To develop an incentive model for GPs to test the performance and effectiveness of current incentives for GPs, so as to put forward suggestions to improve the mechanisms of employing and retaining GPs.

In August 2019, we recruited 204 GPs from 11 community health centers (CHCs) in Shanghai's Fengxian District using stratified random sampling to attend an online survey using a questionnaire named General Practitioners' Perception and Evaluation of Incentive Measures developed based on our previous research results and Herzberg's motivation-hygiene theory. Then from August to September 2019, we invited directors responsible for medical care or public health issues, and medical quality control department heads from the above-mentioned 11 CHCs to attend an online survey using a questionnaire named Workload for General Practitioners in Community Health Centers of Fengxian District developed by usbased on a literature review. Then under the guidance of Herzberg's motivation-hygiene theory, we constructed a structural equation to develop an incentive model for GPs within Fengxian District with representative incentive policies, systems and initiatives related to GPs selected from the survey results as latent independent variables to measure GPs' (responders') perception and assessment of incentives, and with willingness to work, service radius, and workload of GPs as dependent variables.

The average scores for the implementation of incentive measures, and its associated influence on GPs' work status assessed by the 204 GPs were (0.77±0.14) , and (0.73±0.19) , respectively. The assessment score for implementation of incentive measures for GPs differed significantly by personal health condition and administrative position in the respondents (P<0.05) . The assessment score for the influence of implementing incentive measures on GPs' work status differed significantly by personal health condition in the respondents (P<0.05) . Spearman correlation analysis showed that GPs' work status was associated with the implementation of 25 incentive measures (one incentive measure at the subdistrict level was not included for analysis) (P<0.05) . The top three most highly correlated incentive measures were performance distribution (r_s=0.652) , performance-based salary calculation (r_s=0.621) , and wages and benefits (r_s=0.614) . Partial correlation analysis indicated that, after controlling for variables such as the regional policy environment, the street and town government, and the work unit, the assessment score for implementation of either regional, or subdistrict or institutional incentive measures for GPs, was positively correlated with the assessment score for the influence of implementation of incentive measures for GPs on GPs' work status (r=0.381, 0.387, 0.528, P<0.001) . Theaverage assessment score for willingness to work by the respondents was (0.76±0.18) points. Theassessment scores of willingness to work by the respondents differed significantly by work unit (P<0.05) . Structural equation modeling revealed that the implementation of policy incentive measures influenced GPs' workload via GPs' willingness to work. Specifically, the implementation of policy incentive measures increased GPs' willingness to work (w1=0.43) , while GPs' willingness to work decreased with the increase of weighted workload (w2=-0.156) .

The GPs in Fengxian District showed higher level of willingness to work, namely, relatively high job stability, but they had not been incentivized by policy incentives to be competitive in obtaining improved performance, indicating that although policy incentives have produced partial effects on incentivizing GPs, the effects are still unsatisfactory. In the long run, it is necessary to consider whether there will be changes in GPs' retention status due to the lack of work initiative. For GPs, besides a guarantee of stable income, incentives to get more by doing more are also needed, which may be achieved by increasing the incentive authority and proportion of community health institutions when the total amount of incentives is limited.