Psychometric Properties and Measurement Invariance of the Maslach Burnout Inventory–General Survey in Colombia

Context and Objectives

Burnout is an occupational health phenomenon recognized by the World Health Organization (WHO) that affects approximately 40% of the global working population. Studies in Spanish-speaking countries have reported prevalence rates between 7.9% and 14.9% in Spain, Argentina, and Uruguay, while other Latin American countries show rates between 2.5% and 5.9%. The Maslach Burnout Inventory (MBI) is one of the most widely used instruments for assessing burnout, and subsequently a version was developed for general workers (MBI-GS). Although previous psychometric studies exist in Spanish-speaking samples, complete validation of the MBI-GS in Colombian samples had never been conducted. This study aims to analyze the psychometric properties of the MBI-GS (internal consistency, factor structure, measurement invariance) and its convergent validity in a large sample of Colombian workers.

Method

Participants: A total of 978 workers from three private companies in Bogotá were recruited through non-probabilistic convenience sampling. The first company was a construction firm (15.9% of participants), the second a restaurant company (22.5%), and the third from the automotive production sector (51.9%). The sample comprised 650 men (66.9%) and 318 women (32.7%), with a mean age of 38 years (range 18-87 years; SD = 11.0). 52.4% were older than 35 years. Regarding education: 40.4% completed primary or secondary education, 24.9% achieved technical level, and 19.9% obtained university or postgraduate degrees. With respect to socioeconomic status: 54.2% reported low level, 41.0% medium, and 4.8% high. In hierarchical terms: 61.1% were in assistance and operational positions, and 37.5% in managerial and professional positions.

Instruments:

• Maslach Burnout Inventory–General Survey (MBI-GS): A 16-item self-report instrument, 7-point Likert scale (0 = never; 7 = every day) that assesses attitudes toward work across three dimensions: emotional exhaustion, cynicism, and professional efficacy. The Spanish version showed adequate internal consistencies.
• Overall Job Satisfaction Scale (OJS): 15 items, 7-point Likert scale measuring job satisfaction (1 = very unsatisfied; 7 = very satisfied).
• Utrecht Work Engagement Scale (UWES): 15 items, 7-point Likert scale assessing work engagement as a construct opposite to burnout.
• General Health Questionnaire (GHQ-12): 12 items, 4-point Likert scale measuring psychological distress.
• Acceptance and Action Questionnaire-II (AAQ-II): 7 items, 7-point Likert scale assessing psychological inflexibility.

Procedure: Companies provided permission to conduct psychosocial risk factor assessment with their employees. Instruments were administered in groups (20-25 participants) at different company locations. A psychologist with a current occupational health license conducted the assessment. Participants gave voluntary informed consent and received no compensation.

Analysis: Missing data and item-total correlations were examined, and items with correlations below 0.30 were eliminated. Confirmatory factor analysis (CFA) was conducted using LISREL 8.71 with robust diagonally weighted least squares (DWLS) estimation using polychoric correlations. Three models were compared: (a) one-factor model, (b) two-factor model with emotional exhaustion and cynicism collapsed, and (c) original three-factor model. Goodness-of-fit indices were calculated: root mean square error of approximation (RMSEA), comparative fit index (CFI), non-normed fit index (NNFI), expected cross-validation index (ECVI), and standardized root mean square residual (SRMR). Reliability was analyzed using composite reliability coefficients (CFC), convergent and discriminant validity through average variance extracted (AVE), inter-construct correlations, and measurement invariance analyses.

Results

Factor Structure and Item Psychometric Properties: Corrected item-total correlations ranged from 0.66-0.70 for emotional exhaustion, 0.32-0.61 for cynicism, and 0.43-0.62 for professional efficacy. All items showed adequate discrimination. Cronbach's alphas were 0.86 for emotional exhaustion, 0.72 for cynicism, and 0.79 for professional efficacy, indicating appropriate internal consistency.

Dimensionality Model: The one-factor model showed poor fit (RMSEA = 0.17, CFI = 0.83, NNFI = 0.81, SRMR = 0.15). The two-factor model presented acceptable fit (RMSEA = 0.09, CFI = 0.96, NNFI = 0.95, SRMR = 0.08). The original three-factor model achieved the best fit (RMSEA = 0.05, CFI = 0.99, NNFI = 0.98, SRMR = 0.06), with ECVI values confirming the superiority of the three-factor model. Contrary to previous studies in Spanish-speaking countries that required eliminating item 13 (cynicism), this study found that retaining it was appropriate due to its adequate reliability and validity.

Reliability and Validity: CFC values exceeded 0.70 for all factors (emotional exhaustion = 0.88, cynicism = 0.86, professional efficacy = 0.88), indicating adequate reliability. All standardized factor loadings were significant (>0.40, mostly >0.70). AVE values exceeded 0.50 (emotional exhaustion = 0.60, cynicism = 0.58, professional efficacy = 0.55). Inter-construct correlations were below 0.80 (emotional exhaustion–cynicism = 0.75, emotional exhaustion–professional efficacy = −0.29, cynicism–professional efficacy = −0.44), supporting discriminant validity.

Measurement Invariance: Measurement invariance was supported at both metric and scalar levels across gender, age group, hierarchical level, and socioeconomic status. Changes in RMSEA, CFI, and NNFI were below 0.01. At the hierarchical level, items 9 and 13 (cynicism) could be variant, so partial scalar invariance was explored finding that with distinct intercept estimates for these items in both groups, criteria were met.

Convergent Validity: The MBI-GS showed expected correlations with validation measures. Emotional exhaustion and cynicism showed medium-to-strong positive correlations with psychological distress (GHQ-12: r = 0.53 and −0.43 respectively) and psychological inflexibility (AAQ-II: r = 0.42 and −0.44). Both dimensions showed negative correlations with work engagement (UWES): emotional exhaustion (r = −0.37 to −0.46), cynicism (r = −0.30 to −0.40). Professional efficacy showed the opposite pattern, correlating positively with job satisfaction (r = 0.26–0.28) and engagement (r = 0.29–0.44), and negatively with psychological distress (r = −0.23 to −0.22).

Sociodemographic Differences: Statistically significant differences were found: women showed higher levels of emotional exhaustion (M = 8.86, SD = 6.30) compared to men (M = 6.61, SD = 5.64), t(571.49) = −5.41, p < 0.001. In cynicism, women again showed higher scores (M = 6.05, SD = 5.67) than men (M = 4.89, SD = 5.24), t(966) = −3.14, p = 0.002. Professional efficacy was significantly lower in women (M = 30.08, SD = 5.59) versus men (M = 30.99, SD = 5.94), t(966) = 2.29, p = 0.022. Regarding age group, no significant differences were found in emotional exhaustion, but workers aged 35 or younger showed less cynicism (M = 5.47, SD = 5.14) than those older than 35 (M = 4.88, SD = 5.40). At the hierarchical level, participants in managerial/professional positions showed significantly less cynicism (M = 4.50, SD = 5.06) and more professional efficacy (M = 31.59, SD = 4.88) than those in assistance/operational positions (cynicism: M = 5.70, SD = 5.51; efficacy: M = 30.18, SD = 6.24), t(957) = −3.39, p = 0.001 and t(904.88) = 3.90, p < 0.001. Regarding socioeconomic status, participants with low status showed higher cynicism (M = 5.53, SD = 5.65) and lower professional efficacy (M = 30.37, SD = 6.10) compared to medium status (cynicism: M = 4.71, SD = 4.90; efficacy: M = 31.32, SD = 5.10).

Discussion and Conclusions

This study demonstrated that the MBI-GS possesses adequate psychometric properties in Colombian samples. Internal consistency and factor structure were appropriate, validating the original three-factor model. Contrary to previous studies, no items needed to be eliminated. Item 13 (cynicism) showed the lowest discrimination index but was retained due to its adequate reliability and validity. The study provided empirical evidence for measurement invariance of the MBI-GS, an aspect not previously analyzed in Colombian samples and methodologically important. Differences found in women with higher exhaustion and cynicism could be due to factors such as job characteristics, greater family demands, or characteristics of the analyzed companies. Convergent validity was appropriate, correlating in expected directions with job satisfaction, work engagement, psychological distress, and psychological inflexibility. Limitations are acknowledged: the study analyzed only private companies in Bogotá without systematic clinical information about diagnoses that would allow comparison of clinical relevance. Future research is needed with public companies and more diverse samples, sensitivity analyses to psychological interventions, and exploration of variables explaining gender differences in burnout.

Importance and Contribution

This is the first complete psychometric validation study of the MBI-GS in Colombia in a large and heterogeneous sample (n = 978) of workers. It provides robust evidence of appropriate instrument functioning in Colombian and Latin American Spanish-speaking contexts. The inclusion of measurement invariance analysis is methodologically important because it justifies comparison of mean scores across different groups of participants. The study contributes to filling an important gap in the psychometric literature of burnout assessment instruments in Latin American working populations, facilitating its reliable use in occupational psychosocial risk assessments in Colombia.

---