Objective:To explore the effect of family doctor system on the risk stratification of community hypertensive population by application of Markov modelMethods:Retrospective investigation was conducted on hypertensive patients under continuing management from 13 community health service centers in Shanghai Xuhui District from January 2014 to December 2016. Among 98 996 subjects, 50 920 (51.45%) were contracted to family doctors (contracted group) and 48 046 (48.55%) did not contracted to family doctors (non-contracted group). According to the risk stratification of hypertension, the four-state Markov model (low-risk, medium-risk, high-risk, and extremely high-risk) was established. The prediction effect of the model was validated, and the changes in the risk stratification status of hypertension in the study subjects from 2017 to 2020 was predicted by using the Markov model.Results:Among all subjects the number of medium-risk and extremely high-risk accounted for the majority(>80%). Compared with 2014, in 2016 the number of low-risk patients with hypertension was decreased from 9 042 cases (17.76%) to 6 851 cases (13.45%) in contracted group; and from 9 971 cases (20.75%) to 7 906 cases (16.46%) in non-contracted group; the number of people at extremely high risk of hypertension was increased from 15 609 cases (30.65%) to 17 639 cases (34.64%) in the contracted group; from 13 847 cases (28.82%) to 15 641 cases (32.55%) in the non-contracted group. According to the Markov model one year after the risk stratification, the risk status of most subjects remained in the original one. There was not transform from extremely high-risk to low-risk state (0%), but there was transform from low-risk to extremely high-risk state in some extend, and the degree of transform in non-contracted group [2.06%(205/9 971)] was higher than the contracted group [1.85%(167/9 042)]. Predicted by the Markov model, between 2017 and 2020 the number and proportion of the medium-risk>extremely high-risk>low-risk>high-risk in both contracted group and non-contracted group. With the extension of time, low-risk proportion is gradually reduces, and the rate of reduction of the contracted group was lower than that of the non-contracted group, while proportion of medium-risk, high-risk and extremely high-risk is gradually increased.Conclusions:The constructed Markov model is accruable and stable, which can be used in the study of hypertension prognosis. The study indicates that the contracted services of family doctor have positive effects on the management of community hypertensive patients.