The conventional research method of patients’ medical choice behavior in online health community mainly uses CooSeeker software to positively adjust the disease risk, which is vulnerable to changes in the cross frequency of high and low risks, leading to low indicators of medical choice robustness. Therefore, a new research method of patients’ medical choice behavior in online health community needs to be designed based on multivariate data analysis. That is, we extracted the topic of patients’ medical choice in online health community, constructed a research model of patients’ medical choice behavior in online health community, and designed a perception architecture of patients’ online health data using multi-source data analysis, thus completing the research on patients’ medical choice behavior in online health community. The results of the experiment show that the designed research method of online health community patients’ medical choice behavior based on multi-source data analysis has a high index of medical choice robustness, which proves that the designed research method of online health community patients’ medical choice behavior has good research effect, reliability, and certain application value, and has made certain contributions to reducing the risk of patients’ medical choice.