related validity was evi- denced via relationships with other instruments including the MBI-HHS, job sat- isfaction, and self-esteem.

Several additional studies have been con- ducted investigating the construct validity of the CBI in different populations, as recom- mended by Barnes and Moon (2006). Some of

these studies have investigated different types of counselor populations, while other studies have investigated counselors practicing in countries outside the United States. J. Lee et al. (2010) conducted one of the first studies that confirmed the factor structure (using CFA) of the CBI using a sample of sexual abuse and offender counselors (N = 204). One-, two-, and five-factor models were tested and results indicated that a modified five-fac- tor model with error covariances between Items 1 and 6, and Items 10 and 15 was the most appropriate for the sample.

In 2011, Yagi, Lee, Puig, and Lee (2011) investigated the factor structure of a Japanese version of the CBI. A sample of 260 counsel- ors and psychotherapists participated in the study; similar to S. Lee et al. (2010), one-, two-, and five-factor models were tested. Results suggested a modified five-factor model with an error covariance between Items 10 and 15 was the most appropriate for the sample. This was the second study to repli- cate, using a completely different sample, an error covariance between Items 10 and 15.

Another group of researchers, Shin, Yuen, Lee, and Lee (2013), investigated the con- struct validity of a Chinese translation of the CBI among a sample of 489 school counselors in Hong Kong. Similar to previous studies (e.g., S. Lee et al., 2007, 2010; Yagi et al., 2011) a five-factor model was the most appro- priate for the sample; no additional analyses were conducted regarding modification indi- ces.

Last, Carrola, Yu, Sass, and Lee (2012) assessed the measurement invariance of the CBI across United States counselors (n = 363) and Korean counselors (n = 379). The authors concluded that factorial validity existed for the five factors and measurement invariance existed for three of the five factors (i.e., Exhaustion, Devaluing Clients, and Deterio- ration in Personal Life). Items 2, 3, 8, 12, 19, and 20 lacked invariance between the U.S. and Korean samples and suggested using cau- tion when making between culture compari- sons.

While initial support for the factor struc- ture of the CBI had been evidenced among

Gnilka et al. 179

various different samples, no studies have investigated the factor structure of the CBI using a large homogeneous sample of profes- sional school counselors in the United States. In the only study that included professional school counselors in the United States (S. Lee et al., 2007), few school counselors were in the exploratory factor analysis sample (19.4%, n = 50) and CFA sample (43.2%, n = 57). Conducting a CFA with a large sample of pro- fessional school counselors is a compelling next step in the further refinement and valida- tion of the CBI scores.

The first purpose of this study was to deter- mine whether the five-factor structure (i.e., Exhaustion, Incompetence, Negative Work Environment, Devaluing Client, and Deterio- ration in Personal Life) of the CBI (S. Lee et al., 2007) fit the data from a large sample of professional school counselors. The five-fac- tor model was chosen based on the results of the previous studies of the CBI (Carrola et al., 2012; S. Lee et al., 2007, 2010; Shin et al., 2013; Yagi et al., 2011).

In addition, given that professional school counselors work in different environments (schools vs. mental health facilities) and face different types of stressors compared to other samples of counselors, we compared the results of this sample of professional school counselors’ scores on the CBI with a diverse sample of counselors from various specializa- tions and a sample of sexual offender and sur- vivor counselors. This is consistent with Maslach’s (2005) assertion that burnout mea- sures should be more comprehensive and take into account both an individual’s internal and environmental/organizational factors. Based on a review of the professional school coun- selor literature, it was hypothesized that school counselors would have similar levels of Incompetence, Negative Work Environ- ment, and Deterioration of Personal Life when compared to both a diverse sample of counsel- ors from various specializations as well as a sample of sexual offender and survivor coun- selors. Professional school counselors would have lower levels of Incompetence and Deval- uing Client compared to the other groups of counselors.

Methodology

Participants and Procedure

A total of 272 professional school counselors completed the survey. Three cases showed extreme answering patterns in that the same answers were given for each item (e.g., all 5s on all 20 items). Because there was no clear evidence as to why the ceiling effect for these three cases occurred, the cases were removed (Clark & Watson, 1995; Crocker & Algina, 1986).

Participants’ ages ranged from 25 to 67 years (M = 43.86 years, SD = 11.44). A total of 31 participants (11.5%) identified as male and 238 identified as female (88.5%), which was comparable to the gender breakdown (14% and 86%, respectively) of the ASCA membership (ASCA, personal communica- tion, October 8, 2013). In all, 251 (93.3%) participants identified as Caucasian, with the remaining participants representing African American (n = 10), American Indian/Alaskan Native American (n = 4), Hispanic American (n = 1), multiracial (n = 2), and declined to answer (n = 1). While the sample gender dis- tribution was similar to the ASCA, informa- tion regarding ethnicity and age were not available from ASCA. A total of 250 respon- dents (92.9%) completed a master’s degree in school counseling, 9 completed either a PhD or EdD (3.3%), 9 completed an EdS degree (3.3%), and 1 individual declined to answer. Years of experience as a school counselor ranged from their 1st year to 39 years (M = 10.86 years, SD = 7.61) with caseloads of stu- dents ranging from 30 to 1800 (M = 436.86, SD = 231.46).

A recruitment e-mail was sent to several school counselor listservs and investigator personal contacts. Thus, a convenience sam- ple was implemented. Participants were pro- vided a hyperlink, which included the informed consent and the measures. After agreeing to participate in the study, partici- pants completed a demographic questionnaire followed by the CBI. University institutional review board approval was obtained for this study.

180 Measurement and Evaluation in Counseling and Development 48(3)

Measures

Demographics Questionnaire. The demograph- ics questionnaire consisted of items related to age, gender, race/ethnicity, education, years of experience as a school counselor, caseload of students, school location, type of school counselor, salary, amount of time spent in counseling- and noncounseling–related duties each week, and the type and frequency of wellness activities in which the school coun- selor was engaged.

Counselor Burnout Inventory. The CBI (S. Lee et al., 2007) is a 20-item inventory designed to measure counselor burnout through a five- factor structure. Participants respond using a 5-point Likert-type scale ranging from 1 = never true to 5 = always true. The subscales are Negative Work Environment, Devaluing Client, Incompetence, Deterioration in Per- sonal Life, and Exhaustion. Exhaustion reflects the physical and emotional exhaustion due to the counselor’s job duties and a sample item is “Due to my job as a counselor I feel tired most of the time.” Negative Work Envi- ronment assesses the work environment stress beyond personal and interpersonal issues; a sample item is “I feel frustrated with the sys- tem in my workplace.” The Devaluing Client subscale focuses specifically on a counselor’s inability to form an emotional connection with clients. A sample item is “I am no longer concerned about the welfare of my clients.” Incompetence assesses the internal feelings of the counselor’s incompetence, and a sample item is “I feel I am an incompetent counselor.” Last, the Deterioration in Personal Life is assessed as the counselor’s deterioration for personal life. A sample item is “I feel I do not have enough time to spend with my friends.”

Concurrent validity of the CBI subscale scores has been demonstrated through corre- lations with other measures of burnout (S. Lee et al., 2007) and job stress (Wallace, Lee, & Lee, 2010). Discriminant validity has been demonstrated through significant negative relationships with various constructs includ- ing wellness (Puig et al., 2012), and flourish- ing and the working alliance (O’Sullivan &

Bates, 2014), and self-esteem (S. Lee et al., 2007). Evidence regarding the five-factor structure of the CBI as well as the need to determine whether the structure fit for profes- sional school counselors was presented in the introduction of this study. In regards to the reliability of the CBI scores, S. Lee et al. (2007) reported Cronbach’s coefficients alphas (i.e., internal consistency reliability) ranging between .73 and .85 for the subscales (see Table 4 for the alphas on specific sub- scales). Six-week test–retest reliability ranged from .72 to .85 for the subscale scores.

Data Analysis

A CFA was completed to confirm the factor structure and model fit using Mplus (Version 7.11; Muthén & Muthén, 1998–2013). CFA was chosen because there was consistent the- ory and research supporting the factor struc- ture of the CBI (S. Lee et al., 2007). Indices, including Satorra–Bentler scaled chi-square ( χsb

2 ), root-mean-square error of approxima-

tion (RMSEA), standardized root-mean square residual (SRMR), and the comparative fit index (CFI) were used to examine the model fit between the population covariance matrix Σ and the sample covariance matrix S (Schumacker & Lomax, 2010). Because there is no consensus regarding the best measure of fit (Bollen, 1990), using multiple fit indices is the best approach.

In the χ2 test, the differences between the observed and implied variance–covariance matrices are examined (Schumacker & Lomax, 2010). A nonsignificant χ2 is ideal as it indicates that the sample data and the theo- retical model are similar (Schumacker & Lomax, 2010). The RMSEA is another mea- sure of fit and should be .05 or lower indicat- ing that the sample data fit the model well. The SRMR has a range from 0 to 1 and values of less than .05 are desired (Hu & Bentler, 1999; Schumacker & Lomax, 2010). Finally, an index of model comparison was included. The CFI compares the existing model fit with a null model that assumes the latent variables in the model are uncorrelated (Bentler, 1990). The CFI varies from zero to one, with values

Gnilka et al. 181

of .95 (or greater) considered a good fit (Bentler, 1990; Schumacker & Lomax, 2010).

Results

Descriptive Statistics

The correlations between all items, and items means and standard deviations for the CBI are presented in Table 1. Item 14 had the smallest mean (M

14 = 1.16, SD = 0.36), and Item 1 had

the largest mean (M 1 = 3.25, SD = 0.84).

Means and standard deviations on individual items were similar to S. Lee et al. (2010).

Model Identification and Assumption Checking

Before conducting the analyses, it is impor- tant to identify the model. Model identifica- tion involves determining if there is enough information to estimate values for the unknown parameters (Schumacker & Lomax, 2010). In this study, the number of distinct values in the matrix S was 210. There were 50 free parameters in the matrix S, including 20 factor loadings, 20 measurement error vari- ances, 10 correlations among the latent vari- ables, and zero measurement error covariance terms. Thus, there was more than enough information in the matrix S to estimate param- eters (i.e., 50 < 210). Therefore, the number of free parameters to be estimated is less than or equal to the number of distinct values in the matrix S, which indicated that the model was overidentified (Schumacker & Lomax, 2010).

Since all 20 items were on a 5-point Likert- type scale, the covariance matrix was used (Schumacker & Lomax, 2010). The assump- tion of normality was examined using the Kolmogorov–Smirnov test and was not met (p < .05). However, after visual inspection of histograms and normalized Q-Q plots of all items, 3 out of 20 items were positively skewed (i.e., Items 4, 14, and 19), and the other items were approximately normally dis- tributed. Thus, the maximum likelihood parameter estimates with standard errors and the Satorra–Bentler scaled χ2 was used due to the moderate level of nonnormality of the data

(Lei & Lomax, 2005; Schumacker & Lomax, 2010; Yuan & Bentler, 1997).

Initial Model

The results indicated that all 20 factor load- ings between the observed variables and the latent variables were significant (p < .05 for all) and can be viewed in Figure 1. The lowest factor loading was between Incompetence and Item 12 (.55). The highest factor loading was between Negative Work Environment and Item 18 (.90). Eighty-two percent of the vari- ance in Item 18 was explained by Negative Work Environment, and 31% of the variance in Item 12 was explained by Incompetence. All correlations between each pair of latent variables were significant (p < .05 for all) ranging from .23 to .64. All standardized fac- tor loadings are presented in Table 2.

The results of model fit indices indicated that the RMSEA was not an acceptable level of model fit (RMSEA = .06). The SRMR was also slightly above the acceptable level of model fit (SRMR = .057). Additionally, the Satorra–Bentler scaled chi-square test was significant ( χSB

2 332 67= . , df = 160, p < .001), indicating that the specified CFA model was not supported by the covariance data. The CFI was .933, which is below the suggested .95 threshold of good model fit. Overall, most of the indices suggested a poor model fit. A com- parison of all model fit indices is presented in Table 3.

First Modification

Modification indices (MIs) suggested adding an error covariance between Item 15 and Item 10 for the greatest decrease in χ2 (i.e., 40.08). By further examining these two items, Item 15 (i.e., “I feel I do not have enough time to spend with my friends.”) and Item 10 (i.e., “I feel like I do not have enough time to engage in personal interests.”), it was determined that they measured similar constructs (i.e., both items were associated with the same latent variable), Deterioration in Personal Life. This was also consistent with two previous studies using different samples (S. Lee et al., 2010;

182

T a b

le 1

. C

o rr

el at

io ns

B et

w ee

n th

e It

em s,

M ea

ns (

M ), a

nd S

ta nd

ar d

D ev

ia ti o ns

( SD

) o f th

e C

o un

se lo

r B

ur no

ut I nv

en to

ry (

N =

2 69

).

It em

1 2

3 4

5 6

7 8

9 10

11 12

13 14

15 16

17 18

19 20

1 —

2

.3 4*

—

3 .3

8* .3

1* —

4

.1 0

.3 6*

.1 3*

—

5 .4

6* .3

5* .4

1* .2

0* —

6

.7 5*

.3 6*

.4 1*

.1 4*

.4 5*

—

7 .3

9* .5

5* .3

2* .2

8* .3

4* .3

9* —

8

.2 7*

.2 2*

.6 1*

.1 3*

.3 4*

.2 5*

.2 2*

—

9 .2

0* .3

1* .2

5* .4

1* .2

5* .2

0* .3

3* .2

5* —

10

.3 5*

.2 4*

.3 5*

.1 8*

.4 9*

.3 9*

.2 5*

.3 0*

.2 2*

—

11 .6

1* .3

9* .4

2* .1

5* .4

7* .6

6* .4

8* .3

0* .2

7* .5

1* —

12

.1 8*

.4 8*

.1 1*

.3 4*

.2 3*

.1 7*

.3 8*

.1 2*

.2 6*

.1 7*

.2 4*

—

13 .4

4* .3

7* .5

2* .0

9* .4

1* .4

8* .4

1* .4

9* .2

5* .3

9* .4

9* .2

6* —

14

.1 1*

.3 6*

.1 7*

.5 1*

.1 6*

.1 2*

.2 2*

.1 1

.4 6*

.0 9

.1 4*

.3 7*

.1 1*

—

15 .3

0* .2

5* .2

4* .1

2* .4

8* .3

6* .1

9* .2

2* .1

5* .7

1* .4

2* .2

3* .3

1* .1

1* —

16

.5 6*

.2 5*

.3 0*

.0 4*

.4 2*

.5 1*

.3 3*

.2 7*

.1 7*

.3 7*

.5 5*

.0 6

.4 0*

.0 4

.3 4*

—

17 .2

5* .4

5* .2

7* .2

9* .3

1* .3

0* .5

5* .2

8* .3

6* .2

5* .2

9* .3

5* .3

9* .3

3* .2

7* .2

7* —

18

.3 7*

.2 8*

.6 4*

.1 0*

.3 8*

.4 2*

.3 8*

.5 6*

.2 2*

.4 0*

.4 5*

.1 1*

.7 9*

.0 7*

.3 0*

.3 7*

.3 1*

—

19 .1

4 .2

7* .1

2 .4

7* .2

2* .1

6* .2

2* .0

9 .4

2* .1

5* .1

5* .2

4* .1

4* .5

3* .1

2* .0

5 .2

6* .1

1 —

20

.3 4*

.2 5*

.2 5*

.0 7*

.5 0*

.3 2*

.2 4*

.2 4*

.0 8

.4 4*

.3 3*

.2 0*

.2 9*

.1 3*

.4 6*

.2 8*

.2 9*

.2 4*

.1 2

— M

3. 25

1. 98

2. 27

1. 34

1. 98

3. 02

2. 69

2. 26

1. 54

2. 81

3. 02

2. 06

3. 12

1. 32

2. 64

2. 99

2. 44

3. 07

1. 16

1. 92

SD 0.

84 0.

80 1.

04 0.

59 0.

88 0.

97 0.

86 1.

08 0.

64 1.

11 0.

95 0.

94 1.

09 0.

53 1.

11 1.

21 0.

83 1.

13 0.

36 0.

97

*p <

.0 5.

Gnilka et al. 183

Yagi et al., 2011). Therefore, an error covari- ance was added between Item 15 and Item 10. Standardized residuals were also consulted

for potential model modification. After com- pleting the modification of adding an error covariance between Item 15 and Item 10 and

Figure 1. The initial confirmatory factor analysis model tested with the Counselor Burnout Inventory (CBI). The standardized parameter estimates for the CBI are listed. Rectangles indicate the 20 items on the CBI, and ovals represent the 5 latent factors of the subscales.

184 Measurement and Evaluation in Counseling and Development 48(3)

rerunning the model, the error covariance between the two items was significant (p < .05).

In addition to the error covariance being significant, the results indicated that the model with the modification was a better fit than the original model, Δχ2(1) = 25.98, p < .001. In addition, the RMSEA decreased to .053, the SRMR decreased to .051, and the CFI increased slightly to .948. In conclusion, the modification improved the model fit.

Second Modification

Additionally, based on the first modification mentioned above, MIs and standardized residuals were consulted. Adding an error covariance between Item 8 and Item 3 was suggested with the potential outcome of decreasing χ2by 28.09. By further examining these two items, Item 8 (i.e., “I feel negative

energy from my supervisor.”) and Item 3 (i.e., “I am treated unfairly in my workplace.”), it was found that they were associated with the same latent variable, Negative Work Envi- ronment. While this has not been found in previous CFA studies, Carrola et al. (2012) found that these two items lacked invariance between a sample of Korean and U.S. sam- ples. Therefore, an error covariance was added between Item 8 and Item 3. Standard- ized residuals were also consulted for poten- tial model modification. After completing the modification and rerunning the model, the error covariance between the two items was determined to be significant (p < .05). The results indicated that the model with this sec- ond modification had a significantly improved fit, Δχ2(1) = 20.98, p < .001. In addition, the RMSEA decreased to .047, the SRMR decreased to .050, and the CFI increased slightly to .960.

Table 2. Standardized Factor Loadings for the Counselor Burnout Inventory.

Original Model 1st Modification 2nd Modification 3rd Modification 4th Modification

Item LV β LV β LV β LV β LV β

1 Ex .83 Ex .83 Ex .83 Ex .75 Ex .75 2 In .74 In .74 In .74 In .74 In .74 3 NWE .70 NWE .70 NWE .67 NWE .67 NWE .74 4 DC .69 DC .69 DC .69 DC .69 DC .69 5 DPL .67 DPL .78 DPL .78 DPL .77 DPL .77 6 Ex .85 Ex .85 Ex .85 Ex .78 Ex .78 7 In .75 In .75 In .75 In .75 In .75 8 NWE .64 NWE .64 NWE .60 NWE .60 NWE .59 9 DC .62 DC .62 DC .62 DC .62 DC .62

10 DPL .81 DPL .67 DPL .67 DPL .68 DPL .68 11 Ex .79 Ex .79 Ex .79 Ex .83 Ex .83 12 In .55 In .56 In .56 In .55 In .56 13 NWE .85 NWE .85 NWE .86 NWE .86 NWE .90 14 DC .75 DC .75 DC .75 DC .75 DC .75 15 DPL .79 DPL .64 DPL .64 DPL .64 DPL .64 16 Ex .66 Ex .66 Ex .66 Ex .67 Ex .67 17 In .68 In .68 In .68 In .68 In .68 18 NWE .90 NWE .90 NWE .91 NWE .91 NWE .87 19 DC .68 DC .68 DC .68 DC .68 DC .68 20 DPL .60 DPL .63 DC .63 DPL .63 DPL .63

Note. LV = five latent variables (i.e., Ex = Exhaustion, In = Incompetence, NWE = Negative Work Environment, DC = Devaluing Client, and DPL = Deterioration in Personal Life). β = standardized factor loadings. Original model = the model proposed by S. Lee et al. (2007); 1st modification = error covariance between Items 15 and 10; 2nd modification = error covariance added between Items 8 and 3; 3rd modification = error covariance added between Items 6 and 11; 4th modification = error covariance added between Items 13 and 3.

Gnilka et al. 185

Third Modification

After the above second modification and rerun- ning the model, MIs and standardized residuals were consulted for additional improvement. A decrease in χ2by 22.62 was possible by adding an error covariance between Item 6 and Item 1. This was consistent with a previous study of a sample of sexual offender and sexual abuse therapists (J. Lee et al., 2010). On further examination of the two items, Item 6 (i.e., “I feel exhausted due to my work as a counselor.”) and Item 1 (i.e., “Due to my job as a counselor, I feel tired most of the time.”), it was deter- mined that both items were associated with the same latent variable, Exhaustion. Therefore, an error covariance was added between Item 6 and Item 1. After including the modification and rerunning the model, the error covariance between the two items was significant (p < .05). The results indicated that the model with the modification was a better fit than the origi- nal model, Δχ2(1) = 18.91, p < .001. In addi- tion, the RMSEA decreased to .041, the SRMR decreased to .048, and the CFI increased slightly to .968.

Fourth Modification

After rerunning the model with the above- mentioned modifications, MIs and standard- ized residuals were once again consulted. Another reduction in χ2of 11.90 was possible

by inserting an error covariance between Item 13 and Item 3. A review of the content of these two items, Item 13 (i.e., “I feel bogged down by the system in my workplace.”) and Item 3 (i.e., “I am treated unfairly in my work- place.”), showed that these items measured similar constructs and were part of the same latent variable, Negative Work Environment. An error covariance was added between Item 13 and Item 3, and the model was rerun. The error covariance was found to be significant (p < .05), and the modified model is presented in Figure 2. The results indicated that the model with this final modification had an improved fit, Δχ2(1) = 10.22, p < .01. In addi- tion, the RMSEA decreased to .038, the SRMR decreased to .046, and the CFI increased slightly to .974. In conclusion, the final modification improved the model fit. Although there were several additional modi- fication recommendations, most required changes to the factor structure (i.e., which was not supported by theory/the literature) or would not result in a significant change in χ2; therefore, no additional changes were made.

Descriptive Statistics and Effect Sizes

Table 4 provides descriptive data including means and standard deviations of each of the five factors; in addition, effect sizes were

Table 3. Major Indices of Model Testing and Modifications.

Model SBχ2 df SBΔχ2 RMSEA SRMR CFI

Five factor (Model 1) 332.67 160 .060 .057 .933 Modification 1 (Model 2) 293.92 159 .053 .051 .948 Difference between Models 1 and 2 25.98* Modification 2 (Model 3) 264.41 158 .047 .050 .960 Difference between Models 2 and 3 20.98* Modification 3 (Model 4) 242.64 157 .041 .048 .968 Difference between Models 3 and 4 18.91* Modification 4 (Model 5) 226.59 156 .038 .046 .974 Difference between Models 4 and 5 10.22**

Note. SBχ2 = Satorra–Bentler scaled chi-square; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CFI = comparative fit index. Model 2 = error covariance added between Items 15 and 10; Model 3 = error covariance added between Items 8 and 3. Model 4 = error covariance added between items 6 and 11; Model 5 = error covariance added between Items 13 and 3. *p < .01. **p < .01.

186 Measurement and Evaluation in Counseling and Development 48(3)

calculated to compare the mean scores of the CBI subscales between this study’s sample (i.e., professional school counselors) and two other samples of counselors: (a) a diverse sample of counselors from various specializa- tions (N = 132; S. Lee et al., 2007) and (b)

counselors working with sexual offender and sexual abuse clients (N = 204; J. Lee et al., 2010). As noted by Cohen (1992), effect sizes (d) were designated as small (.20), medium (.50), and large (.80). Effect size differences were noted between professional school counselors and a

Figure 2. The fourth modification confirmatory factor analysis model tested with the Counselor Burnout Inventory (CBI). The standardized parameter estimates for the CBI are listed. Error covariances were added between Items 15 and 10, Items 8 and 3, Items 6 and 11, and Items 13 and 3. Rectangles indicate the 20 items on the CBI, and ovals represent the 5 latent factors of the subscales.

Gnilka et al. 187

diverse sample of counselors from various specializations. For example, professional school counselors had higher levels of Exhaustion, Negative Work Environment, and Deterioration in Personal Life (d = .26, .09, and .03, respectively); however, only Exhaus- tion had an effect size classified between small and medium, while the other two did not even meet the small effect size cutoff. In addi- tion, professional school counselors had lower mean levels of Incompetence and Devaluing Client (d = −.12 and −.50, respectively). Devaluing Client was classified as a medium effect size, while Incompetence failed to reach the small effect size cutoff.

Next, effect size differences were found between professional school counselors and counselors working with sexual offender and sexual abuse clients. Specifically, profes- sional school counselors had higher levels of Exhaustion and Negative Work Environment (d = .27 and .14, respectively); however, only Exhaustion had an effect size classified between small and medium. Professional school counselors also had lower mean levels of Incompetence, Devaluing Client, and Dete-

rioration in Personal Life (d = −.19, −.82, and −.23, respectively). The Devaluing Client was classified as a large effect size while the Dete- rioration in Personal Life was classified between a small and medium effect size.

Discussion

This study sought to determine whether the five-factor structure of the CBI was an appro- priate model fit for professional school coun- selors. There was evidence that the original five-factor model of the CBI was a marginally appropriate fit due to three of the four indices being close to the suggested cutoffs. The five- factor structure included the following five factors: Exhaustion, Incompetence, Negative Work Environment, Devaluing Client, and Deterioration in Personal Life. Exhaustion refers to the emotional and physical fatigue caused by various issues and types of clients a counselor may work with. Incompetence reflects a counselor’s internal belief in his or her ability to perform his or her work as a counselor. Negative Work Environment focuses on a counselor’s attitudes and beliefs

Table 4. Means, Standard Deviations, Internal Consistencies, and Effect Sizes of the Five Factors Across Samples.

Study Sample Ex In NWE DC DPL

1. Current study: Professional school counselors M 3.03 2.23 2.64 1.31 2.31 SD 0.83 0.65 0.90 0.38 0.82 α .85 .78 .85 .78 .82 2. S. Lee et al. (2007): U.S. counselors M 2.81 2.30 2.56 1.53 2.29 SD 0.88 0.56 0.83 0.49 0.72 α .85 .73 .83 .80 .78 3. J. Lee et al. (2010): Sexual offender and sexual abuse therapists M 2.82 2.35 2.51 1.69 2.48 SD 0.72 0.60 0.96 0.53 0.67 α .80 .73 .88 .78 .73 Effect size: 1 vs. 2 Cohen’s d .26 −.12 .09 −.50 .03 Effect size 1 vs. 3 Cohen’s d .27 −.19 .14 −.82 −.23

Note. Ex = Exhaustion, In = Incompetence, NWE = Negative Work Environment, DC = Devaluing Client, and DPL = Deterioration in Personal Life.

188 Measurement and Evaluation in Counseling and Development 48(3)

toward the work environment. The Devaluing Client factor focuses specifically on a coun- selor’s lack of emotional connection and empathy toward clients. Lastly, Deterioration in Personal Life focuses on the impact work may have on a counselor’s personal life including social relationships and hobbies.

While the five-factor structure was accept- able for the sample based on fit indices such as the RMSEA, SRMR, and CFI, modifica- tions were needed to significantly improve the fit. As a result, within-factor error covariance for pairs of items (i.e., 10 and 15, 8 and 3, 6 and 1, and 13 and 3) were included resulting in a significantly better fitting model. In other words, the five-factor model was the most appropriate fit for the professional school counselor data set, with the addition of a few error covariances between items. It was inter- esting that each of these modifications were within-factor, suggesting that the overall five- factor structure of the CBI was a good model for the professional school counselor popula- tion.

Results of this study were consistent with previous research which found and confirmed a modified five-factor burnout structure (Car- rola et al., 2012; S. Lee et al., 2007, 2010; Shin et al., 2013). For example, S. Lee et al. (2010) added error covariances between two similar pairs of items found in this study (i.e., 10 and 15, and 6 and 1) and Yagi et al. (2011) also added an error covariance between Items 10 and 15. Lastly, Carrola et al. (2012) noted Items 3 and 8 were invariant between a sam- ple of U.S. and Korean professional school counselors, which was consistent with the results of this study. One additional item pair (8 and 3), which had not been found in previ- ous research on the CBI, was also found in this study. Bentler and Chou (1987) suggested that forcing all error terms to be uncorrelated is rarely appropriate with these kinds of data. Taken together, future researchers should investigate these item pairs closely in addi- tional populations and consider revising redundant items in the CBI.

For example, both Item 13 (i.e., “I feel bogged down by the system in my work- place.”) and Item 3 (i.e., “I am treated unfairly

in my workplace.”) fall under the factor of Negative Work Environment. Both items reflect feeling pressure or stress in the work- place. Perhaps collapsing these two items into one item would be helpful; however, the cor- relation between the two items is only .52 and the items seem to be worded differently.

Under the factor Exhaustion, Items 6 (i.e., “I feel exhausted due to my work as a coun- selor.”) and Item 1 (i.e., “Due to my job as a counselor, I feel tired most of the time.”) reflect feeling tired or exhausted because of work and had a correlation of .75. Consistent with the earlier recommendation for Items 13 and 3, future researchers may want to con- sider collapsing these items into one single item (e.g., “I feel tired most of the time due to my job as a counselor”).

Another suggested change would be look- ing more closely into Item 8 (i.e., “I feel nega- tive energy from my supervisor”) and Item 3 (i.e., “I am treated unfairly in my work- place.”), which have a correlation of .61 and both fall under the Negative Work Environ- ment Factor. Both items reflect negativity with regard to either a person (supervisor) or the environment (workplace). While these seem conceptually different, Sterner (2009) concluded that the working alliance formed with a supervisor had a significant negative relationship (r = −.56) with perceptions of work-related stress. Future researchers should look more closely into rewording these items to more clearly delineate supervision from other workplace variables. If one of the two items should be dropped, Item 8 seems more appropriate given that it could be included under Item 3 (i.e., the workplace overall).

Finally, Item 15 (i.e., “I feel I do not have enough time to spend with my friends.”) and Item 10 (i.e., “I feel like I do not have enough time to engage in personal interests.”), which fall under the Deterioration in Personal Life factor, have a correlation of .71. These items reflect not having enough time for friends or personal interests. One possibility is that the social support outside of work is considered a large component of personal interests. Item 15 seems the more appropriate item to drop given that the content of the item (time with friends)

Gnilka et al. 189

could be subsumed by Item 10 (i.e., personal interests).

Previous studies have suggested that school counselors do experience burnout (e.g., Moyer, 2011; Wilkerson & Bellini, 2006), which has led to increased awareness of the importance of wellness within the coun- seling profession. Several factors contribute to an individual’s burnout level. Understand- ing contributions to burnout in individual school counselors can provide potential inter- ventions to levels of burnout and increase lev- els of job satisfaction. When an individual recognizes increased levels of stress on the job, understanding whether the individual is experiencing burnout, the extent of burnout, and in which dimensions, interventions can be tailored to best suit the individual’s wellness needs. One way to compare how burnout may be experienced between different types of practicing counselors is to compare the mean scores of the five factors of the CBI, thus war- ranting evidence of the factor structure from previous studies in the professional school counselor population.

Through mean comparison and effect size calculations, professional school counselors had higher scores on the Exhaustion and Neg- ative Work Environment factors of the CBI compared with a diverse mix of professional counselors (S. Lee et al., 2007) and counsel- ors working with sexual offenders and survi- vors (S. Lee et al., 2010). Only the effect size for the Exhaustion factor was classified between small and medium suggesting that school counselors feel chronic emotional and physical exhaustion. This result is consistent with the burnout and stress literature about school counselors including increased expo- sure to organizational stressors (Wilkerson & Bellini, 2006), a lack of professional supervi- sion (Moyer, 2011), a high level of nonguid- ance responsibilities (Coll & Freeman, 1997; Moyer, 2011), and increased stress when fac- ing acts of school violence (Paine, 2009). This suggests that one of the key ways of reducing professional school counselors from burnout may be larger schoolwide and districtwide interventions that assist in reducing school counselor–student ratios, increase ability for

supervision and continuing education oppor- tunities, and reductions in nonguidance activi- ties.

School counselors also reported lower mean levels of Devaluing Client and Incom- petence than both a diverse mix of profes- sional counselors (S. Lee et al., 2007) and counselors working with sexual offenders and survivors (S. Lee et al., 2010). The effect sizes for Devaluing Client was classified as large suggesting that school counselors are likely to maintain high levels of empathy and emo- tional warmth toward their students and are less likely to act out their frustrations by devaluing clients. This is an interesting find- ing that suggests that most professional school counselors do not burnout or act out their frus- trations due to client related issues; rather, organizational and overwork seem to be larger factors in overall burnout.

Finally, since the present study determined the modified five-factor CBI was a good fit for a sample of professional school counsel- ors, the CBI may be an appropriate tool to use in the supervision setting. The CBI could be administered to determine levels of burnout overall and among the five factors within the supervisee. This would provide the supervisee with an overall understanding of various areas of burnout and could offer insight into well- ness and modifications in their duties to lessen symptoms of burnout. Moyer (2011) dis- cussed the positive role that supervision may have with school counselors because those receiving supervision reported lower burnout scores compared to school counselors who were not receiving supervision. Completing the CBI may be another step in supervision to further identify higher levels of burnout in a school counselor before an individual becomes impaired in his or her job duties.

Limitations and Future Research

There are several limitations to this study. First, convenience sampling was completed using several school counseling listservs. This limits generalizability to school counselors employed abroad; in addition, there may be unequal representation from professionals

190 Measurement and Evaluation in Counseling and Development 48(3)

who are members of the Listserv versus those who are not members (i.e., listserv members may generally may have more experience or a vested interest in the profession than non- members). While the sample gender distribu- tion was similar to ASCA membership, information regarding ethnicity and age were not available from ASCA. Therefore, it is not possible to ensure the sample would mirror ASCA membership patterns on these addi- tional variables. Future researchers should consider obtaining randomized samples from professional school counselor associations on the national and state levels as well as obtain more ethnically diverse samples of school counselors to see if similar or different factor structures are found.

Future research confirming the factor struc- ture within other counseling specializations should also be completed. For example, there are multiple specializations of professional counselors that could experience burnout dif- ferently from other groups (e.g., college coun- selors, faith-based counseling agencies). This may give insight how burnout impacts each counseling specialization. In addition, given that very few participants endorsed Likert- type values of 4 or 5, which was similar to S. Lee et al. (2010), future researchers may want to investigate this using Rasch analysis. Lastly, future research of the impact of internal and external factors of burnout within the environ- ment and counseling specialties would provide insight into wellness strategies to safeguard again professional burnout in within the indi- vidual’s personal and environmental factors.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

American School Counselor Association. (2012). The ASCA national model: A framework

for school counseling programs (3rd ed.). Alexandria, VA: Author.

Barnes, K. L., & Moon, S. M. (2006). Factor struc- ture of the psychotherapy supervisor develop- ment scale. Measurement and Evaluation in Counseling and Development, 39, 130–140.

Bentler, P. M. (1990). Comparative indexes in structural models. Psychological Bulletin, 107, 238–246.

Bentler, P. M., & Chou, C. (1987). Practical issues in structural modeling. Sociological Methods & Research, 16, 78–117.

Bollen, K. (1990). Overall fit in covariance struc- ture models: Two types of sample size effects. Psychological Bulletin, 107, 256–259.

Butler, S. K., & Constantine, M. G. (2005). Collective self-esteem and burnout in profes- sional school counselors. Professional School Counseling, 9, 55–62.

Carrola, P., Yu, K., Sass, D., & Lee, S. (2012). Measurement invariance of the Counselor Burnout Inventory across cultures: A com- parison of U.S. and Korean counselors. Measurement and Evaluation in Counseling and Development, 45, 227–244.

Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in objective scale devel- opment. Psychological Assessment, 7, 309–319.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.

Coll, K. M., & Freeman, B. (1997). Role con- flict among elementary school counselors: A national comparison with middle and second- ary school counselors. Professional School Counseling, 3, 251–261.

Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory. Belmont, CA: Wadsworth Group/Thomson Learning.

Farmer, E. Z., Burns, B. J., Phillips, S. D., Angold, A., & Costello, E. J. (2003). Pathways into and through mental health services for children and adolescents. Psychiatric Services, 54, 60–66.

Freudenberger, H. J. (1974). Staff burnout. Journal of Social Issues, 30, 159–165.

Hu, L., & Bentler, P. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.

Kesler, K. D. (1990). Burnout: A multimodal approach to assessment and resolution. Elementary School Guidance and Counseling, 24, 303–311.

Lambie, G. W. (2007). The contribution of ego development level to burnout in school counselors: Implications for professional

Gnilka et al. 191

school counseling. Journal of Counseling & Development, 85, 82–88.

Lee, J., Wallace, S., Puig, A., Choi, B. Y., Nam, S. K., & Lee, S. M. (2010). Factor structure of the Counselor Burnout Inventory in a sample of sexual offender and sexual abuse therapists. Measurement and Evaluation in Counseling and Development, 43, 16–30.

Lee, S., Baker, C., Cho, S., Heckathorn, D., Holland, M., Newgent, R., . . .Yu, K. (2007). Development and initial psychomet- rics of the Counselor Burnout Inventory. Measurement and Evaluation in Counseling and Development, 40, 142–154.

Lee, S., Cho, S., Kissinger, D., & Ogle, N. (2010). A typology of burnout in professional coun- selors. Journal of Counseling & Development, 88, 131–138.

Lei, M., & Lomax, R. G. (2005). The effect of varying degrees of nonnormality in struc- tural equation modeling. Structural Equation Modeling, 12, 1–27.

Maslach, C. (2005). Understanding burnout: Work and family issues. In D. Halpem, S. Murphy, & S. Elaine (Eds.), Work-family balance to workfamily interaction: Changing the meta- phor (pp. 99–114). Mahwah, NJ: Lawrence Erlbaum.

Maslach, C., & Jackson, S. E. (1981). The mea- surement of experienced burnout. Journal of Occupational Behavior, 2, 99–113.

Moyer, M. (2011). Effects of non-guidance activi- ties, supervision, and student-to-counselor ratios on school counselor burnout. Journal of School Counseling, 9(5). Retrieved from http://jsc.montana.edu/articles/v9n5.pdf

Muthén, L., & Muthén, B. (1998–2013). Mplus user’s guide (7th ed.). Los Angeles, CA: Muthén & Muthén.

O’Sullivan, D., & Bates, J. K. (2014). The relation- ship among personal and work experiences: Implications for rehabilitation counselor well- being and service provision. Rehabilitation Research, Policy, and Education, 28, 45–60.

Paine, C. K. (2009). School crisis aftermath: Care for the caregivers. Principal Leadership, 10, 12–16. Retrieved from http://www.principals.org/

Puig, A., Baggs, A., Mixon, K., Park, Y. M., Kim, B. Y., & Lee, S. M. (2012). Relationship between job burnout and personal wellness in mental health professionals. Journal of Employment Counseling, 49, 98–109.

Schumacker, R. E., & Lomax, R. G. (2010). A beginner’s guide to structural equation mod- eling (3rd ed.). New York, NY: Routledge.

Shin, H., Yuen, M., Lee, J., & Lee, S. (2013). Cross-cultural validation of the Counselor Burnout Inventory in Hong Kong. Journal of Employment Counseling, 50, 14–25.

Sterner, W. R. (2009). Influence of the supervisor working alliance on supervisee work satisfac- tion and work-related stress. Journal of Mental Health Counseling, 31, 249–263.

Wachter, C. A., Clemens, E. V., & Lewis, T. F. (2008). Exploring school counselor burnout and school counselor involvement of par- ents and administrators through an Adlerian theoretical framework. Journal of Individual Psychology, 64, 432–449.

Wallace, S. L., Lee, J., & Lee, S. M. (2010). Job stress, coping strategies, and burnout among abuse-specific counselors. Journal of Employment Counseling, 47, 111–122.

Warnath, C. F., & Shelton, J. L. (1976). The ulti- mate disappointment: The burned-out coun- selor. Personnel and Guidance Journal, 55, 172–175.

Wilkerson, K. (2009). An examination of burnout among school counselors guided by the stress- strain-coping theory. Journal of Counseling & Development, 87, 428–437.

Wilkerson, K., & Bellini, J. (2006). Intrapersonal and organizational factors associated with burnout among school counselors. Journal of Counseling & Development, 84, 440–450.

Yagi, D., Lee, J., Puig, A., & Lee, S. (2011). Cross-cultural validation of factor structure in the Japanese Counselor Burnout Inventory. Japanese Psychological Research, 53, 281– 291.

Yuan, K. H., & Bentler, P. M. (1997). Mean and covariance structure analysis: Theoretical and practical improvements. Journal of the American Statistical Association, 92, 767–774.

Author Biographies

Philip B. Gnilka, PhD, is an Assistant Professor in the Clinical Mental Health Counseling Program in the College of Education at DePaul University in Chicago.

Aryn C. Karpinski, PhD, is an Assistant Professor in the Evaluation and Measurement program in the College of Education, Health, and Human Services at Kent State University in Kent, Ohio.

Heather J. Smith is a doctoral candidate in the Coun- seling and Human Development Services Program in the College of Education, Health, and Human Ser- vices at Kent State University in Kent, Ohio.

Copyright of Measurement & Evaluation in Counseling & Development (Sage Publications Inc. ) is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

Copyright of Measurement & Evaluation in Counseling & Development is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.