Role stressors and turnover intention among doctors in Malaysian public hospitals: work–family conflict and work engagement as mediators

Purpose – This study aims to investigate the role of work–family conflict (WFC) and work engagement (WE) and its influence on role stressors and turnover intention among medical doctors in Malaysian public hospitals. Doctors who experience higher work stress will inevitably experience WFC, consequently triggering their intention to quit. Design/methodology/approach – A total of 202 structured questionnaire responses were collected from medical doctors in four Malaysian public hospitals. The study used partial least squares structural equation modeling (SmartPLS 3.0) for hypotheses testing. Findings – As hypothesized, WFC encourages turnover intention while WE mitigate the relationship between role ambiguity (RA)-TI. RA is also observed to reduce WE, which, in turn, increases the intention to quit. Findings showed that both role conflict andWFC share a positive relationship with TI while WE and TI are negatively related. Furthermore, it was found that WE and WFC are significant mediators in the RA and TI relationship.WFC is also reported to be a significant mediator between the RC and TI relationship. Research limitations/implications – The causality effects remain limited due to the nature of the cross-sectional design. Future studies should use a longitudinal approach to gauge a better understanding of these relationships. Practical implications – This study provides insights for policymakers in resolving the increase of turnover issues by providing support and relieving medical doctors’ stress levels. Health directors should be encouraged to focus on the key aspects that may directly affect the well-being of medical doctors and eventually reduced staff turnover. Originality/value – The study contributes to existing knowledge by measuring variables such as job demand (RS and WFC), personal resources (WE) and job outcomes (TI) in the public health care sector. Additionally, research involving COR theory in Asian countries like Malaysia remains relatively underexplored.


Introduction
With the rise of emerging illnesses and a recent global pandemic, the burden of the medical staff's workload in the health care sector also increases. The health care sector remains one of the most critical sectors in confronting a dearth of labor shortage globally (Kirk, 2017). According to the International of Labor Organization (ILO), there was a 13.6 million gap in the global health workforce that was expected to fulfill the demands of health-care in 2015. A similar notion was reported by the World Health Organization (WHO) where a total of 27 million health care personnel are required by 2030 to fully cater to the unmet health care expectations. If this crisis remains unresolved, the likelihood of achieving both the Sustainable Development Goal (SDGs) and Millennium Development Goal (MDGs) seems far-fetched (ILO, 2017). At present, the turnover rate in the global health care industry is reported to be 20% (ILMIA, 2018), where the Malaysian health care sector alone accounts for 4% of the total amount. This issue has sparked the interest of many academic scholars and industry key-players. Health workers are often highly susceptible to stress and may succumb to burnout. This detrimental effect on their well-being is one of the contributing factors in medical staff resigning from their jobs, thereby increasing staff shortage when the workforce is already stretched. The situation is worsened by the constraint of the limited availability of current resources in preventing staff turnover.
Turnover has always been a perennial concern among managers while among scholars, it has been an actively studied subject matter for over a decade (Hom et al., 2017). The seminal work of Mobley (1977) classically described the term employee turnover as an experience to be voluntarily dismissed from current employment ties. The implications of increased turnover events are cumbersome considering the unnecessary costs incurred besides the productivity impairment and stress that could happen due to the shortage of manpower and inexperience substitutes (Hom et al., 2017). Turnover trajectories are often closely-associated with stressful working environments and taxing demands of the job. High work demand could hinder health care professionals from performing their role to the best of their ability and may consequently lead to stress. A higher work stress level could subsequently produce detrimental effects on the staff's health and well-being. Undeniably, health care workers are left to deal with role stress, lengthy working hours, increased job obligations, besides shouldering excessive workload (Hämmig, 2018). These job obstacles pose a potential threat to patients' care and safety. This adverse circumstance has led researchers to focus and emphasize the interactions between work role stressors and work environment besides investigating physical and psychological symptoms, as well as individual attitudes and behavior (Brough and Driscoll, 2005;Driscoll and Dewe, 2001).

Purpose
The novelty of this study lies in the conceptual connections between stress, work engagement (WE) and work-family conflict (WFC) in studies related to health care management practices and their impact on organizational outcomes (Meyers et al., 2019;Van De Voorde et al., 2016). This study examines how WE act as a plausible barrier in deterring employee turnover. Podsakoff et al. (2007) found that role stressors can exert adverse implications on performance and attitude besides impairing turnover intuitions. However, the relationship between role stressors and critical work outcomes from an empirical perspective remains unclear and needs to be further elucidated (Moura et al., 2014). Similarly, the direct impact of WFC on work stress or burnout has also been widely investigated. However, the effects of role stressors on WFC, WE and TI are still not fully explored and understood. Taking into account the numerous empirical evidence, this study establishes the link between role stressors and the tendency to quit via WE and WFC in the context of Asian countries.

Literature review
Role stressors as job demands and depleted resources Generally, workers are more exposed and disposed to role stress whether in the work or family environment. Therefore, it is not surprising that role stressors are central in most discussions on work stressors in the health care literature (Jensen, 2016;Portoghese et al., 2017). Role stressors are often identified by its two distinct variables, namely, role ambiguity (RA) and role conflict (RC). Collectively, both stressors are said to induce and exacerbate distressful emotions and energy depletion in the workplace (Eatough et al., 2011;Gilboa et al., 2008). The RA variable refers to one's belief that the job requirements were not clearly articulated and in turn, causes uncertainty in where to channel one's effort (Bedeian and Armenakis, 1981;Kahn and Byosiere, 1992). On the other hand, RC involves employees receiving incompatible or contradictory demands from different parties which may lead to confusion and difficulty in task accomplishment (Fisher and Gitelson, 1983). RC is correlated with frustration and quantitative workload (Che et al., 2017), burnout (Kleiner and Wallace, 2017) and performance (Arnold et al., 2009). The increased prevalence of new diseases and the low doctor-patient ratio often lead to the intense workload which makes work a strenuously demanding obligation for health personnel.
Considering employees' susceptibility to role stress in the contemporary work environment, it can be asserted that the critical focus should be on, both, the predictors and outcomes of role stress. Based on previous stress literature, the effects of role stressors on different outcomes and occasions demonstrated contradictory results (Bowling et al., 2015). Additionally, little empirical evidence was found on the underlying causes which linked these stressors to either individual or organizational outcomes (Karatepe et al., 2014), especially in the context of Asian countries where the collectivist culture is practiced. Undoubtedly, employees who encounter exposure to the high level of stress tend to suffer adverse effects on their health and psychological well-being (Larner et al., 2017;Tucker et al., 2005). Thus, this study assessed the role of stressors (RA and RC) in turnover intentions. The following hypothesis is proposed for this variable: H1. (a) RA is associated positively with TI, (b) RC is associated positively with TI.
Work engagement as personal and protective resources WE is a motivational construct which remained a scholarly interest over the past two decades (Schaufeli et al., 2006). The definition of WE, as accorded by Schaufeli et al. (2002, p. 74) mainly refers to "a positive, fulfilling work-related state of mind that is characterized by vigor, dedication and absorption." WE is a personal resource, which has been considered as one of the most crucial factors of well-being at the workplace (Bakker and Demerouti, 2008). The components of WE are relatively distinct, but they are generally correlated to each other. A component of WE is vigor, which is defined as the psychological endurance and elevated energy level during work, eagerness in investing effort at work and resilience when experiencing stress at work. In contrast, the absorption component of WE refers to the state of becoming fully focused and absorbed in one's work where it would be difficult to pry them away from work. The last component of WE, dedication, is a feeling of being important, passionate, motivated, dignified amid the struggle when at work.
Several scholars claimed WE as the inverse situation of burnout (Willard-Grace et al., 2019). In health care settings, medical personnel is expected to build rapport and caring connection with their patients. Being empathetic and engaging patients is one of the most important elements in their job scope. An engaged employee is often considered as the most prized asset in most organizations (Huynh et al., 2014). This notion is further supported by the fact that WE Role stressors and turnover intention potentially encourage vast improvement in organizations (Bailey et al., 2017). Although long working hours are deemed as detrimental, it does contribute to impactful work experience through WE (Dåderman and Basinska, 2016). The seminal work of Schaufeli and Bakker (2004) found that a declined in WE could further enhance the intention of leaving employment. Moreover, WE are also found to be negatively connected to turnover intentions (Saks, 2019). Another study by Coetzee and De Villiers (2010) revealed that RA causes a decline in WE after researching staff in a financial institution based in South Africa. Previous literature on role stressors was not able to provide conclusive findings on its impact on work outcomes. As the connection between role stressors and work-outcomes cannot be fully established, the existence of a mediating mechanism is possible (Hämmig, 2018;Montani and Dagenais-Desmarais, 2018). Greater attention needs to be given to this notion as there were mixed results reported concerning work stressors framework in the current health care literature. Previous studies have overlooked the role of WE in relation to role stressors-turnover intentions, as most researchers are inclined to study these variables independently (Vandenberghe et al., 2011). As such, the establishment of the empirical connection between role stressors and WE requires further investigation (Curran and Prottas, 2017;Olugbade and Karatepe, 2019). Based on the empirical and theoretical findings from past literature, it is anticipated that an increased level in each stressor will reduce the level of WE, thus, elevated turnover intention. Thus, this study predicts: H2. WE is negatively associated with TI.
H3. (a) WE is a significant mediator of the RC and TI relationship, (b) WE is a significant mediator of the RA and TI relationship.
Work-family conflict as a threat of resource loss WFC is an essential element to both individual and organizational outcomes. Studies over the past three decades demonstrated the damaging impact of WFC toward one's health and well-being (Frone et al., 1997). The negative repercussions of WFC often stemmed from taxing demands of work conditions which in turn, caused the decline in work outcomes (Sirgy and Lee, 2018). The present study adopts the unidimensional terms of WFC from the works of Netemeyer et al. (1996) which advocated WFC as a type of inter-RC where the common job demand, time spent and strain of work intervenes the fulfillment of family obligations. WFC usually happens when one is incapable of balancing the demands between work and family as these demands are incompatible. As proposed by the COR theory, WFC also potentially leads to resource depletion as a consequence of employees juggling both work and family responsibilities (Grandey and Cropanzano, 1999). A study by Michel et al. (2011), classified four distinct groups of work-related factors impacting WFC, namely, work stressors, work characteristics, work social support and work involvement. Of these four factors, work role stressors dominated the mainstream literature as it significantly impacts WFC. This finding is consistent with the work by Rubel et al. (2017), who denoted that the positive relationship between work role stressors and WFC is through stressorsstrain relationship, a predominant find in several stress models. Considering the significance of role stressors, the present study adopted two components of role stressors, namely, RA and RC due to the prevalent empirical and theoretical relationship these stressors share with WFC (Matthews et al., 2010a(Matthews et al., , 2010bStoeva et al., 2002). Besides indicating an inconsistency when it comes to the relationship between RS, WFC and TI, this literature also noted inconclusive deductions on the role stressors-work outcome. For example, Rubel et al. (2017) concluded that all the three role stressors (RA, RC and role overload) predicted turnover through a taxing WFC/ PRR FWC while Matthews et al. (2010aMatthews et al. ( , 2010b reported otherwise (RA shares a negative relationship with work-to-family conflict while RC shares a positive relation). These limitations warrant further investigation to understand the dynamics of the relationship between the said variables better. Hence, based on the findings of these studies, we postulated that: H4. WFC is positively associated with TI.
H5. (a) WFC is a significant mediator in the RC and TI relationship, (b) WFC is a significant mediator in the RA and TI relationship.

Theoretical frameworks
Hobfoll's conservation of resources theory Stressors' theoretical frameworks are frequently dependant on role theory. However, when it comes to studying WFC, the role theory may not be sufficient as role theory pay less focus to family roles, which an essential aspect in fully understanding WFC (Grandey and Cropanzano, 1999). Moreover, critics also argued that the role theory is unable to explain the motivational aspect of one's behavior and failed to examine the individual subjective experience (Kossek et al., 2017). In other words, role theory failed to explicitly elucidate why people behave the way they do besides neglecting an individual's experiences in attempting to amend the role's boundaries by changing current social practices. Recognizing the limitations of role theory, there is a need for an equally established and appropriate theoretical framework for the purpose of the present study.
The COR theory appears to be one of the most viable theoretical perspectives which could be used to comprehensively clarify the conflict relating to both, work and family aspects. Theoretically, this study attempts to address a few research gaps in the stressor framework using the COR theory. Developed by esteemed scholars (Hobfoll, 1989(Hobfoll, , 2011Hobfoll et al., 2018), the COR theory presents that stress is most likely to occur when there is a lack of available resources and when one is unable to conserve and sustain these resources. In this context, resources could refer to personal characteristics, various energies or states that are of significance or value to oneself. This theory states that people aspire to sustain, acquire and protect their available resources (quality of life, health, family life and well-being) from being depleted. Individuals would make effort in safeguarding these resources while establishing some new resources in a bid to diminish stress. The present study examines the personal resources component (WE) in the COR theory besides investigating the issue of WFC as job demands (resource depletion) increases psychological role stressors and weakened WE. The depletion of resources often takes place whenever employees are exposed to stressful work environment factors, consequently leading to withdrawal in behavior. This study extends the findings of previous literature on WFC and stress literature (Dettmers, 2017;Rubel et al., 2017). The proposed research framework as stipulated in Figure 1.

Role stressors and turnover intention
Methodology Sample and procedure Adopting a quantitative framework, a cross-sectional design survey was implemented as the instrument of the study while the systematic random sampling approach was used for participant selection. The survey was distributed to 400 participants, all of whom were medical doctors and specialists in 4 selected public hospitals in Malaysia. Seven departments took part in the survey, namely, the emergency and trauma department, the surgical department, the psychiatric department, the medical department, the orthopedic department, the obstetrics and gynecology department and the anesthesiology department. These departments were selected as there was a relatively high number of physicians available. After two weeks, follow up visits were conducted where 282 surveys were collected. There were a total of 202 responses available after the elimination of invalid questionnaires, charting a 50.5% overall response rate. The reported sample size exceeded the minimum requirement of the PLS-SEM software (Hair et al., 2014(Hair et al., , 2016(Hair et al., , 2019. Harman's one-factor test was implemented to test the common method variance (CMV). The findings revealed only 29.56% which means that the CMV is not a main concern in this study. Table 1 presents a summary of respondents' characteristics.

Measuring instruments
Each participant was required to complete the self-report survey online or manually using a pen or pencil. The survey comprises five main parts, namely, role stressors, WFC, WE, turnover intention and demographics. The first part of the survey and encompasses 14 questions on role stressors: eight questions on RC and six questions on RA. RS was measured using the scales found in a study by Rizzo, House and Lirtzman (1970). The second part of the survey and is related to WFC. This variable was gauged using three  Schaufeli et al.'s (2006) study. The fourth part of the survey involving six questions to measure turnover intention. These questions were adapted from a study by Bothma and Roodt (2013). The question format in the survey was in the form of a Likert scale with responses ranging from 1 (strongly disagree) to 7 (strongly agree). Table 2 presents the statistical description of the variables tested in this study.

Measurement model
PLS-SEM analysis begun with the testing of measurement traits of the outer model (Hair et al., 2019). The measurements that can be churned out by the Smart PLS ranges from the Cronbach's alpha to t-values. To test the router model, the constructs and their indicators were tested to establish the relationship between them. Indicators with loadings below 0.70 were eliminated after the examination of loadings was carried out (Hair et al., 2014). The items were deemed reliable as it achieves a significance level (t > 1.96) indicated by the standardized outer loading between 0.723 and 0.953. The internal consistency of the constructs was also deemed adequate as the composite reliability value was between 0.83 and 0.92 for all five constructs exceeded 0.70. On the other hand, convergent validity was also confirmed as the measurement model recorded an average variance extracted (AVE) above 0.60 for all constructs with a cut-off of 0.50 (Hair et al., 2014). The results of the measurement model as tabulated in Table 3.
As for the confirmation of discriminant validity, all AVEs were reported to have a higher value compared to the squared inter-construct correlations (Table 4) according to the Fornell-Larcker criterion (Henseler et al., 2014). The heterotrait-monotrait (HTMT) value for all indicators was also recorded to be below 0.90 (Table 5). These values confirmed the significance of the discriminant validity for all the indicators in this study.

Structural model
The important components in the assessment of the structural models are the size and significance of the path coefficients, the level of the coefficient of determination (R 2 ) values and the predictive relevance as measured by Q 2 (Hair et al., 2014). Based on Cohen's (1988) study, the R 2 values in this study were greater than 0.10 (TI = 0.310 (substantial); WE = 0.111 (weak); WFC = 0.181 (moderate). R 2 reflects the cumulative effects of exogenous latent variables on the endogenous latent variable where its value should be sufficient for a model to be deemed usable for the explanation (Urbach and Ahlemann, 2010). In this case, the target value should equal to or greater than 0.10 for R 2 to be considered sufficient (Falk and  Notes: * Correlation is significant at the 0.05 level (two-tailed); ** Correlation is significant at the 0.01 level (two-tailed) Role stressors and turnover intention Miller, 1992). On the other hand, all constructs show Q 2 values more than 0, indicating their predictive relevancy. A Q 2 value that is larger than zero signifies that the path model is relevant to the endogenous constructs and reflective indicators. Q 2 or predictive relevance is usually calculated using the blindfolding technique. Another important value that needs to be taken into consideration is the f 2 values which indicate the significance of the exogenous construct in relation to the endogenous construct. In this study, the f 2 values mostly showed small and medium effect sizes (Table 6).

PRR
Once the reliability and validity of all the constructs were confirmed, the results of the structural model were examined to extract occurring patterns in the data, thus commencing the hypotheses testing. The relevant components that were tested include the size and significance of the path coefficients which represented the hypotheses. The bootstrapping analysis of 5,000 subsamples was conducted to identify the levels of significance of the hypotheses (Hair et al., 2016). Figure 2 shows the final structural model used for hypotheses testing. The summary of findings for the hypotheses testing is shown in Table 7.

Discussion
The findings of the current study provide an improved insight into the role of stress in triggering employees' turnover intention and aligned with the postulation of the COR theory (Hobfoll, 1989). Furthermore, these findings also corresponded to the studies by scholars (Vandenberghe et al., 2011b;Hwang and Kim, 2017;Wen et al., 2020) where they found a positive correlation between RC and TI. Although the current study discovered that RA does not share a relationship with TI, a study done by Kim et al. (2015) echoes the same sentiment where RA did not significantly influence turnover intention. This inconsistency may be due to the difference in how employees perceived the notion of RC and RA. The similar results on RA could be attributed to the fact that individuals are not diligent enough in identifying details, which can aid in lessening ambiguity. Besides that, the inability of

Role stressors and turnover intention
organizational surroundings in providing clear strategic objectives could also be a contributing factor to the consistency results reflecting RA (Kammeyer-Mueller and Wanberg, 2003). In regard to the link between WE and TI, results demonstrated that WE have a significant negative effect on TI, thus decreasing it. Despite the consistent role of WE in relation to the other stressors, its role as a mediator was not as predicted in the hypotheses of the current study. For example, it was observed that the transmission of the RA effect on TI was done by WE where it serves as a mediating role. This not the same in the case of transmitting the RC effect. Findings also reported that only RA decreases WE, subsequently elevating the levels of quitting intentions. The COR theory and the principle of the spiral of loss resources support these findings. As the depletion of family resources may induce future loss, WE plays an important role in enhancing employees' well-being in and outside the workplace. Another study by Babakus et al. (2016) found that both challenge and hindrance stressors are considered triggers in employees' intention to quit their jobs. The findings of this study are also supported by the works of Karatepe et al. (2018) where they revealed that hindrance stressors (RC and RA) are often damaging on the outcomes (e.g. quitting intentions). Their study also reported that the mediating role of WE is a potential solution in resolving the impacts caused by other stressors. The mixed results on the role of stressors, WE and job outcomes imply that further research in this area can be done to arrive at a more conclusive finding.
As for the impacts of WFC on turnover intentions, findings were reported to be in line with previous studies (Amstad et al., 2011) and consistent with the expectations of the current study. WFC is observed to hold a positive impact on turnover intention among medical doctors. This implies that when health workers are highly stressed, they are more prone to conflict in handling patients. This strenuous circumstance can eventually lead to job resignation as the burden becomes too heavy to bear. This result conforms with Dåderman and Basinska (2016), who demonstrated the positive relationship between WFC, quantitative workload and quitting intentions. The works of Nohe and Sonntag (2014) and Chen et al. (2016), also denoted the impact of WFC only on turnover intentions as FWC is found to have no relation. Another study by Wang et al. (2018) study among employees in the automobile industry reported that a work-family climate which focused on sacrifice and sharing increases the level of work attention among employees. As a result, employees may experience a higher level of WFC which subsequently, leads to higher number of staff leaving their jobs.
Overall, the findings also highlighted the positive impacts of RS on TI as mediated by WFC. The accumulation of stress among employees is undoubtedly straining work and family obligations. The damaging effects of stress are often exhibited physically,  (Rajkonwar and Rastogi, 2018;Wang et al., 2017;Wen et al., 2020). Studies on the interactive effects relating to job conditions postulated that workplaces, which provide supportive resources to cope with stress can counter the effects of role stressors (Wu et al., 2012). On the same note, Choi (2008) highlighted the effect of family demands on personal stress where it was found that work demands affect personal stress through WFC. Medical staff experience WFC as a result of the failure of meeting their work and family obligation. This inability is often due to the depletion of valuable resources where they find their time and energy completely spent. Dealing with high work demands is most likely where they used most of their resources. With no other support or ways to replenish their resources, the development of the conflict between work and family roles is bound to happen. Additionally, Jain and Nair (2016) also discovered that WFC plays a mediating role in the relationship between work demands and turnover intentions which aligned with the works of Netemeyer et al. (1996) where WFC was found to be a mediator of the RA and RC effects on TI. The theory of Hobfoll et al. (2018) states that when employees encounter resource depletion caused by RS, employees can undertake certain strategies to preserve their resources and better deal with the overlapping work and family responsibilities. Some of these strategies include WE a potential factor in buffering against WFC, thus lowering turnover intentions.

Practical contributions
This study benefits both health care management and staff. Based on the findings of this study, health care managers should introduce initiatives that can strengthen the balance of both work and personal life for their employees. By focusing on family-centered practices such as the provision of child-care at the workplace and the implementation of flexible working hours (e.g. reducing on-call time), the impact of RS and WFC can be alleviated. With long work-shifts, health workers often dedicate a large amount of time to their work, thus, allowing them more time to fulfill their personal needs is crucial. This strategy is particularly useful for employees who are free from family commitments. A variety of programs should be made available to cater to the different needs of individual employees.

Limitations
There are a few limitations reported in this study. First, as this is a quantitative study, the results may not reflect the necessary justifications and genuine perspectives of the health care employees. Hence, future studies can use in-depth interviews or a diary study for a deeper exploration of the issue. The testing of the research model for validation in other relevant sectors should also be conducted. Next, the sample in this study is cannot represent the viewpoint of public hospitals as there were only 4 selected public hospitals in the Klang Valley. The cross-sectional design of the current study also posed a limitation in the data collection method as this implies that the causality effect is limited. Therefore, future studies should use a longitudinal design approach to rectify this shortfall. This study can also be extended to include managers and upper management, as the focus of participants was only on doctors who directly handles patient care.

Conclusions
This study elucidated the factors that impact RS and TI, consequently providing insights for both academic researchers and health care researchers in overcoming the longstanding issues of high turnover rates. By understanding the role of RS in inducing WFC, better working conditions and proper support can be made available to health care staff. These resources ensure that staff continues to provide exceptional care to their patients and perform their job well, maintaining a competitive advantage in the health care sector. Furthermore, this study also included the understudied role of WE on individual well-being, fulfilling a research gap. Additionally, this study also contributes to existing knowledge by: Conducting the study from the Asian perspective, covering the Eastern context; and Further supports the existing theory of conservation resources of Hobfoll et al. (2018), particularly on the loss of resources.