The initial focus that gave rise to modelling in this study was the social exchange theory developed by Homans in 1961. Social exchange theory is a theory in social science that states that in social relationships, there are elements of reward, sacrifice and profit that interact. This theory explains how humans perceive relationships with each other based on their perceptions of (1) the balance between what is given in the relationship and what is excluded from the relationship; (2) the type of relationship that is carried out; (3) how humans relate to the society (between communities), where small groups are households and have better relationships with others (Meira & Hancer, 2021). Social exchange theory includes the basic theory of rationality, more integrative exchange theory, network exchange theory and rational choice theory. Although there are various modes of exchange, Homans focused his studies on dyadic exchange (Cropanzano & Mitchell, 2005). This theory underlies the model of antecedents and consequences of employee engagement that is applied in this study.

Work engagement is determined by a variety of factors, including job characteristics, perceived organisational support (POS), perceived supervisor support (PSS), perceptions of distributive justice and perceptions of procedural justice, all of which predict a variety of outcomes, such as job satisfaction, organisational commitment and organisational citizenship behaviour directed at either the individual or the organisation (Saks, 2006). Employees determine their level of involvement in their work and organisation based on the resources available to them. The antecedent is a resource in the model, and when employees receive it, they will respond with a higher level of engagement (Saks, 2019). Employees believe they are compelled to put more effort into their work because of the rewards they receive from their employer. The amount of cognitive, affective and physical resources an individual is willing to dedicate to the performance of their function is determined by the organisation’s economic or socio-emotional resources (Žnidaršic & Bernik, 2021).

The Utrecht Work Engagement Scale (UWES) has been utilised in the majority of WE studies. According to a recent evaluation of engagement research, UWES was used in 86% of investigations (Bailey et al., 2017). This makes it impossible to compare Saks’ (2019) findings to those of other engagement research, and it raises doubts about the model’s validity and generalisability.

Based on the review and comparison conducted by Saks through the revision of the method in 2019, the overall results of measuring WE using UWES are the same as the results of his prior study, which can provide some support for generalising its findings to UWES WE measures. However, UWES is related to all consequences, and UWES mediates the relationship between antecedents and consequences as well as WE (Saks, 2019).

Engaged workers have superior performance and financial outcomes as a result of their great dedication and focus on their work activities. Furthermore, engaged workers have more creativity and are more inclined to innovate and be entrepreneurial because of their openness to new ideas (Gawke et al., 2017). In addition to these results at the individual level, research has indicated that workers with high WE are more likely to assist their coworkers. Work engagement is favourably associated with team performance at the team level (Saks, 2019). Because engagement spreads from one person to the next, it has a significant impact on the team.

Little is known about the WE of emergency physicians, although their key role can determine the quality of services provided. Lack of job support, long working hours and challenges in managing life-threatening cases are associated with increased stress and lower job satisfaction (Shams & El-Masry, 2013). The authors applied the revised Saks’ (2019) antecedent model to examine the factors that could have relations with the WE of emergency physicians.

Physician autonomy (PA) implies that physicians should have full freedom to treat patients according to their best judgement (Zhang et al., 2021). Physicians lose their professional autonomy under managed care. Professional (clinical) autonomy is compromised under managed care, such as frequently having to seek consent before starting treatment, prescribing only authorised drugs and following a specific treatment plan for a particular disease. Things like the above will greatly affect the WE of the physician concerned about carrying out his or her obligations in the hospital.

Public service motivation (PSM) is defined as an individual orientation, in this case, that of emergency physicians who work in both district and private hospitals, towards providing services to people with the intention of doing good for the larger community (Palma & Sepe, 2017). Public service motivation has been associated with major positive consequences, such as organisational citizenship behaviour as described by Koumenta (2015), organisational commitment as studied by Naff and Crum (1999) and job satisfaction as described by Steijn (2008). A meta-analysis by Cerasoli et al. (2014) showed that performance improvements and increased probability of performance were more visible with the help of intrinsic motivational sources than extrinsic motivators. According to Petrovsky and Ritz (2014), there is a significant positive relationship between PSM and organisational performance.

Work–life balance (WB) as a form of work flexibility provides a balance between the professional and personal life of the workforce, which leads to high performance and the overall improvement of the organisation (Davidescu et al., 2020). Job adjustment theory states that job flexibility leads to higher engagement and increased job performance. If Bal and De Lange (2014) proved that the relationship between work flexibility and work performance was mediated by labour commitment, it also showed that time flexibility significantly affects labour productivity.

A good and healthy relationship with colleagues (RC) can lead to social support provided by the organisation and is described as interpersonal exchanges in which one individual helps another individual in the organisation, as seen in the interspeciality relationship that is well established within the pillars of the medical profession. Social support is comfort, attention, appreciation or other forms of assistance that individuals receive from other people or groups. Peer support is positive feelings, trust and attention from others in the individual’s work life, recognition, one’s trust and direct assistance in a certain form (Vassos et al., 2019).

Perceived organisational support for employees is an effort to realise welfare and various awards that can be given for contributions and evaluations of employee performance (Atingabili et al., 2021). Based on this understanding, employees who receive POS will positively perceive the support, and vice versa, POS will give a negative evaluation of employees that both determines WE and employee performance, including physicians (Imran et al., 2020). Therefore, the physician’s perception of the support provided by the company becomes interesting to study more deeply because it can determine how the physician behaves at work.

Several efforts to motivate and improve employee performance are through rewards and recognition (RR). If the remuneration is arranged correctly and fairly, accompanied by good rewards, the employees will feel satisfied and motivated to achieve the goals and objectives of the organisation. The level of remuneration of workers will determine the scale of life and, in turn, will show the dignity and social status in society. Therefore, if employees think that the salary they receive is inadequate, performance satisfaction decreases, they are not motivated and this ultimately affects work performance, so organisational goals are not achieved (Rehman et al., 2010).

Jain and Khurana (2017) argue that employees possessing the required knowledge and skills are essential for any organisation seeking growth and success in business. In today’s dynamic business scenario, the opportunity to learn and develop (OLD) has gained importance in keeping employees engaged and maintaining a competitive advantage. Proper learning and development plans to nurture talent can increase engagement rates. Khan et al. (2016) interpret training and development and job satisfaction as important parameters of employee performance. They tried to study the effect of training and development on employee performance through job satisfaction. A sample of 105 employees of a telecommunication company in Pakistan was collected and analysed. The results showed that there was a positive impact of training and development on employee performance. Thus, they concluded that by investing in training and development programs, WE among employees would potentially increase.

Physicians who spend more time completing paperwork (PW) and have a bigger administrative workload have poorer job satisfaction (Ly & Glied, 2014). There is also evidence that using an electronic health record is linked to reduced professional satisfaction and increased burnout (Shanafelt et al., 2016). Administrative work, on the other hand, can be beneficial, as there is evidence that telephone and electronic communication with patients can reduce office visits and enhance patient outcomes (Apaydin, 2020).

Physical, psychological, social and organisational components of a workplace that are good or functional in attaining work goals, minimising working demands and accompanying physiological and psychological costs and encouraging human growth, learning and development have been classified as hospital resources (Bakker & Demerouti, 2007). Only physical resources (hospital facility [HF]) were investigated in this research. A literature review indicates that job resources have been found to mediate the relationship between behaviour and organisational workforce performance (Shin & Hur, 2019).

Extending the aforesaid reasoning, the authors propose the following hypotheses:

There has been no research linking WE and PC. Despite minimal evidence in healthcare, previous research has linked WE to higher performance in a variety of industries. Engaged health workers are more likely to be genuinely driven, achieve their goals and learn from their mistakes, which has been linked to organisational-level quality outcomes. Physician engagement is considered to be higher in the healthcare industry than in other industries, and hospitals with more physician involvement deliver better patient care and have a better safety record than those with low WE (Wake & Green, 2019).

Although there is still limited research on this subject, the historical relationship between job satisfaction, WE and service quality was reported in a study by West et al. (2011), where patient satisfaction, hospital mortality rate, infection rate, absenteeism, staff turnover and annual health check-up ratings were significantly impacted. Based on the explanation above, the authors established the following hypothesis:

From the signified literature reviews mentioned above, the authors have proposed the research framework as illustrated in Figure 1. In this model, there are 11 variables with 10 paths marked with arrows to describe the research hypothesis.

A quantitative survey with a cross-sectional approach for collecting the primary data was applied to answer the research questions. The conceptual framework was empirically tested with a purposive sampling technique on the population obtained from nonmanagerial emergency physicians who have worked for at least a year in the district and private hospitals in East Nusa Tenggara in March 2022. The authors explained the main objective of the research to the physicians to obtain permission for the data collection.

The survey continued for 4 weeks, and a total of 190 responses were gathered. A few respondents did not meet the inclusion criteria, because of their position that also doubled as managerial staff in the hospital. After elimination, 183 eligible samples were collected. This amount met the criteria of the minimum sample requirement based on guidance for analysis with partial least squares structural equation modelling (PLS-SEM) (Kock & Hadaya, 2018). The respondents participated anonymously and voluntarily, in which they were convinced of the confidentiality of their responses.

A set of indicators in a structured questionnaire instrument are used to measure the constructs in the proposed conceptual framework. The questionnaire utilised in this study was developed from a previous study and then tweaked to fit the needs of the investigation. Translators have already translated this questionnaire into the local language to guarantee that all questions are easily understood. Before being sent, this questionnaire was examined by specialists in the field of service marketing. On a scale of 1–5, respondents were asked to rate their level of agreement with the assertions on a scale of 1 (strongly disagree) to 5 (strongly agree), as specified in the questionnaire. The questionnaire for PA was adopted from Fida and Najam (2019); PSM from Zubair et al. (2021); WB from Irawanto et al. (2021); RC, PW and HF from Mehta (2020); POS from Atingabili et al. (2021); RR from Baqir et al. (2020); OLD from Vuong et al. (2021); WE from Lovakov et al. (2017); and PC from Srivastava and Singh (2020).

The conceptual framework consists of 11 constructs and is considered a complex research model; therefore, the PLS-SEM method was preferable as it could analyse complex models in exploratory research. Partial least squares structural equation modelling approaches are preferred when the orientation of the study is more on the explanatory abilities of the model (Hair et al., 2019). The PLS-SEM analysis was applied through SmartPLS version 3.3.8 (SmartPLS GmbH, Oststeinbek, Germany), which was selected as it provides a bootstrapping menu to test significance (Memon et al., 2021). The main procedure with PLS-SEM is based on two types of models, namely measurement and structural models. The measurement model is established to measure reliability and validity between indicators and their respective constructs in the model. The reliability testing phase includes indicator reliability (outer loading) and construct reliability (Cronbach’s alpha and composite reliability). The validity testing phase includes construct validity (average variance extracted [AVE]) and discriminant validity (heterotrait–monotrait [HT–MT] ratio). If these four things have met the requirements of reliability and validity, it can proceed to the next stage. The structural model is deployed to test the significant relationship between each construct in the research model. Further, a mediation analysis will be conducted to evaluate the variables that act as a mediator in this model. A more advanced PLS technique, in this study, importance–performance map analysis (IPMA), is utilised for more specific managerial implications (Ringle & Sarstedt, 2016).

The Ethics Committee of the Universitas Pelita Harapan, with regard of the protection of human rights and welfare in research, has carefully reviewed the research protocol including the information given to the potential subjects entitled: ‘Linking Work Engagement to Patient Centricity of Emergency Physicians in Underdeveloped Regions’.

The Department of Hospital Administration, Universitas Pelita Harapan, approves the above-mentioned protocol including the information given to the potential subjects (ref. no. 002/MARS-FEB-UPH/I/2022).

The demographic data of the 183 eligible respondents are represented in Table 1. The majority of the respondents were male (52.5%), while women represented 47.5%. Based on age group, most of the participants were from the age group 20 to 29 years (64.5%). Also, most of the participants had 1–5 years of job experience. In terms of working duration, most of the respondents had 40 h – 80 h per week.

The outer loading from the reflective model was done to assess the indicator of reliability. The results of the analysis show that all 59 research indicators meet the outer loading criteria, with a loading value above 0.708. From the internal consistency test, all constructs showed Cronbach’s alpha greater than 0.7 and composite reliability as an upper threshold ranging between 0.7 and 0.95 (Hair et al., 2019), indicating the constructs’ reliability of the respective model. Assessment of the convergent validity was done by measuring the AVE. This validity check shows that each construct has an AVE ≥ 0.50 as required (Hair et al., 2019), showing all constructs can explain at least 50% of item variance in the model, thus establishing convergent validity. The results of the reliability and validity tests can be seen in Table 2. A list of questionnaire statements concerning the indicators can be seen in Appendix Table A1.

The final step of the measurement model analysis is to check the discriminant validity by HT–MT ratio. This approach was used as it is known to have a more precise value (Hair et al., 2019; Henseler et al., 2014). The recommended threshold value for HT–MT ratio is below 0.9, referring to Hair et al. (2019) to establish that each construct indicator is conceptually different. Table 2 shows that all HT–MT values are well below the 0.9 thresholds; thus, it is concluded that all indicators used in this research model have adequate discrimination to measure their respective constructs. This measurement model analysis had passed the four parameters of the reliability and validity test sequentially. Therefore, it can be concluded that all indicators in this research model are reliable and valid to measure their respective constructs specifically. The model fit indices in this study were assessed by standardised root mean square (SRMR), and the value obtained was 0.06, lower than 0.08, which was the required value, indicating a good model fit (Sarstedt et al., 2022).

This study performed coefficient determinant or R² to measure accuracy and Q² to measure the cross-validated redundancy of the model. Priorly, the inner variance inflation factor (VIF) test was conducted to check multicollinearity issues. The findings showed all the constructs had inner VIF below 5 as suggested (Hair et al., 2019); thus, it can be said that there is no multicollinearity issue found in this model. The result found that PC had R² = 0.713 and was categorised as moderate accuracy of the estimation (Hair et al., 2019). Thus, it can be said that the respective model has adequate capability to estimate PC. Work engagement had R² = 0.579, indicating that it had a moderate accuracy. Furthermore, WE had a large effect size on PC with an f² value of 2.484.

The out-of-sample redundancy value was applied from the blindfolding feature of PLS-SEM (Hair et al., 2019). All the Q² are found > 0, whereas PC shows Q² = 0.597 and was categorised as large value (> 0.5). Work engagement has Q² = 0.494. Therefore, it can be said that endogenous constructs in the out-of-sample model approach have sufficient cross-validated redundancy.

Hypothesis testing by bootstrapping feature was conducted to determine the association of the variables in the model and confirm whether the hypothesis proposed in this study was supported. The bootstrap approach was used to establish the significance of the data analysis in PLS-SEM (Memon et al., 2021). The cut-off value of T-statistics > 1.645 (one-tailed with alpha 0.05) and accordance with confidence interval (CI) 5% and CI 95% direction with the directional hypotheses were used as a criterion to determine whether the hypothesis is significant (Hair et al., 2019; Sarstedt et al., 2022). The results are shown in Table 3.

It can be seen in Table 3 that there are eight hypotheses supported by T-statistics > 1.645 and p < 0.05, with positive or negative direction (coefficient, CI 5% and CI 95%) following the directional hypotheses. The other two hypotheses such as H6 and H9 were not supported. The coefficient of PA to WE was 0.474, greater than others in estimating WE. Given its predominant relationship with WE, the authors did a subgroup analysis of PA with the other variables in this study and found that differences in employee length of service led to differences in PA in its contribution to WE and vice versa. Emergency physicians with more than 5 years of service had a standardised coefficient of PA 2.98 times higher than those with less than 5 years of service (0.763 vs. 0.256). Because of its prominent relationship with WE, the authors did a subgroup analysis of PSM dividing the respondents into two categories, high PSM and low PSM, based on the average of their scores in the concerned variable. Emergency physicians with high PSM had a standardised coefficient of PA 6.11 times higher than emergency physicians with low PSM (0.757 vs. 0.124). This shows the contribution of PSM to PA in estimating WE. Moreover, it was found that POS and RC were higher in respondents with high PSM. R² for the WE variable in respondents with high PSM was higher than low PSM (0.854 vs. 0.382), which was categorised as substantial strength of relationship, giving the interpretation that this research model was more accurate when applied to respondents with high PSM.

Furthermore, mediation analysis was also carried out to determine the mediation significance through the specific indirect effects, as recommended by Nitzl et al. (2016). Based on the result of mediation analysis in Table 4 as below, WE as the mediator construct tested had T-statistics above 1.645 thresholds for the supported hypotheses. This result indicates that WE is proved as a significant mediator towards PC.

Importance–performance map analysis is a useful tool to identify indicators, providing input to managers to prioritise their improvement activities (Ringle & Sarstedt, 2016). This method is based on the importance that resulted from the total effect and performance based on the mean value. Importance–performance map analysis could be seen as four quadrants, whereas the focus is more on the quadrant with indicators that have more importance, whether the performance is sufficient or vice versa. Figure 2 shows the mapping to identify the respective indicators that need to be maintained or improved.

This graphic illustrates an indicator to which hospital management must pay more attention. Indicator PSM4, which contains respondents’ responses that serving public health would give them good feelings even if they were not paid enough for it, is not sufficient. Therefore, this matter needs to be prioritised by policymakers and hospital management because it is considered important for emergency physicians but has not shown adequate performance.

As seen in the data analysis from SmartPLS 3.3.8, the empirical model can be described as the following (Figure 3). Work engagement and PC have moderate strength of relationship. Therefore, this research model can be suggested for research on WE and PC of emergency physicians.

In this study, several research limitations can be identified as a concern for further research with similar topics related to WE and PC. Firstly, the sample of this study came from one province with a variety of hospital types and numbers of emergency physicians in each hospital, which affected the duration of work and work-related stress faced by each respondent who participated in this study. Therefore, it can be suggested in further research to take the respondents from the same hospital type and scale, so that the sample will be more representative with more stringent criteria. Secondly, the majority of respondents in this study were 20–29 years old. This can affect the results because of different perceptions. Therefore, it is recommended to involve respondents of more diverse ages to get more accurate results in future research. Thirdly, this study has limitations in the sampling method carried out through an online questionnaire, which was thus conducted because of the COVID-19 pandemic. Questionnaires conducted online have a weakness where the condition of the respondents cannot be known with certainty. Emotional factors or the respondent’s mood when filling out the questionnaire can affect the results. A suggestion for further research is for data to be collected or questionnaires to respondents face to face, while still implementing community health protocols.

Based on the results of the subgroup analysis in this study, testing the model using the length of service as control or moderating variable is also suggested, considering the differences in the results obtained in respondents based on the length of work, both on intrinsic and extrinsic factors affecting WE. Moreover, PSM can be beneficial as one of the antecedents of WE in future research, because this study has proven the importance of PSM in creating WE in health workers. Furthermore, more research is needed to analyse the relationship between PC and patient experience to assess differences in service quality from the perceptions of health workers and patients. Finally, this research model can be used for other research involving more developed areas to assess the relationship between WE and PC.

Eight out of 10 hypotheses were proven to be supported with significant and appropriate directions of relationship. This research model has moderate strength of relationship and large cross-validated redundancy on PC as the dependent variable. Therefore, this model is replicable for further research with different populations. Based on the results, PA and PSM are two factors with the most predominant positive relationships on WE of emergency physicians, followed by POS, RC, WB and OLD. Paperwork showed a negative connection that weakens the WE.

Hospital management needs to pay attention to the well-being of emergency physicians, particularly in making decisions according to their respective competencies, which will have an impact on experience and confidence in serving patients. It also needs to look at the work motivation of emergency physicians, especially during the recruitment and selection process, and needs to assess whether the personal values of the employees concerned are consistent with organisational goals. To reduce the potential for disengagement because of PW, hospital management can apply to the health office to employ administrative staff who specifically work on administrative matters related to claims or counting consumables in the unit concerned, providing a computerised system in the form of digital applications to simplify the tasks. At the level of policymaker, the local public health officials can evaluate the emergency physicians who will be or have been placed in the area concerned through the assessment of PSM. Maintaining the quality of the existing WE antecedents can support the achievement of an optimal PC.