A cross-sectional study was conducted among undergraduate medical students in Ramathibodi Hospital, Mahidol University, Bangkok, Thailand, using convenience sampling (n=545, 76.1% response rate, female 52.1%). Participants who incompletely responded and those with extreme responses to multiple questions in the questionnaire were excluded. Four hundred fifty-one completed responses (62.9%) were finally included. In the preclinical years, 361 medical students completed the survey (48.3% female). In the clinical years, 190 medical students completed the surveys (52.1% female). Medical students in the preclinical and clinical years were asked to complete the written questionnaire in the last semester of the academic year 2016. The curriculum consisted of a three-year preclinical course (mostly classroom learning) and three-year clinical clerkships. Year 2–5 undergraduate medical students were enrolled to represent both preclinical and clinical years. First- and sixth-year medical students studying outside Ramathibodi Hospital were excluded. All participants provided written informed consent prior to responding and subsequent data collection was confidential. Data handling and accessibility were allowed only for the members of the research team. This study was approved by the Ethics Committee of the Faculty of Medicine of Ramathibodi Hospital, Mahidol University.

Permission to translate the Maslach Burnout Inventory-general survey for students into a Thai version was obtained from Mind Garden, Inc. The Thai version of the Maslach Burnout Inventory-general survey for students (MBI-SS) was developed to measure burnout among Thai undergraduate medical students by a two-stage process, namely, the translation process and testing the resulting version for factorial validity and psychometric properties. The translation process—‘forward’ translation from English to Thai and ‘backward’ into English—was done by two independent experienced translators. The Thai translation was reviewed by a panel of experts including three faculty medical instructors (from paediatrics, family medicine, and clinical epidemiology departments) and based on their suggestions, minor cultural adaptations to fit with Thai context, and to maintain translation quality was performed. The expert panel and a native English speaker (a physician from the family medicine department) then rated for degree of agreement with the original version and the back-translation into English. Finally, pilot testing was performed that showed 10 medical students had no difficulties in understanding and completing the questionnaire.

The next phase consisted of psychometric properties testing of the Thai version of the MBI-SS, 15- item questionnaire consisted 3 subscales: Emotional exhaustion (5 items), Depersonalization (4 items), Professional efficacy (6 items) and rated by 7-point frequency rating scale ranging from 0 (never) to 6 (always). Emotional exhaustion scores over 14 are considered as high level, depersonalization scores over 6 are considered as high level, and personal efficacy score lower than 23 is considered as low level. A study in Spanish (n = 621), Portuguese (n = 723), and Dutch (n = 309) revealed that the three-factor structure (emotional exhaustion, depersonalization, and personal efficacy) of the MBI-SS fits to the data  in the Spanish (χ²/df=209.48/81, CFI=0.95, TLI=0.94 and RMSEA=0.05), Portuguese (χ²/df=280/82, CFI=0.93, TLI=0.92 and RMSEA=0.06), and Dutch (χ²/df=127.14/83, CFI=0.97, TLI=0.96 and RMSEA=0.04). Intercorrelation and internal consistencies (Cronbach’s alpha values) of the MBI-SS subscales of Spain, Portugal, Netherland were 0.74-0.79, 0.69-0.86, and 0.67-0.86, respectively. Most of them met the criterion of 0.7.⁴

Data collected included gender, study year, grade point average (GPA), non-academic time, sleeping time and screening scores for students’ depression (see below). The GPA refers to the cumulative average of grades during their course of study and reflects academic performance on a scale of 0–4 (4 is the best academic performance score). A GPA of less than two is considered to be a ‘fail’ in terms of summative assessment, and the student would then be expected to repeat certain blocks or subjects in the next academic year.

The depression screening tool used in this study is the Thai version of the Patient Health Questionnaire-9 (PHQ-9), a 9-item validated questionnaire. The Thai version of the PHQ-9 had satisfactory internal consistency (Cronbach’s alpha = 0.79) and has acceptable psychometric properties for screening for major depression in general practice with cut-off score of greater than 9.¹⁵

The 33-item written questionnaire including demographic data was subsequently distributed in written format among pre-clinical and clinical year medical students after lecture hall activities by the first author in the last semester of the academic year in February to June 2017, a strategy empirically known to provide a higher response rate in Thai culture.

The interrater reliability (IRR) by means of Kappa was assessed to identify the degree of agreement. Data from the questionnaire was tested for psychometric properties. Internal consistency was analysed to demonstrate the level of reliability, measured with a standardised Cronbach’s coefficient alpha. A Cronbach’s alpha ≥ 0.7 was considered acceptable, although we preferred alpha ≥ 0.8.¹⁶

As cross-culture and language differences between Western and non-Western countries have to be taken into account, a confirmatory factor analysis (Principal Component Analysis) was used to confirm the structure validity of the Thai version MBI-SS and to compare similarity to the original hypothesised measurement model MBI-SS, English version. First, the psychometric sensitivity evaluation was conducted by measuring central tendency and shape. Items with skew (Sk) above three and kurtosis (ku) above 3, in absolute values, were directed to sensitivity problems. Fit indices used for confirmatory factor analysis were the chi-square and degree of freedom ratio (χ²/df), Comparative Fit Index (CFI), the Goodness of Fit Index (GFI), the Non-normed Fit Index (NNFI), Akaike information criterion (AIC) and the Root Mean Square Error of Approximation (RMSEA). The model fit was considered suitable for χ²/df (chi-square and degree of freedom ratio) values of less than 5. When the CFI, GFI, and NNFI values were greater than 0.90, the model was considered an adequate fit.^4,17,18 A RMSEA value of less than 0.08 indicated an acceptable model-data fit, whereas a value above 0.10 indicated that the model should be rejected.^4,19Spearman and Kendall’s tau-b method was used to identify correlations between burnout and other factors. Confirmatory factor analysis, descriptive data, and correlation analysis was performed by using AMOS® version18.0 program, SPSS Statistics for Windows, version x.0 (SPSS Inc., Chicago, IL, USA).

In the translation process from the English to the Thai version, the backward translation and the original English version were compared and analysed for interrater reliability. The Thai version of MBI-SS showed an acceptable value of Kappa = 0.83.

The internal consistency of the questionnaire was good with a Cronbach’s alpha value of 0.80. The Cronbach’s alpha coefficient values for emotional exhaustion, depersonalisation and person accomplishment were 0.89, 0.81, and 0.70, respectively. Mean, standard deviation and intercorrelations between each subscale are reported in Table 1.

The CFI and GFI values of the three-factor model (M1) were 0.85 and 0.85, which did not satisfy the respective criteria. The modified three-factor model (M2) was slightly adjusted using modification indices provided by AMOS. The modified three-factor model (M2) improved the model–data fit to acceptable values of CFI=0.97 and GFI= 0.95. The confirmatory factor analysis demonstrated good fit indices of the Thai version of the MBI-SS, namely, X²/df=197.62/83, CFI=0.97, GFI=0.95, NNFI=0.96, AIC=271.62, and RMSEA=0.06 (Table 2).

The path diagram with standard factor loadings of the 15-item Thai version of MBI-SS (Figure 1) indicated standardised coefficients of the relationship between factors and items between 0.51–0.90 (Table 3).

The average GPA was 3.27 and 3.22 among pre-clinical and clinical year medical students, respectively (Table 4). Forty percent of the participants had a high risk of depression (score above 9) as identified by the PHQ-9 screening tool. The prevalence of burnout on all three subscales among medical students was 28.4% (n=128) as identified both by low personal accomplishment, high emotional exhaustion, and high depersonalisation. Medical students’ perceptions of burnout on subscales were 54.8% for low personal accomplishment, 57.4% for high emotional exhaustion and 65% for high depersonalisation.

Statistically significant, yet weak negative correlations between burnout and GPA (r =−0.119, df=3, p=0.002) and gender (male) (r =−0.139, df=1, p=0.003) were found. Students who had better academic performance and male gender had lower risk for burnout. Burnout also had weak positive associations with the depression score (PHQ-9, r=0.294, df=2, p<0.001). Risk of depression had a significant, yet modest positive association with emotional exhaustion (r=0.469, df=4, p<0.001) and depersonalisation (r=0.411, df=4, p<0.001), and a weak inverse association with professional efficacy (r=−0.273, df=4, p<0.001). These findings indicate that students that experienced emotional exhaustion and/or depersonalisation were at risk for depression as well as students who perceived low personal efficacy. Female gender was positively associated with professional efficacy (r=0.104, df=2, p=0.020). A higher GPA was weakly and negatively associated with depersonalisation (r=-0.112, df=6, p=0.017) (Table 5).

The strengths of this study include the adequate sample size of the students that helped facilitate the factorial validation of the items and test of the psychometric properties of the questionnaire. This study also provides evidence that the psychometric properties of the Thai version of MBI-SS are valid and allow opportunities for further research on burnout within the Thai context. It is important to note that the prevalence of depression among medical students was not included in this study, and the PHQ-9 was used as a screening tool that only indicates risk of depression and not a diagnostic tool. A limitation of this study was the self-administered questionnaires, which might have been answered dishonestly, due to social desirability. Also, some students who were experiencing burnout or depression might not have responded, so the reported frequency might be an underestimate.

Figure 1. Path diagram with standard factor loadings of the 15-item Thai version of MBI-SS (n=451)

This research was funded by the Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand. The authors would like to express gratitude to Assoc. Prof Dr Pornthep Tanpowpong for aspiring guidance, constructive feedback and Prof Dr Manote Lotrakul for giving permission of the Thai version of the Patient Health Questionnaire-9.

The authors declare that they have no conflict of interest.