The Correlation Between Work Motivation and Turnover Intention Among Primary Health Workers in China

Background: High turnover intention, as a manifestation of low work motivation, is a crucial barrier to strengthening primary health systems worldwide, including in China. Targeting those being less motivated will be a realistic choice to retain primary health workers. This study translate, adapt, and validate the Work Motivation Scale for Health Workers (WMSHW) scale to directly measure and rate health workers’ motivation composition based on Self-Determination Theory, and assessed how health workers with different levels of motivation being associated with the turnover intention. Methods: The process of translation and cross-cultural adaptation followed the recommendations of international guidelines. Participants include 1341 health workers within 75 primary health institutions from 6 provinces in China. The reliability and validity of the scale was analyzed. Cluster analysis in a person-centered approach and logistic regression analysis was used to understand how different combinations of motivations related to intention to leave. Results: Conrmatory factor analysis indicated that the modied ve-factor model had a better t than the other models in accordance with the original English version. The factor loads were high and ranged from 0.70 to 0.9(cid:0)Cronbach’s alpha coecients for ve dimensions of the Chinese WMSHW ranged from 0.81 to 0.94, indicating the scale’s high internal consistency. Four distinct clusters of work motivation were found in this study, representing low motivated, highly controlled, highly autonomous and highly motivated primary health workers. Compared with low motivation group, both controlled and autonomous motivation groups were more likely to have lower turnover intention. The negative relationship between motivation and intention to leave became stronger with the level of motivation increasing: highly controlled cluster (OR: 0.47; 95% CI: 0.35-0.63), highly autonomous (OR: 0.27; 95% CI: 0.18-0.41) and highly motivated (OR: 0.20; 95% CI: 0.15-0.27). Conclusion: The Chinese version of WMSHW showed satisfactory reliability and validity and can be used as an instrument for measuring and rating the work motivation of Chinese health workers. The primary health workers were grouped into four motivation levels based on this scale. Both controlled and autonomous motivation could work in reduce the turnover intention, and the inuence of autonomous motivation on retaining was stronger.


Introduction
The availability of motivated health human resources is critical for the health system to high-quality health care [1]. Just as a common problem worldwide, the low performance of primary health system has been recognized as the bottleneck problem in improving the overall health system performance in China.
Work motivation is de ned as the drivers of workers for choice of action and persistence in doing it [2].
The drivers working with environmental factors determined a psychological process that energizes, directs and sustains work behaviors [3,4][], and "be motivated" manifests as a status "the willingness to exert efforts at work towards certain goals" [5]. In health sector researches, it is widely accepted that the health workers' motivation was signi cantly related to the choice of work behaviours, including whether to retain in existing position and whether to perform as the desire of managers or supervisors [6][7][8]. In China's primary health system, attraction and retaining of PHWs is always a challenge, and turnover intention of PHWs is higher than that in hospital health workers [9]. One systematic review showed that the pooled prevalence of turnover intention was 30.4% [10], and the Health Workers Survey under the Sixth National Health Service Survey in 2018 showed that 55.3% of the investigated primary health workers have a middle or higher level of intention to leave [9]. Turnover intention is one predictor of turnover behaviour so it was reported that the mobility of health workers in primary health institutions (PHIs) is high and those leaving PHIs are more likely to be those with higher professional title level and longer working experience [11].
Health workers' willingness to stay is the basis for better performance. Studying the relationship between intention to leave and its predictor -motivation, will be closely relevant to health policy makers and managers of PHIs to take more targeted interventions to retain health workers. There have been some studies exploring the relationship between determinants of motivation and the attrition or retention of health workers in different countries [6,12,13], and from investigations in rural and urban PHIs in China, these factors included income and payment methods, learning and training opportunities, promotion and personal development space [14]. These ndings can potentially help policy makers to provide relevant incentives for retaining PHWs. However, providing these incentive factors usually involves extra resources or system reforms, which may not be applicable.
An alternative is to transfer the focus of interventions from factors to individuals, to measure the individual motivation level and composition, and then to help with the design of incentive tailored to individuals [15,16]. However, studies directly analyzing the relationship between health workers with different level of motivation and intention to leave are still lacking. The key reason is the lack of measurement tools for work motivation [17]. Most of the existing measurements of motivation are to indirectly measure the external or internal in uencing factors of motivation [18], or measure the affective or cognitive outcomes of motivation as a proxy, like job satisfaction [19,20]. There are few tools to directly measure the drivers of work behavior. Based on the de nition of motivation as a driver of behavior, some people may more likely be motivated by the enthusiasm and interest of the task, while others may more likely be motivated by external pressure and bene ts. Different sources and levels of motivation may contribute to completely different outcomes.
This study aims to use a scale to measure motivation directly and assess relations between the PHWs with different motivation levels and intention to leave in China. This study will contribute to existing knowledge in two aspects. Firstly, a well-structured and psychometrically sound tool to measure the work motivation of Chinese health workers could help further studies on predictors and outcomes of motivation in Chinese health care settings. To our knowledge, there are no validated instruments in China to measure health motivation among health workers. Secondly, this study will nd the relationship between different compositions of motivation and intention to leave of PHWs, which is of great value to policy makers in designing targeting strategies for health workers with different drivers.

Theoretical basis
Self-Determination Theory (SDT) provides a way to measure motivation itself and distinguish different motivation compositions. This theory was rst described by Edward Deci and Richard Ryan in the mid-1980s, who conceptualized motivation as a continuum of self-determination representing different motivation dimensions with regulatory styles [21,22]. The rst dimension of motivation, which is labelled as intrinsic motivation, mainly relates to pure enjoyment inherent in the activity itself. The second dimension of motivation is labelled as integrated regulation, which is driven by the congruence between one's behavior and values, identity and needs. The third dimension of motivation is labelled as identi ed regulation, which is derived from a conscious valuing of goals and values. The fourth dimension of motivation is labelled as introjected regulation, in which behaviors are undertaken to attain ego involvement or to avoid guilt such as pride. The fth dimension of motivation is labelled as external regulation, which occurs when behaviors are performed for external rewards or punishments [23].
SDT has become the mainstream motivation theory in the elds of education and sports and other life domains accompanied by the booming of well-validated motivation scales development based on SDT [24][25][26][27]. In terms of work domain, the rst motivation scale -"L' Inventaire des Motivations au Travail de Blais" (Blais Inventory of Work Motivation) based on SDT was developed in France [28]. There has been a growing trend in the development and evaluation of work motivation measurement such as Work Extrinsic and Intrinsic Motivation Scale [29] and Multidimensional Work Motivation Scale [30]. Most recent studies focus on the whole employees, while less attention is paid to the occupational particularity of health workers. The Work Motivation Scale for Health Workers (WMSHW) is a 15-item multidimensional scale developed to measure work motivation composition, which has been explicitly designed for use in the medical context [31].

Measurement of motivation
Considering the specialty in healthcare contexts, WMSHW has been developed tailored for health workers and has previously been applied and tested in 1142 nurses from government health institutions [31]. All respondents were asked to answer the guiding question: "Why are you motivated to work?", to capture their underlying motivation composition for hardworking. The original version of the WMSHW is a 15-item inventory that covers and assesses ve dimensions of work motivation: 3 pertaining to intrinsic motivation (e.g. "Because I enjoy doing what I do at work every day"); 1 to integrated regulation (e.g. "Because being a health worker is a fundamental part of who I am."); 2 to identi ed regulation (e.g. "Because my work is extremely important for my patients."), 2 to introjected regulations (e.g. "In order to feel good about me.") and 7 external regulations (e.g. "Because my supervisor recognizes and appreciates me." or "In order to earn money"). All items are rated on an 11-point Likert scale ranging from 0 ("No, I disagree completely") to 10 ("Yes, I agree"). The ve types of regulation represent different levels of motivation sources, which were distinguished by the extent to which they stem from contingencies outside the person (controlled motivation) or originate within the person (autonomous motivation) [32].
The WMSHW has only been available in the English language to date, there is a need to adapt and validate this tool within a Chinese-speaking population to investigate different compositions of work motivation of health workers.
Translation and cross-cultural adaptation were carried out following the previously published international common guidelines [33]. To ensure an accurate translation process, the scale was translated in 4 steps: forward translation from English into Chinese, achieving consensus, backtranslation into English, and pretesting on health workers. Firstly, two bilingual academics who are familiar with health human resource management translated the original English version into Chinese respectively. Secondly, the English and the Chinese versions were discussed by the expert committee to obtain an exact translation. Thirdly, two language specialists who were blinded to the original English version translated the Chinese version into English respectively. Then, a panel of experts experienced in pediatric nursing was then asked to compare and examine discrepancies between these two versions (original English and back-translated English versions) until a uni ed Back translation scale formed through consensus discussion. Lastly, the Chinese version of the WMSHW was nalized after being welltested in a representative piloted group of forty health workers.

Measurement of turnover intention
The turnover intention of health workers was measured by the scale developed by Mobley et al. [34] and Cammannet et al. [35], and has been widely used in the National Health and Services Survey in China [36]. This Turnover Intention Scale consisted of four items including thoughts of leaving the organization you served now; thoughts of leaving a career in the medical profession; the possibility of nding other jobs recently; and the possibility of looking for new jobs next year. All items are rated on a 6-point Likert scale ranging from 0 ("No, I disagree completely") to 10 ("Yes, I agree"), with a higher score indicating stronger intention to leave. The total score of turnover intention was rated by adding the scores in all 4 items ranging from 4 to 24. Additionally, the internal consistency of the Turnover Intention Scale was tested using Cronbach's alpha coe cient (α = 0.90).

Study design and participants
A cross-sectional study was conducted for data collection in six provinces throughout the eastern, central and western regions of mainland China. A strati ed cluster sampling method was applied in this study. Data were collected from 76 primary health institutions (47 in urban area and 28 in rural areas). All health workers present on the day of the investigation were invited to participate. Participation was informed that they were voluntary to participate in the study and their responses were anonymous questionnaires.
The nal sample was made up of 1341 health workers comprising 60.2% of health workers were selected from the eastern region, 9.9% from central and 29.9% from the western region of China. Informed consent from PHWs was obtained.

Statistical analysis
Quantitative variables for scale description were expressed as item means, standard deviations (SDs), skewness, and kurtosis. The structural validity and internal consistency of the WMSHW Chinese version were carried out to assess the psychometric properties of the tool. Con rmatory factor analysis (CFA) with maximum likelihood was conducted to test how well the items adequately represented their intended dimensions of the work motivation. Goodness-of-t indices were assessed using the chi-square (χ 2 ), degrees of freedom (df), the comparative t index (CFI), the root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), and Akaike information criterion (AIC). For the CFI indices, values close to or greater than 0.95 are taken to re ect a good t to the data. RMSEA values of less than 0.08 indicate a satisfactory t, with values ranging to 0.10 representing a mediocre t [37]. SRMR values of less than 0.08 indicate a good t [38]. Pearson correlation coe cients were referred to in order to explore intercorrelations among subscales. Internal consistency estimates for both total scale and subscales were evaluated using Cronbach's alpha. An alpha value between 0.70 and 0.90 was considered good and values greater than 0.90 were considered excellent [39].
Due to the scarcity of studies on the motivation level category of health workers, cluster analysis, an analytic tool of person-centered approach was applied to explore the motivation pro les of PHWs in China. In contrast to the variable-centered approach, this approach aimed to understand how a health worker standing on multiple motivations might be related to turnover intention, instead of how particular work motivations relate to turnover intention. We used logistic regression models to explore the in uence of work motivation on turnover intention among primary health workers in China.
Statistical analysis was conducted using STATA 14.0 for Windows, while CFA was performed through Mplus software version 7.0 [40]. All two-tailed p values less than 0.05 are considered statistically signi cant.

Description of participants in this study
The demographic and job characteristics of 1341 participants are presented in Table 1  Description of scale measuring work motivation   CFA was conducted to assess how well the three hypothesized models represented the data using the maximum likelihood estimation method. The goodness-of-t indices for the three models are present in Table 3. We found low goodness-of-tness indices for both four-factor structure and two-factor structure, while the ve-factor model yielded acceptable t indices: χ 2 (80) = 1015, p < 0.001, CFI = 0.94, RMSEA = 0.093, and SRMR = 0.059. The result revealed that ve-factor model proposed by the original authors [31] was accepted as the best-tting model compared to the other two models.  As seen in Figure 1, all standardized factor loadings were over 0.5 and statistically signi cant, ranging from 0.69 to 0.94. The correlation among ve factors ranged from 0.31 to 0.81, indicating that they are related, but measure different dimensions of work motivation. More speci cally, the correlations between each subscale followed the theory hypothesized as the less nonadjacent subscales are less strongly positively related.

The correlation between work motivation and turnover intention
We used cluster analysis to group PHWs based on the composition of regulation types. The result of cluster analysis was shown in Figure 2. To facilitate the naming of different groups, we grouped IM and IDEM as autonomous motivation, and group INTRO, EXT-S and EXT-E as controlled motivation. Participants in cluster 1 presented low levels on all dimensions of work motivation compared with other clusters, which could be described as "low motivated" workers. Participants in cluster 2 presented low levels of autonomous motivation and middle levels of controlled motivation, which could be described as "highly controlled" workers. Participants in cluster 3 presented middle levels of autonomous motivation and low levels of controlled motivation, which could be described as "highly autonomous" workers. Participants in cluster 4 presented high levels on all dimensions of work motivation, which could be described as "highly motivated" workers. Each cluster accounted for 19.6%, 32.2%, 10.1% and 38.1% of the PHWs respectively.
In Table 4, the relationships between the work motivation and outcomes turnover intention were tested. It was found that was four distinct clusters of work motivation differently related to turnover intention. The highly autonomous (OR: 0.27; 95% CI: 0.18-0.41) and highly motivated (OR: 0.20; 95% CI: 0.15-0.27) clusters were strongly related to low level of turnover intention, while highly controlled cluster (OR: 0.47; 95% CI: 0.35-0.63) had less favorable levels of correlates.

Discussion
In the present study, we sought to examine the correlation between the motivation of PHWs and their intention to leave, at the same time to examine the psychometric properties of a Chinese translation of the WMSHW. The current results provide supportive evidence regarding the cross-cultural reliability and validity of the inventory implying that WMSHW opens a new eld of research in relation to work motivation of Chinese health workers by introducing a scale that measures motivation compositions and levels. The data show that the majority of PHWs investigated had a moderate level of motivation with controlled motivation dominantly existed, and 65% of PHWs were with moderate or high level of turnover intention. It is also found that higher autonomous motivation and those being highly motivated by all types of drivers are more likely to be with lower turnover intention.
This study is the rst time to measure the work motivation of PHWs in China as the drivers of working hard and ensuring the validation of scale is the basis for further exploration of the correlation between motivation and intention to leave. The translated and cross-culturally adapted WMSHW demonstrated good structural validity and internal consistency among Chinese sample. We used the well-established translation and back-translation guidelines to assure equivalence in varied aspects between source and target versions. During the translation of the English WMSHW into Chinese, a minor cultural discrepancy was encountered and one item of the WMSHW has been modi ed accordingly: 'supervisor' was replaced with 'leader' as this word is closer to Chinese culture. Compared with previous studies that measure work motivation, this is the rst study to con rm the structure of WMSHW with the population working at health institutions in China without sacri cing any item.
The result of CFA showed that the modi ed ve-factor model had a better t than the other models and was selected as the best-tting model, which was consistent with the original scales. In line with a previous study testing the latent structure of work motivation based on SDT in seven languages [30], t indices in the Chinese version demonstrated slightly lower than the original English version. Culture may have played a role in shaping the understanding of work motivation. Although values of RMSEA are barely within the scope of adequate model t, our option was not to modify the model considering that the current results suggest that the ve-factor model may be a theoretically adequate model for the data based on SDT. Therefore, the current results showed that a Chinese translation of the WMSHW demonstrates a ve-dimensional factor structure representing IM, IDEN, INTRO, EXT-S, EXT-E as hypothesized. Speci cally, the IDEN dimension combines integrated regulation and identi ed regulation which has proven di cult to separate in previous research [30,41].
In line with previous studies [29,42], the correlation between the two factors was identi ed as moderate, indicating that they are related although they measure different compositions of work motivation.
Furthermore, strong internal consistency of the Chinese WMSHW was demonstrated, as evidenced by all Cronbach's alpha coe cients over 0.8. In the Chinese WMSHW, the Cronbach's alpha for the total scale (0.94) was much higher than in the English (0.89) versions, and Cronbach's alpha for the ve subscales was also higher than in the English version (0.92) [31].
This study rstly used the WMSHW to measure and categorized PHWs into four types based on motivation composition. As explained by SDT, the self-determination continuum of motivation includes ve major dimensions of drivers covering from controlled motivation to autonomous motivation. So we rated four types of groups as PHWs without any controlled or autonomous drivers for work (low motivation), PHWs with more external controlled drivers (controlled motivation), PHWs with more internal autonomous drivers (autonomous motivation), PHWs with strong drivers from both controlled and autonomous factors (highly motivated). Previous studies [43,44] mainly used job satisfaction as a proxy to measured and categorized the motivation of PHWs. The boundaries between motivation and job satisfaction measurements are blurred in existing studies. The reasons include: rstly, the in uencing factors of job satisfaction have a lot of overlap with determinants of motivation [45,46]; secondly, job satisfaction is an effective consequence of motivation, and a status of harmony between drivers' of health workers and organization environments [5]; lastly, scales directly measuring motivation per se as the drivers of working hard are lacking. The measurement of job satisfaction cannot substitute the measurement of motivation. Compared with the category of being satis ed or dissatis ed, categorybased WMSHW can identify complex compositions of drivers and diverse levels of motivation.
It was found that the health workers with low motivation, controlled motivation, autonomous motivation and highly motivated (higher scores on both autonomous and controlled drivers) took 19.6%, 32.2%, 10.1% and 38.1% of the total number of participants respectively. Based on the assumption of STD, human beings tend to make decisions by themselves and health workers with autonomous drivers, i.e. doing what they love and can make decisions for themselves, so those with the highest autonomous motivation are the most motivated compared with health workers with other drivers or those without drivers [7]. So the four groups of health workers provide a possible method to order the motivation level from the lowest level (low motivation group) to the highest level (highly motivated group). We found that  [49]. The dominant role of existence need can also be re ected by the ndings of this study: health workers mainly driven by controlled factors took 32.2% of investigated PHWs in this study; at the same time, in the driving forces of highly motivated participants (38.1%), the impetus of controlled factors was also as strong as the forces of autonomous factors; the sum of the above two groups means that the majority of participants are in the stage being driven by monetary rewards and other forms of material bene ts.
The data analysis showed that compared with the "low motivation" group (those having no drivers to work hard), the groups with any type of motivation, including controlled, autonomous, both controlled and autonomous, were all more likely to have lower turnover intention. Though there are no studies that directly measure motivation for PHWs in China, previous studies using job satisfaction as a proxy to measure reported similar ndings: lower satisfaction as an important indicator of low motivation, was correlated with higher turnover intention no matter in China and other LMICs [10,50,51].
Autonomous motivation is more effective in reducing turnover intention compared with controlled motivation, and this result implies those PHWs with real interest or pure enjoyment in the activities related to primary health care have less intention to leave the existing work position. As agreed by SDT and other motivation theories, being motivated is a status of harmony between personnel goals and organization goals, so those PHWs with natural enthusiasm on their tasks are nearer to the status of harmony even under disadvantaged environments. Autonomous motivation can help health workers keep optimistic in work, and cope with the feelings of dissatisfaction, stress, and burnout. Though there are no similar studies on the relationship between different compositions of motivation and turnover intention of health workers, the studies in other industries have found that employees with the pursuit of individual development, interests and achievement were more likely to have positive feelings compared with those with nancial bene ts and prestige as major drivers [52,53].
We also found that controlled motivation can work in reducing the extent of turnover intention compared with those without any driving forces. In this study, the PHWs in the "low motivation" group are those having the lowest score on any drivers of working hard, which means those people having no purposes or desires, thus they usually do not expect any achievements from their efforts or do not have con dence on the competency of themselves. The PHWs with no motivation usually perform as having no willingness to make efforts in work and responding passively to external environments. PHW workers with controlled motivation, though they are rated as the lower level of motivation in the STD continuum of motivation, their pursuit for the basic living substances can work to some extent in retaining them in the existing work position.
The present study has several limitations that should be noted. First, the Chinese WMSHW has been tested only among health workers working at primary health care facilities, so further psychometric testing in secondary-level and third-level hospitals and other health settings would contribute to reject or con rm the generalizability of the instrument. Secondly, although our study translated and adapted the English-language scale into the Chinese language using standard procedures, the cultural differences between Western countries and China may limit the ability to measure work motivation in Chinese cultural contexts. Thirdly, because limited studies assessing the motivation level directly are published for us to compare, our ndings on categories of PHWs based on motivation level, and the relationship between motivation and turnover intention are mainly discussed based on motivation theories.

Conclusion
In general, cross-cultural translation was completed, and the Chinese version of WMSHW showed good reliability and validity. The self-determinant perspective provided insights into a broader set of measure work motivation composition that are associated with work behavior. Speci cally for turnover intention, we found that both controlled and autonomous motivation can work in reducing turnover intention, and the in uence of autonomous motivation on retention was stronger. To retaining PHWs, the interventions should be tailored to different workers with varied motivation compositions, those measures inspiring the interests and enthusiasm of primary health workers may have stronger effects.

Declarations
Ethics approval and consent to participate The study has been approved by the Ethics Review Board of the School of Public Health, Peking University (code of ethics: PKU201412128). Informed consent was obtained from all participants prior to questionnaire administration. Standardized parameter estimates for ve-factor model Legend: Coe cients attached to arrows pointing at items from the right: residual variances (i.e. non-explained item variance, error terms); coe cients attached to arrows from factors to items: factor loadings; factor variances were xed to 1 for model identi cation; coe cients attached to curved arrows represent factor correlations