Burnout in Singapore’s public primary healthcare workers: a cross-sectional study Authors: Pei Zheng Tan, Anna Szücs, Yong Yi Tan, Lay Hoon Goh Published: 2025 ABSTRACT AbstractBackgroundBurnout in healthcare workers results in higher staff turnover and poorer patient care. Knowledge of burnout and its associated factors among polyclinics’ (public primary care clinics) healthcare workers in Singapore is lacking. This study aims to evaluate the extent of burnout and identify associated factors in polyclinic staff.MethodsThis was a cross-sectional study conducted in Singapore’s National University Polyclinics. From 11 July to 4 September 2024, an anonymous self-administered online questionnaire was sent to eligible doctors, nurses, and allied health staff. The Maslach Burnout Inventory Human Services Survey (MBI-HSS) with domains of emotional exhaustion, depersonalisation, and personal accomplishment was used to measure extent of burnout and tested for associations with staffs’ demographics, working-related factors, patient-related experiences, lifestyle, resilience levels (measured by the Brief Resilience Scale), and intention to quit using multivariable linear regression.ResultsOf the 228 participants, 82 were doctors, 79 were nurses, and 67 were allied health. Mean total MBI-HSS scores were 25.98 (SD = 12.40, min-max = 1–53) for emotional exhaustion, 10.22 (SD = 7.24, min-max = 0–30) for depersonalisation, and 31.69 (SD = 8.33, min-max = 8–48) for personal accomplishment. Emotional exhaustion was associated with being a doctor (B = 4.55, SE = 1.60, p = 0.005) and nurse (B=-4.16, SE = 1.70, p = 0.015) compared to allied health workers. Emotional exhaustion was also associated with fewer average hours of exercise a week (B=-0.51, SE = 0.22, p = 0.024), and lower levels of resilience (B=-4.44, SE = 1.12, p < 0.001). Depersonalisation was associated with being a doctor (B = 2.56, SE = 0.97, p = 0.009) and perception of less recognition by patients (verbally or in writing) (B=-6.25, SE = 2.56, p = 0.016 for neutral; B=-8.99, SE = 2.54, p < 0.001 for agree; B=-8.69, SE = 2.82, p = 0.002 for strongly agree). Personal accomplishment was associated with being male (B = 3.38, SE = 1.37, p = 0.015), having full-time domestic help (B = 3.12, SE = 1.17, p = 0.008), being a member of a teamlet (B = 3.02, SE = 1.33, p = 0.024), perceiving greater recognition by patients (B = 8.60, SE = 3.21, p = 0.008 for agree; B = 12.46, SE = 3.55, p < 0.001 for strongly agree), and higher levels of resilience (B = 0.60, SE = 0.13, p < 0.001).ConclusionsThe study highlighted the extent of burnout in healthcare workers in Singapore’s public primary care and its associated factors. Strategies that could potentially manage burnout include improving patient-provider relationships, improving work environment, increasing resilience in healthcare workers, and optimising team-based care models.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12875-025-03093-5.Keywords: Burnout, Primary care, Health services, Surveys, Singapore Background Burnout is a mental health condition “resulting from chronic workplace stress that has not been successfully managed” [1], and is characterised by energy depletion or exhaustion, increased negativism or cynicism related to one’s job, and reduced professional efficacy. It can result in health implications, including alcohol abuse [2], depression, and suicidal ideation [3–5]. Burnout is also associated with a lower quality of patient care [6, 7], major medical errors [8, 9], and decreased patient satisfaction [10]. Lastly, burnout impacts organisations, being associated with higher staff turnover [11, 12]. As the cornerstone of Singapore’s healthcare system, primary care in Singapore comprises around 2,000 private general practitioner (GP) clinics and 26 polyclinics (public primary care clinics) [13]. Although most primary care physicians work in the private sector, the National Health Population Survey 2023 reported that polyclinics manage about 60% of the chronic disease patient-load [14]. This high patient-load requires a large and steady workforce. There are many reasons for burnout in primary care. For example, physicians deal with diagnostic uncertainty [15], are being tasked to gatekeep scarce resources, and have to balance the doctor-patient relationship [16] with competing time constraints. Nurses face pressure to assume new roles previously provided by physicians and multi-task administrative duties [17]. A previous Singapore study [18] found that compared to specialists, GPs in private practice are more likely to be satisfied with the amount of time for each patient, initiate changes in their work, and maintain harmony of relationship among colleagues and staff associates. However, this may not be the case in the public sector. In a qualitative study on the experiences of polyclinics [19], doctors shared less favourable perspectives, including a heavy workload, short consultation times, and a perceived lack of management’s concern about doctors’ welfare. Though, positive aspects of polyclinics also emerged, including opportunities for academic activities such as teaching and research, as well as continuity of care within teamlets (care teams serving a regular pool of patients and comprising doctors, a care manager who is a chronic disease management-trained nurse, and care coordinators). There was a wide range in prevalence of burnout symptoms reported in studies conducted among healthcare workers in Singapore, from 37.8% with elevated scores in emotional exhaustion (a domain of burnout) in healthcare workers from Singapore Health Services (Singapore’s eastern public healthcare cluster out of 3 clusters in Singapore) during the pre-COVID-19 period from July 2019 to January 2020 [20], to 82.1% of physicians in various private primary care networks and the National University Polyclinics (NUP; Singapore’s western public primary healthcare cluster) between April to September 2020 during the COVID-19 pandemic [21]. To our knowledge, there has been no reassessment of burnout and its associated factors post-COVID-19 amongst primary care providers in Singapore, despite the concerningly high numbers reported during the pandemic. Such an evaluation is particularly timely because Singapore is now shifting its focus from curative care in tertiary hospitals to preventive health in primary care in the context of the new national health campaign: Healthier SG [22]. Healthier SG is transforming Singapore’s healthcare landscape by mobilising primary care in promoting preventive care and early intervention while continuing to provide care to those with chronic conditions [23]. To meet these growing demands on primary care, Singapore’s Ministry of Health is looking to expand its network of polyclinics from 26 to 32 by 2030 [24]. To cope with a greater primary care load and to continue providing quality patient care, a large and psychologically sturdy polyclinic workforce is essential, which in turn requires a more granular understanding of factors associated with burnout in the polyclinic workforce. As of now, there is inconsistent evidence on the association between burnout and individual characteristics of primary healthcare workers. For example, findings were mixed for demographic factors such as sex and work-related factors such as length of tenure as a predictor of burnout [25]. Extant interventions tend to target individual coping strategies, e.g., by promoting mindfulness [26], yet interventions targeting system-level changes may be more effective [27]. There is a need to examine a broad range of individual- and work environment-related factors together, to devise new, combined multi-level interventions that can address the actual causes of burnout in the context of Singapore’s public primary care. To address these gaps, the present study aims to assess the extent of burnout among NUP’s healthcare workers post-COVID and identify factors associated with burnout, which can potentially provide avenues to address burnout in public primary healthcare in general and at Singapore’s polyclinics in particular. Methods Participants A total of 738 NUP full-time and part-time staff comprising 220 doctors, 242 nurses, and 276 allied health workers (from pharmacy, dietetics, physiotherapy, podiatry, psychology, medical social work, and radiography) were eligible for the study. Doctors who run clinics infrequently in NUP (e.g., training doctors who run residency continuity clinics once a week), locums (i.e., doctors or allied health professionals who are not salaried NUP employees and only fill in on an ad hoc basis when there is a lack of staff), as well as non-clinical, administrative, and dental staff were excluded from the study. A voluntary response sampling method was used to recruit participants (see Procedure below). Data collection Data was collected using a self-administered questionnaire hosted on FormSG [28], a secure site for government forms. The questionnaire comprised items: (i) participants’ demographics (age, sex, nationality, ethnicity, marital status, number of children, caregiving status, availability of domestic help, occupation, highest level of qualification, designation, number of years working in the profession, number of years working in the polyclinic), (ii) working conditions (number of domains of work, number of hours exceeding knock-off time, duration of lunch break, number of patient encounters a week, whether one works in a teamlet), (iii) lifestyle factors (duration of sleep, self-care, exercise), (iv) workplace support (perceived support received from colleagues, reporting officer, upper management, in terms of training opportunities) answered on a Likert scale, (v) patient-related experiences (recognition by patients, verbal/physical abuse by patients) also answered on a Likert scale, (vi) intention to leave their current position within the next 6 months to measure potential staff attrition, (vii) resilience (detailed below), and (viii) three domains of burnout (detailed below). The survey did not assess services that were equally available to all NUP staff, such as clinic facilities, in-house counselling, or access to whistle-blowing channels. No personal identifiers were collected. Burnout was assessed with the Maslach Burnout Inventory Human Services Survey (MBI-HSS) for medical personnel. MBI-HSS measures burnout in three domains: emotional exhaustion, depersonalisation, and personal accomplishment [29]. Emotional exhaustion assesses feelings of being emotionally overextended and exhausted by one’s work. Depersonalisation measures an unfeeling and impersonal response toward recipients of one’s care. Personal accomplishment has an opposite direction to the two domains and assesses feelings of competence and successful achievement in one’s work with people. The scale comprises 22 question statements on job-related feelings (nine on emotional exhaustion, five on depersonalisation, and eight on personal accomplishment). Respondents answer how often they ever feel this way about their job on a Likert scale from 0 to 6 (0 = never, 1 = a few times a year or less, 2 = once a month or less, 3 = a few times a month, 4 = once a week, 5 = a few times a week, 6 = every day). The three domains are intended to be used as a continuum and have no specifically prescribed cut-offs for clinical levels of burnout [30]. Scores can be calculated using either the total sum or average of responses for each subscale. In our study, we reported total and mean scores of each subscale to aid interpretability but subsequent bivariate and multivariate analysis utilised the subscale’s total score. In terms of internal consistency, the Cronbach alpha coefficients for the MBI-HSS domains of emotional exhaustion, depersonalisation, and personal accomplishment in our study were 0.92, 0.83, and 0.81 respectively, showing high internal consistency. MBI has also been administered in numerous studies among healthcare workers in Singapore [20, 31–36], thus allowing comparison of differences in burnout between our study populations. Resilience, defined as the ability to bounce back or recover from stress, was measured with the Brief Resilience Scale [37]. This self-reported assessment comprises six items, each being scored from 1 (strongly disagree) to 5 (strongly agree). The scale’s scores are calculated as item averages, where an average score of 1.00–2.99 indicates low resilience, 3.00–4.30. normal resilience, and 4.31–5.00 high resilience. The Brief Resilience Scale had been administered in prior studies in Singapore, including among mental health professionals in a tertiary psychiatric hospital [38]. The Cronbach alpha coefficient for the Brief Resilience Scale in our study was 0.86, showing high internal consistency. Sample size calculation Based on the results of a previous study on burnout in a healthcare cluster in Singapore using the MBI-HSS [20], the burnout domain with the largest standard deviation (SD) was 13.0. Considering 738 eligible NUP staff and a significance threshold of 0.05, the sample size was estimated at 208 participants for the study to evaluate levels of burnout in NUP healthcare workers with a 1.5 points margin of error. Procedure Ethical approval was received from the National Healthcare Group Domain Specific Review Board (study reference: 2023/00979). The questionnaire was first piloted among six NUP staff from different health professions to evaluate participants’ understanding of the relevance and clarity of the items and response options. Between July 2024 and September 2024, all eligible NUP staff except those who participated in the pilot were invited via their work email to complete the voluntary, anonymous questionnaire on FormSG. After following a link to the survey, participants were provided with a detailed explanation about the study and needed to give their informed consent to participate by ticking a checkbox before proceeding with the survey. The survey took approximately 20 minutes to complete although there was no time limit once participants logged in. No financial rewards were offered for participation. Statistical analysis The software R 4.3.2 was used for statistical analysis. A p-value of < 0.05 was considered statistically significant. First, descriptive statistics and histograms were employed to assess the distribution of individual independent and dependent variables. Missingness was managed using complete case analysis. Overall, 22 data points were missing or needed to be excluded for other reasons. Twelve participants’ data (5.3% of the total sample) were excluded for the question on average lunch break minutes per day due to ambiguity (these participants reported ‘1’ and it is uncertain if participants meant one minute or one hour). Four participants (1.8%) had missing data in number of work domains. Finally, three participants’ data (1.3%) were excluded from the question on hours of exercise per week, two (0.9%) were excluded for the question on hours of sleep per day, and one (0.4%) was excluded for the question on number of patient encounters per week as the numbers provided were unrealistically large. Next, bivariate analyses were performed to examine the relationships between the independent variables and the three MBI-HSS domains (emotional exhaustion, depersonalisation, and personal accomplishment). Most collected variables were included in the bivariate analysis, with the exceptions being variables that were not equivalent across different job types (e.g., highest academic qualification, designation) as well as categorical variables for which most responses pertained to the same category with no other categories reaching 15% (e.g., nationality, ethnicity). The bivariate analyses were performed with Pearson correlation for continuous and normally distributed independent variables, Spearman correlation for continuous and non-normally distributed independent variables, Student’s t-test for categorical independent variables with two categories, and analysis of variance (ANOVA) for categorical independent variables with more than two categories. All independent variables with a p-value of < 0.05 in at least one of the three MBI-HSS domains of the bivariate analyses were selected into confirmatory multivariate analyses using multivariable linear regression. To ensure complete case analysis in the multivariate analysis, sixteen observations with missing values were removed prior to analysis (twelve from average lunch break minutes per day, three from hours of exercise per week, and one from patient encounters per week). Separate regression models were built for each of the three MBI-HSS domains as dependent variable. This additional step involving multivariate analyses enabled adjusting for potential confounders, and allowed evaluation of each identified associations’ robustness and importance relative to the others. Consistently, the study’s main findings and conclusions were based on the multivariate analysis results. Results Participant characteristics 228 staff (30.9% of eligible staff) participated in the study (Table 1), including 82 doctors (35.96%), 79 nurses (34.65%), and 67 allied health workers (29.39%). Their mean age was 37.31 years (SD = 8.62, min-max = 21–63). 181 (79.39%) were women. Table 1Characteristics of study participantsCharacteristics (N = 228)Count (%)Age, mean ± SD, min-max37.31 ± 8.62, 21–63Sex Male47 (20.61%) Female181 (79.39%)EthnicityChinese166 (72.81%)Malay26 (11.40%)Indian18 (7.89%) Others18 (7.89%)Nationality Singaporean183 (80.26%) Permanent Resident30 (13.16%) Others15 (6.58%)Occupation Doctors82 (35.96%) Nurses79 (34.65%) Allied health67 (29.39%)Designation of doctors Medical officer6 (7.32%) Family medicine resident12 (14.63%) Resident physician9 (10.98%) Family physician33 (40.24%) Associate consultant9 (10.98%) Consultant8 (9.76%) Senior consultant5 (6.10%)Designation of nurses Enrolled nurse2 (2.53%) Staff nurse9 (11.39%) Senior staff nurse44 (55.7%) Assistant nurse clinician16 (20.25%) Nurse clinician/nurse manager/nurse educator5 (6.33%) Senior nurse clinician/nurse manager/nurse educator2 (2.53%) Others1 (1.27%)Designation of allied health Allied health support5 (7.46%) Pharmacy technician4 (5.97%) Pharmacy technician supervisor2 (2.98%) Allied health professional27 (40.30%) Senior allied health professional24 (35.82%) Principal allied health professional5 (7.46%)Highest academic qualification of doctors MBBS and equivalent25 (30.49%) GDFM15 (18.29%) MMED/MCFPS24 (29.27%) FCFPS17 (20.73%) Others1 (1.22%)Highest academic qualification of nurses NITEC in nursing2 (2.53%) Diploma in Nursing13 (16.46%) Bachelors in nursing56 (70.89%) Masters in nursing7 (8.86%) Others1 (1.27%)Highest academic qualification of allied health O-Levels1 (1.49%) NITEC1 (1.49%) Diploma20 (29.85%) Bachelor’s degree37 (55.22%) Master’s degree7 (10.45%) Others1 (1.49%)Number of years working in the profession, median (IQR), min-max13 (7–19), 0–45Number of years working in the polyclinic, median (IQR), min-max6 (2–10), 0–30Marital status Single51 (22.37%) In a long-term relationship or married173 (75.88%) Divorced2 (0.88%) Widowed2 (0.88%)Number of children, median (IQR), min-max1 (0–2), 0–7Presence of domestic help Yes, full-time59 (25.88%) Yes, part-time7 (3.07%) No162 (71.05%)Caregiver status Yes119 (52.19%) No109 (47.81%)Number of work domains (out of clinical, education research, and administrative)* 1 domain29 (12.95%) 2 domains65 (29.02%) 3 domains104 (46.43%) 4 domains26 (11.61%)Reported number of hours above knock-off time per week, median (IQR), min-max1 (0–4), 0–44Reported average time for lunch break in minutes per day*, median (IQR), min-max52.5 (45–60), 0–180Reported number of patient encounters per week*, median (IQR)100 (50–160), 0–1400Is a member of a teamlet Yes85 (37.28%) No143 (62.72%)Perception of support by colleagues Strongly agree60 (26.32%) Agree124 (54.39%) Neutral36 (15.79%) Disagree6 (2.63%) Strongly disagree2 (0.88%) Prefer not to answer0 (0%)Perception of support by reporting officers Strongly agree71 (31.14%) Agree105 (46.05%) Neutral41 (17.98%) Disagree7 (3.07%) Strongly disagree3 (1.32%) Prefer not to answer1 (0.44%)Perception of support by upper management and leadership team Strongly agree28 (12.28%) Agree106 (46.49%) Neutral66 (28.95%) Disagree16 (7.02%) Strongly disagree7 (3.07%) Prefer not to answer5 (2.19%)Perception of opportunities for training and upskilling Strongly agree45 (19.74%) Agree140 (61.40%) Neutral31 (13.60%) Disagree11 (4.82%) Strongly disagree0 (0%) Prefer not to answer1 (0.44%)Perception of recognition by patients (verbally or in writing) Strongly agree21 (9.21%) Agree107 (46.93%) Neutral71 (31.14%) Disagree21 (9.21%) Strongly disagree7 (3.07%) Prefer not to answer1 (0.44%)Perception of verbal and/or physical abuse from patients affecting own mental health Strongly agree19 (8.33%) Agree72 (31.58%) Neutral72 (31.58%) Disagree54 (23.68%) Strongly disagree11 (4.82%) Prefer not to answer0 (0%)Intention to leave current position in the next 6 months Certainly not65 (28.51%) Unlikely77 (33.77%) I don’t know45 (19.74%) Likely12 (5.26%) Certainly yes9 (3.95%) I prefer not to answer20 (8.77%)Average hours of sleep per day*, mean ± SD, min-max6.44 ± 0.85, 4.00–9.00Average hours of recreational or self-care activities per week, median (IQR), min-max5 (2–10), 0–72Average hours of exercise per week*, median (IQR), min-max2 (1–3), 0–28Brief Resilience Scale average score, mean ± SD, min-max3.17 ± 0.63, 1.50–5.00Numbers indicate count (proportion in %) unless otherwise specifiedLegend: MBBS Bachelor of Medicine, Bachelor of Surgery, GDFM Graduate Diploma in Family Medicine, MMED Master of Medicine in Family Medicine; MCFPS Collegiate Membership of the College of Family Physicians Singapore, FCFPS Fellowship of the College of Family Physicians Singapore, NITEC National Institute of Technical Education Certificates* Variables with missing or excluded data (refer to the statistical analysis subsection under Methods for more details) Characteristics of study participants Numbers indicate count (proportion in %) unless otherwise specified Legend: MBBS Bachelor of Medicine, Bachelor of Surgery, GDFM Graduate Diploma in Family Medicine, MMED Master of Medicine in Family Medicine; MCFPS Collegiate Membership of the College of Family Physicians Singapore, FCFPS Fellowship of the College of Family Physicians Singapore, NITEC National Institute of Technical Education Certificates * Variables with missing or excluded data (refer to the statistical analysis subsection under Methods for more details) In terms of work-related factors, 104 (46.43%) participants reported working in three and 65 (29.02%) in two professional domains (out of clinical, education, research, and administrative). Participants reported a median of one hour exceeding knock-off time a week (IQR = 0–4, min-max = 0–44), and a median of 100 patient encounters per week (IQR = 50–160, min-max = 0–1,400). Most participants (n = 143, 62.72%) were not members of a teamlet. Most participants rated a generally adequate level of support (defined as strongly agreeing or agreeing) by colleagues (n = 184, 80.71%), reporting officers (n = 176, 77.19%), and upper management/leadership team (n = 134, 58.77%). In terms of patient-related experiences, majority of participants also felt recognised by patients verbally or in writing (n = 128, 56.14% strongly agreed or agreed) while 71 of them (31.14%) remained neutral on this point. Yet, 91 of the participants (39.91%) strongly agreed or agreed that they perceived verbal and/or physical abuse from patient to be affecting their mental health, while 72 (31.58%) were neutral on this point. When asked of their intention to leave their current position in the next six months, 9 participants (3.95%) answered “certainly yes”, 12 (5.26%) answered “likely”, 45 (19.74%) “I don’t know”, and 20 (8.77%) “I prefer not to answer”. In terms of lifestyle factors, participants reported exercising a median of two hours per week (IQR = 1–3, min-max = 0–28). The mean resilience score was 3.17 (SD = 0.63, min-max = 1.50–5.00), indicating normal resilience. Extent of burnout The mean total score for emotional exhaustion on the MBI-HSS was 25.98 points (SD = 12.40, min-max = 1.00–53.00) (see Table 2), the mean total score for depersonalisation was 10.22 points (SD = 7.24, min-max = 0.00–30.00), and the mean total score for personal accomplishment was 31.69 points (SD = 8.33, min-max = 8.00–48.00). Table 2Participants’ Maslach burnout inventory human services survey scoresVariables (range of scores)Mean (SD)Participants’ MinParticipants’ MaxMBI-HSS mean total score Emotional exhaustion (0–54)25.98 (12.40)1.0053.00 Depersonalisation (0–30)10.22 (7.24)0.0030.00 Personal accomplishment (0–48)31.69 (8.33)8.0048.00MBI-HSS mean item score Emotional exhaustion (0–6)2.89 (1.38)0.115.89 Depersonalisation (0–6)2.04 (1.45)0.006.00 Personal accomplishment (0–6)3.96 (1.04)1.006.00Legend: MBI-HSS Maslach burnout inventory human services survey, min minimum, max maximum Participants’ Maslach burnout inventory human services survey scores Legend: MBI-HSS Maslach burnout inventory human services survey, min minimum, max maximum Supplementary Table S1 summarises the correlation between each domain. There was a strong positive correlation between emotional exhaustion and depersonalisation (r = 0.74), but not between emotional exhaustion and personal accomplishment (r=−0.27) or between depersonalisation and personal accomplishment (r=−0.26). Results of bivariate analysis Twenty-one out of 25 variables were associated with at least one MBI-HSS domain with a p-value of < 0.05 in the bivariate analysis (see Table 3). These variables were included in the multivariate analysis. Table 3Bivariate analyses of Maslach burnout inventory human services survey subscale score and other variablesIndependent variablesMBI-HSS emotional exhaustionMBI-HSS depersonalisationMBI-HSS personal accomplishmentTest statisticsp-valueTest statisticsp-valueTest statisticsp-value Socio-demographics Ager = −0.29 < 0.001 r = −0.30 < 0.001 r = 0.17 0.011 Sext = 30.14< 0.001t = 18.77 < 0.001 t = 55.17 < 0.001 Marital statusF = 3.95 0.009 F = 3.79 < 0.001 F = 1.240.295 Number of childrenρ =−0.25 < 0.001 ρ = −0.24 < 0.001 ρ = 0.15 0.024 Is a caregivert = 29.75 < 0.001 t = 18.09 < 0.001 t = 54.56< 0.001 Has domestic helpF = 0.440.647F = 1.180.308F = 7.94< 0.001 OccupationF = 6.25< 0.001F = 10.14 < 0.001 F = 3.20 0.043 Work- and patient- related factors Number of years working in the professionρ = −0.30< 0.001ρ = −0.33 < 0.001 ρ = 0.090.170 Number of years working in the polyclinicρ = −0.120.068ρ = −0.070.278ρ = −0.010.938 Number of domains of workF = 0.210.891F = 0.110.954F = 1.750.159 Reported number of hours exceeding knock-off time per weekρ = 0.040.522ρ = −0.070.326ρ = 0.20 < 0.001 Reported average time for lunch break in minutes per dayρ = −0.16 0.017 ρ = −0.100.151ρ = −0.16 0.017 Is a member of a teamlett = 29.93 < 0.001 t = 18.41 < 0.001 t = 54.84 < 0.001 Reported number of patient encounters per weekρ = 0.22 < 0.001 ρ = 0.29 < 0.001 ρ = 0.010.914 Perception of support by colleaguesF = 5.34 < 0.001 F = 1.710.149F = 6.00 < 0.001 Perception of support by reporting officersF = 3.96 0.002 F = 1.080.375F = 7.56 < 0.001 Perception of support by upper management and leadership teamF = 7.62 < 0.001 F = 5.69 < 0.001 F = 6.15 < 0.001 Perception of opportunities for training and upskillingF = 4.90 < 0.001 F = 1.260.286F = 4.45 < 0.001 Perception of recognition by patients (verbally or in writing)F = 12.03 < 0.001 F = 15.99 < 0.001 F = 9.90 < 0.001 Perception of verbal and/or physical abuse from patientsF = 9.50 < 0.001 F = 7.77 < 0.001 F = 4.27 0.002 Intention to quit in the next 6 monthsF = 15.14 < 0.001 F = 10.43 < 0.001 F = 2.190.056 Lifestyle factors Average hours of sleep per dayr = −0.120.066r = −0.040.524r = 0.100.141 Average hours of recreational or self-care activities per weekρ = −0.100.135ρ = −0.030.629ρ = 0.130.055 Average hours of exercise per weekρ = −0.120.073ρ = −0.090.167ρ = 0.16 0.019 Personal attributes Brief Resilience Scale average scorer = −0.43 < 0.001 r = −0.26 < 0.001 r = 0.42 < 0.001 Statistically significant values are boldedLegend: MBI-HSS Maslach burnout inventory human services survey, r Pearson’s correlation coefficient, t t-statistic (Student’s t-test), F F-statistic (ANOVA), ρ Spearman’s rank correlation coefficient Bivariate analyses of Maslach burnout inventory human services survey subscale score and other variables Statistically significant values are bolded Legend: MBI-HSS Maslach burnout inventory human services survey, r Pearson’s correlation coefficient, t t-statistic (Student’s t-test), F F-statistic (ANOVA), ρ Spearman’s rank correlation coefficient Results of multivariate analysis A total of 212 observations were utilised for the multivariate analysis after removing 16 observations with missing data (as described under methods). Supplementary Table S2 depicts the model selection statistics in the first step of the forward-backwards stepwise model selection approach for the three sets of models, having MBI-HSS emotional exhaustion, depersonalisation and personal accomplishment domain scores respectively as the dependent variable. All three final linear regression models predicting MBI-HSS domain total scores were statistically significant (Table 4), respectively predicting 57% of the total variance for emotional exhaustion (F(28, 183) = 8.75, p < 0.001, R2 = 0.57, adjusted R2 = 0.51), 50% for depersonalisation scores (F(19,192) = 9.91, p < 0.001, R2 = 0.50, adjusted R2 = 0.45), and 35% for personal accomplishment scores (F(13, 198) = 8.37, p < 0.001, R2 = 0.35, adjusted R2 = 0.31). The estimates, standard errors, and p-value of each factor compared against the MBI-HSS domains are further summarised in Table 4. Table 4Final linear regression models predicting Maslach burnout inventory human services survey domain scoresDependent variableMBI-HSS emotional exhaustionMBI-HSS depersonalisationMBI-HSS personal accomplishmentestimate (standard error)t-valuep-valueestimate (standard error)t-valuep-valueestimate (standard error)t-valuep-valueSocio-demographicsMale sex (ref: female)Not retainedN/AN/ANot retainedN/AN/A3.38 (1.37)2.470.015Is a caregiver (ref: no)Not retainedN/AN/ANot retainedN/AN/A−1.88 (1.02)−1.840.067Has domestic help (ref: no domestic help) Yes – full-timeNot retainedN/AN/ANot retainedN/AN/A3.12 (1.17)2.660.008 Yes – part-time1.84 (2.72)0.680.499Occupation (ref: allied health) Doctor4.55 (1.60)2.840.0052.56 (0.97)2.650.009−2.38 (1.61) −1.470.143 Nurse−4.16 (1.70)−2.440.015−0.88 (1.05)−0.830.406−2.44 (1.26) −1.940.054Work- and patient- related factorsNumber of years working in the professionNot retainedN/AN/A−0.10 (0.05)−1.980.050Not retainedN/AN/AIs a teamlet member (ref: not a teamlet member)Not retainedN/AN/ANot retainedN/AN/A3.02 (1.33)2.270.024Reported number of patient encounters per week0.01 (0.01)1.360.1740.01 (< 0.01)2.870.005Not retainedN/AN/APerceived support by colleagues (ref: strongly disagree) Disagree12.43 (7.70)1.620.108Not retainedN/AN/ANot retainedN/AN/A Neutral5.53 (6.78)0.820.415 Agree3.05 (6.63)0.460.646 Strongly agree1.51 (6.72)0.230.822Perception of opportunities for training and upskilling (ref: disagree) Neutral8.68 (3.94)2.200.029Not retainedN/AN/ANot retainedN/AN/A Agree6.85 (3.66)1.870.062 Strongly agree5.88 (3.94)1.490.136 Prefer not to answer27.20 (9.62)2.830.005Perceived recognition by patients (verbally or in writing) (ref: strongly disagree) Disagree1.94 (4.53)0.430.668−1.30 (2.74)−0.470.6362.26 (3.49)0.650.519 Neutral−1.81 (4.27)−0.420.672−6.25 (2.56)−2.440.0165.48 (3.25)1.690.093 Agree−5.59 (4.22)−1.30.186−8.99 (2.54)−3.54< 0.0018.60 (3.21)2.680.008 Strongly agree−3.37 (4.74)−0.710.478−8.69 (2.82)−3.090.00212.46 (3.55)3.51< 0.001 Prefer not to answer−12.48 (10.57)−1.180.239−14.11 (6.11)−2.310.0221.64 (7.60)0.220.830Perception of verbal and/or physical abuse from patients affecting own mental health (ref: strongly disagree) Disagree−3.29 (3.10)−1.060.290−0.53 (1.87)−0.280.778Not retainedN/AN/A Neutral1.16 (3.03)0.380.702−0.52 (1.83)−0.290.775 Agree2.83 (3.14)0.900.3691.66 (1.86)0.89 0.374 Strongly agree4.25 (3.79)1.120.2623.16 (2.27)1.39 0.165Intention to quit within 6 months (ref: certainly not) Unlikely2.90 (1.70)1.700.9052.32 (1.00)2.310.021Not retainedN/AN/A I don’t know6.97 (2.01)3.46< 0.0013.78 (1.16)3.250.001 Likely11.83 (3.11)3.81< 0.0017.58 (1.89)4.01< 0.001 Certainly yes3.68 (3.74)0.980.3261.58 (2.23)0.710.480I prefer not to answer10.84 (2.54)4.27< 0.0015.66 (1.48)3.83< 0.001Lifestyle factorsAverage number of hours of exercise per week−0.51 (0.22)−2.280.024−0.20 (0.13)−1.510.093Not retainedN/AN/APersonal attributesBrief Resilience Scale score−4.44 (1.12)−3.94< 0.001Not retainedN/AN/A0.60 (0.13)4.42< 0.001Overall model statisticsR20.570.500.35Adjusted R20.510.450.31F statistic8.75 (df = 28; 183)< 0.0019.91 (df = 19; 192)< 0.0018.37 (df = 13; 198)< 0.001Statistically significant values are boldedLegend: MBI-HSS Maslach burnout inventory human services survey, ref reference, N/A Not Applicable, DF Degrees of Freedom Final linear regression models predicting Maslach burnout inventory human services survey domain scores Statistically significant values are bolded Legend: MBI-HSS Maslach burnout inventory human services survey, ref reference, N/A Not Applicable, DF Degrees of Freedom After controlling for other independent variables, multiple factors were identified to be statistically significant (all p < 0.05). For socio-demographic factors, compared to allied health, doctors had higher emotional exhaustion and depersonalisation while nurses had lower emotional exhaustion. Men were found to have higher personal accomplishment scores than women, and having a full-time domestic helper was associated with higher personal accomplishment. For work- and patient-related factors, being a member of a teamlet was associated with higher personal accomplishment. Fewer years working in the profession and having a higher reported number of patients per week was associated with higher depersonalisation, although effect size for both were small. Compared to those who strongly disagreed that they perceived recognition by patients verbally or in writing, those who answered neutrally, agreed, strongly agreed, or who preferred not to answer this statement had lower associated depersonalisation and those who agreed or strongly agreed had higher personal accomplishment. Finally, compared to responses of “certainly not” in participants’ intention to quit in the next 6 months, most responses, including not knowing, being likely, or preferring not to answer were associated with higher emotional exhaustion (except “certainly yes” and “unlikely”) and higher depersonalisation (except “certainly yes”). With respect to lifestyle factors, more time spent exercising per week was associated with lower emotional exhaustion. Last, higher Brief Resilience Scale average scores were associated with lower emotional exhaustion and higher personal accomplishment. Discussion Summary of main findings Our study highlighted the extent of burnout faced in a population of polyclinic healthcare workers and identified multiple factors associated with the three domains of burnout. Higher emotional exhaustion was associated with doctors (compared to allied health), fewer hours of exercise, and lower resilience, while lower emotional exhaustion was associated with nurses (compared to allied health). Higher depersonalisation was associated with doctors and perceiving less recognition by patients. Lower personal accomplishment was associated with being female, not having full-time domestic help, not being a member of a teamlet, perceiving less recognition by patients, and lower resilience. Comparison of burnout with other studies Burnout scores using the MBI-HSS have mainly been assessed in tertiary institutions, where they showed comparable results to ours. This included a study in Singapore Health Services’ healthcare professionals [20] that showed a mean emotional exhaustion of 23.2 (SD = 13.0), depersonalisation of 7.2 (SD = 6.5), and personal accomplishment of 31.3 (SD = 9.5). However, the inclusion of non-clinical staff in this study decreased mean scores of burnout. Unlike tertiary institutions, polyclinic staff have regular hours and no night shifts, and patients’ medical conditions are not as severely acute. However, polyclinic staff face the challenge of managing patients holistically within a short consult duration. These differences make it difficult to compare burnout scores between the two types of work environments. Burnout had been assessed in Singapore’s GPs during the COVID-19 pandemic with the Oldenburg burnout inventory, where it found that 82.1% met the scale cut-off for burnout (2.25 for exhaustion and 2.1 for disengagement) [21]. However, it is difficult to compare the extent of burnout with ours as a different scale was used. Nevertheless, in our study, 9.21% of participants indicated “likely” or “certainly yes” in their intention to leave their currently position in the next six months, which is concerning. This figure could even be higher as there are also 8.77% of participants who indicated “I prefer not to answer”, which could conceal intentions to leave. Comparatively, there was only a 4% resignation rate among doctors, nurses, and allied health professionals from Singapore’s public healthcare institutions in the first half of 2021 during the COVID-19 pandemic [39]. As our study found a positive association between emotional exhaustion and depersonalisation with intention to quit, this signals a greater urgency to tackle burnout to reduce staff attrition. Compared to international studies, our finding that burnout is common in primary care aligns with what has been found in the United States, where the Medscape Physician Burnout & Depression Report 2024 found that a significant percentage (51%) of US physicians in primary care experienced burnout [40], although the tool used to assess burnout was not reported. Demographical factors associated with burnout domains In our study, men were found to have higher levels of personal accomplishment than women. Comparatively, in a systematic review of burnout among primary care providers in the United states [25], findings were mixed on which sex was associated with higher burnout. Another qualitative review of burnout literature [41] had also found that males scored higher in depersonalisation and females scored slightly higher on exhaustion. Possible explanations for our study findings could be that men find more satisfaction in their professional goals than women. Our study also found that those with full-time domestic help were found to have higher personal accomplishment. In Singapore, it is common to hire domestic help, which is more affordable than in Western countries. Future directions for research and practice: potential strategies to reduce burnout based on our findings At present, there are programmes that seek to reduce burnout in NUP, such as the availability of in-house counsellors and an online platform for patients to provide their healthcare workers with compliments. Our study highlighted four strategies which primary care institutions could potentially further adopt to improve burnout outcomes in healthcare workers – focusing on improving patient-provider relationships, improving work environment, increasing resilience in healthcare workers, and optimising team-based care models. As our study found that perception of recognition by patients (verbally or in writing) was associated with decreased depersonalisation and increased personal accomplishment, polyclinics could encourage patients to show greater recognition to healthcare workers, such as through awareness campaigns and getting patients to give regular words and notes of appreciation. As our study also found that a higher duration of exercise per week was associated with lower emotional exhaustion, primary care organisations could look into implementing exercise programmes at work, such as supervised yoga programmes [42] and team-based incentivised exercise programs [43]. Furthermore, our study found that higher resilience on the Brief Resilience Scale was associated with lower emotional exhaustion and higher personal accomplishment. This finding aligns with the results of a study in the United States, which found that physicians without overall burnout had higher mean resilience scores [44], although a different resilience measure (2-item Connor-Davidson Resilience Scale) was used. Thus, programmes that seek to improve resilience (e.g., resiliency training programmes, discussion groups, mindfulness [45]) may be considered by primary care organisations in improving burnout in healthcare staff populations. Additionally, our study found that being a member of a teamlet was associated with higher personal accomplishment. These findings corroborate with findings of some studies that isolation at work was associated with increased burnout [46] and working in a tight team structure and high team culture with lower exhaustion in clinicians [47]. However, one study found that physicians who do not practice in a teamlet or team had significantly lower rates of burnout [48]. The differences in findings could stem from variations in terms of team member composition and how teamlets function in Singapore’s public primary care compared to other countries and healthcare settings. Based on our study, a teamlet-based approach model not only provides multi-disciplinary collaboration and teamwork to improve the quality of care of patients, it could also be a potential strategy to manage healthcare worker burnout in polyclinics in Singapore. The study additionally revealed several socio-demographical factors associated with burnout that are non-modifiable (or difficult for polyclinics to modify), such as gender, occupation, and presence of full-time domestic help. Nevertheless, this information can be useful for polyclinics to be better aware of sub-populations at higher risk of burnout when devising preventive interventions to help ensure that these sub-populations will be able to benefit from them in priority. For future research, mixed-methods designs could be considered. Qualitative studies could help in understanding the complexities of the identified factors associated with burnout to better craft burnout-reducing interventions. Repeating the quantitative study could be useful for longitudinal monitoring of burnout scores when interventions are implemented and Healthier SG is more extensively rolled-out. Additionally, such studies could be conducted in other public primary care organisations for a more representative evaluation of burnout in polyclinic healthcare workers in Singapore. Strengths of the study This is the first quantitative study on burnout in primary public healthcare workers in Singapore post-COVID pandemic. Our study used a valid and reliable scale with three dimensions, which allowed a more complete and nuanced understanding of burnout than other scales. The broad and relevant range of factors investigated explained up to 57% of the variance of burnout domains in the final models which demonstrated the strength of the models. Lastly, the use of anonymous online questionnaires may have mitigated social desirability bias. Limitations of the study First, as this was a cross-sectional study, burnout was captured at a single time point without considering fluctuations in burnout levels over time. Second, a cause-effect relationship between the variables and burnout could not be determined in this cross-sectional design. Third, although universal sampling was used to reach out to all eligible staff, we could not exclude self-selection and non-response bias resulting from certain individuals being more inclined to answer questions about their experiences on burnout. Participants who completed the questionnaire may have different characteristics such as extent of burnout than those who did not complete the questionnaire. Fourth, a self-rated, anonymous questionnaire was used to collect information from participants which could have a risk of recall bias as compared to using objective measurements. Despite thorough consistency checks, it cannot be completely excluded that participants filled out the questionnaire twice with different demographic data or responded incorrectly in other ways. Lastly, although the measures included in the survey were based on an extensive literature review, it remains possible that some risk and protective factors were not captured by it. For instance, there were no questions asking participants about the presence of workplace bullying or the frequency of meet-ups with superiors. Factors typically linked to burnout in settings other than healthcare including screen time and enterprise community involvement were also not asked in our questionnaire [49]. Conclusion Our study highlights the extent of burnout experienced by Singapore public primary care healthcare workers in the domains of emotional exhaustion, depersonalisation, and personal accomplishment, as well as the associated factors from the various socio-demographical, work-related, patient-related, and lifestyle domains and personal attributes. These study findings potentially guide polyclinics and primary health organisations in designing interventions addressing burnout in healthcare staff. Supplementary Information Supplementary Material 1 Supplementary Material 2 Acknowledgements We would like to thank Mr Zachary Lim for assistance in statistical analysis. Authors’ contributions All authors made contributions to the study. P.Z.T was responsible for the conceptualisation, methodology, investigation, formal analyses, manuscript writing, fund acquisition, and project administration. A.S was responsible for the conceptualisation, methodology, supervision, and review & editing of the manuscript. Y.Y.T was responsible for the conceptualisation, methodology, formal analyses, and review & editing of the manuscript. L.H.G was responsible for the conceptualisation, methodology, supervision, and review & editing of the manuscript. All authors read and approved the final manuscript. Funding The study was funded by the National University Polyclinics research fund. Data availability The data underlying this work could not be made publicly available given the limited size of the sampled population (healthcare providers working at National University Polyclinics), which may enable reidentification based on the demographic and professional information collected. The data and survey instrument remain available from the first author upon reasonable request. Declarations Ethics approval and consent to participate The study was submitted and approved by the National Healthcare Group Domain Specific Review Board (study reference: 2023/00979) and adhered to the Declaration of Helsinki. Informed consent was obtained from all survey respondents. No personal identifiers were collected in the survey. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The data underlying this work could not be made publicly available given the limited size of the sampled population (healthcare providers working at National University Polyclinics), which may enable reidentification based on the demographic and professional information collected. The data and survey instrument remain available from the first author upon reasonable request.