Doctor Recommendations System (DRS) have been developed to increase the quality and efficiency of patients seeking medical service in Online Health Communities (OHCs). However, existing DRS did not take patients’ individual attributes into account, which are required for improving patients’ health outcomes under patient-centered care circumstances. To address this issue, this study develops a patient-centered DRS, in which the personalized characteristics and demands of patients are emphatically considered. To set up the DRS, this study draws on the Multidimensional Sequence Alignment (MSA) method measuring "biological distance" among doctors based on their patient’s and doctor-patient interaction attributes, which are identified by a combination of Levenshtein Distance model and opinion mining technology. Based on the results of MSA, 12 doctor recommendation sets with different characteristics are obtained from a sample of 241 doctors. XGBoost model is further applied to assess the importance of each attribute, which could be integrated to optimize the DRS. The results of performance evaluation show that the precision of patient-centered DRS is 72.6%, and the recommended doctors are significantly different from the doctors chosen by patients lacking medical knowledge. There are multiple benefits of the patient-centered DRS in assisting patients choose doctors, which could increase their satisfaction with consultation experience. When patients who arriving OHCs for the first time filling out their personal information, the DRS can provide them with a tailored doctor recommendation sets. The findings also suggest the importance of the doctor’s information completeness in increasing the probability of being selected by patients, and provide practical guidance for OHCs operators.