A total of 1,177 Spanish nurses participated in this study. The mean age of the participants was 44.25 years (SD=8.45). Of the sample, 35.2% were male and 64.8% were female; 18.8% were single, 73.5% married, and 7.7% separated, divorced, or widowed. Of the subjects, 34% often had lengthy work shifts (24-hour on-duty service). In regards to hospital unit, 28% worked in the emergency room and intensive care; 18.9% were in surgical units; 23.9% were in patient care units; 21.8% were in primary healthcare; and 7.4% worked in the maternal and child healthcare services. In relation to work shift, 55.9% of the subjects had rotating shifts; 41.9% always worked in the morning; 0.7% always worked in the afternoon; and 1.5% always worked at night.

The Granada Burnout Questionnaire (GBQ) (De la Fuente et al., 2013) is a paper-and-pencil test that was created according to the guidelines in Downing (2006). It is based on the operative definition of burnout provided by Maslach et al. (2001) for the specification of the different dimensions.

The questions are answered on a five-point Likert scale, where 1 means total disagreement and 5 signifies total agreement. The items measure burnout dimensions either positively or negatively. A high score signifies a high value in the feature evaluated. Accordingly, 11 items should be redirected when the correction is performed (items 1, 4, 10, 12, 13, 19, 20, 23, 24, 25 and 26).

We also applied the MBI (Maslach & Jackson, 1981) in its version adapted to the Spanish population (Seisdedos, 1997) to examine the convergent validity of the GBQ. The MBI is composed of 22 items, which are answered on a seven-point Likert scale. Following the recommendations of meta-analytic studies (Aguayo, Vargas, De la Fuente, & Lozano, 2011), we calculated the reliability of each of the MBI dimensions in the sample. These dimensions are Emotional Exhaustion (9 items, α=.89), Depersonalization (5 items, α=.70), and Personal Accomplishment (8 items, α=.84).

The Neo Reduced Five-Factor personality inventory (NEO-FFI; Costa & McCrae, 1992) was administered in the version adapted to the Spanish population (Costa & McCrae, 2002) to evaluate concurrent validity. This inventory consists of 60 items answered on a five-point Likert scale with 12 items in each dimension. In this sample, Cronbach's alpha for each dimension was the following: Neuroticism (.79), Extroversion (.80), Openness (.70), Agreeableness (.75), and Responsibility (.83).

The Educational-Clinical Questionnaire: Anxiety and Depression (CECAD; Lozano, García-Cueto, & Lozano, 2007) was also used to assess concurrent validity. This questionnaire is composed of 50 items with a five-point Likert scale format. It provides an overall evaluation of emotional disorders based on the scores in the five dimensions. In our sample, its reliability was the following: Depression (26 items, α=.94), Anxiety (19 items, α=.92), Uselessness (9 items, α=.87), Irritability (7 items, α=.87), and Problematic Thoughts (8 items, α=.86).

Once legal permission was obtained, the guidelines of Campion and Miller (2006) were used to edit the data collection booklet.

This research is an instrumental study (Hartley, 2012; Montero & León, 2007). We contacted the Nurses’ Union (SATSE, for its initials in Spanish). A committee composed of the authors and members of the Provincial Delegation of the SATSE in Granada were in charge of coordinating data collection. This made it possible for nurses from other provinces in the region of Andalusia to also participate. The subjects were contacted individually. Those who voluntarily agreed to be part of the study anonymously filled out the questionnaires, outside of their work schedule.

In order to evaluate the evidence of validity based on the internal structure of the questionnaire, the factorial validity was verified by cross-validation. For this purpose, an Exploratory Factor Analysis (EFA) was performed with a Minimum Rank Factor Analysis (MRFA) and oblique factor rotation (PROMIN). The polychoric correlations between items were used as the input matrix, all of which were implemented in the program FACTOR v.9.2 (Lorenzo-Seva & Ferrando, 2006).

EFA was performed on approximately 40% of the sample. A Confirmatory Factor Analysis (CFA) was conducted on the remaining 60% by using the maximum likelihood for the estimation of parameters, standard error, and goodness of fit, and was corrected when it did not show a multivariate normality. This option was suitable in our case since the items had a Likert-scale response format with moderate asymmetry indices (< 2) and kurtosis (< 7) (Finney & Distefano, 2013; Rhemtulla, Brosseau-Liard, & Savalei, 2012) implemented in MPLUS 7.2 (Muthén & Muthén, 1998–2012). The goodness-of-fit criteria used were the following values: RMSEA ≤ .06, CFI ≥ .95, TLI ≥ .95, and SRMR ≤ .08 (Beauducel & Wittmann, 2005; Hu & Bentler, 1999; Yu & Muthén, 2002).

The evidence of convergent validity was evaluated by estimating the correlation of GBQ dimension scores with the MBI dimensions. Evidence of criterion-related validity was also studied through concurrent validity that analyzed the correlations between the GBQ dimensions with those of the NEO-FFI and CECAD. Attenuation formulas for measurement errors were used to correct all of the estimated correlations (American Educational Research Association, American Psychological Association, & National Council on Measurement in Education, 1999).

An analysis was then performed of the items in the dimensions of the GBQ. More specifically, the focus of the study was on the discrimination, mean, standard deviation, and standard error of the scores of the items in each dimension. Cronbach's alpha was also calculated. Analysis of variance (ANOVA) was then performed to ascertain whether there were statistically significant differences in the burnout dimension scores among different groups.

Finally, the scores of the questionnaire were transformed to a scale of a mean of 50 and a standard deviation of 10, following a normal distribution as well as their corresponding centile scale, both widely used in Health Sciences.

An Exploratory Factor Analysis was performed on 40% of the sample (n=462). The Minimum Rank Factor Analysis extraction method with Promin rotation was used to extract three factors that explained 73.21% of the total variance of the questionnaire (see Table 1). The number of factors extracted was determined by a Parallel Analysis (Timmerman & Lorenzo-Seva, 2011) as well as by the theoretical model supporting the questionnaire and the suitability of the three-dimensional model.

The correlation between Depersonalization and Personal Accomplishment was -.63, the correlation between Depersonalization and Emotional Exhaustion was .13, and the correlation between Personal Accomplishment and Emotional Exhaustion was -.48. The mean value of the residuals of the model was -0.0006 (SD=0.045), GFI [goodness-of-fit index]=.99, and RMSR=.045. This indicates a good fit of the three-dimensional model to the data.

The next step was a Confirmatory Factor Analysis of the remaining 60% of the sample (n=715). The model obtained in the previous Exploratory Factor Analysis was verified and the relation between factors was taken into account. Furthermore, guided by the modification indexes, the correlations between residuals were estimated. This is common in factorial solutions with related factors. However, for the sake of simplicity, these correlations are not shown in the graphical model in Figure 1.

Figure 1.

Confirmatory Factor Analysis of Granada Burnout Questionnaire (60% of the sample).

(0.37MB).

When the parameters of the model were calculated by the MLM calculation method implemented in MPLUS (Muthén & Muthén, 1998–2012), the result was χ2SB (257)=515.97, p<.001. Moreover, values of RMSEA=.038 (90%C.I.: .033, -.042) and p(RMSEA ≤ .05)=.99, CFI=.957, TLI=.945, SRMR=.061 were obtained. These values indicated the good fit of the model.

Because of the high correlation between Emotional Exhaustion and Personal Accomplishment as well as between Personal Accomplishment and Depersonalization, the corresponding two-dimensional models were checked with the chi-square difference test in comparison to the three-dimensional model. The result obtained was χ2diff (2) (PA+D)=164.76, p<.001 and χ2diff (2) (PA+EE)=272.41, p<.001, which indicated that the three-dimensional model had a better fit than the two-dimensional models (Byrne, 2001).

Evidence for convergent validity was calculated by means of the correlation between final scores of the various dimensions of the GBQ and the MBI. The values obtained were greater than .51 (Table 2). These correlations were corrected to eliminate the negative effect of the measurement error.

To obtain evidence of criterion-related validity, the concurrent validity was analyzed by calculating the correlation between the scores for the different dimensions of the GBQ as well as the MBI and the scores obtained in the various dimensions of the NEO-FI and the CECAD. Generally speaking, the values obtained for the GBQ and the MBI in relation to the NEO-FI and CECAD were similar (see Table 2). As in the previous case, the correlations were corrected.

Table 3 shows the descriptive analyses, the discrimination index of the GBQ items by dimension. Also included are the α coefficients of each dimension of the GBQ and the MBI as well as their dispersion.

The α coefficients obtained for the GBQ and MBI were similar. Higher reliability values corresponded to the dimensions of Depersonalization and Personal Accomplishment in the GBQ in regard to those obtained for the MBI.

When the reliability coefficient was calculated for the dimensions of the GBQ, and the variance set to that obtained in the dimensions of the MBI, the following reliability values were obtained: .99 for Emotional Exhaustion; .96 for Depersonalization; and .94 for Personal Accomplishment.

This research focused on the differences in the scores of the GBQ dimensions, depending on group membership, with a view to establishing different scales for those groups that required them. For this purpose, a fixed-effect ANOVA was performed with four factors: gender, marital status, 24-hour duty shifts, and hospital unit. No statistically significant differences were found (p>.01) for any of the dimensions as a function of the various factors or their interactions. Both T and centile scores were used for the scales of each dimension since they are frequently used in the field of healthcare (see Table 4).