Objective To explore the health science popularization talent training mechanism for district-level public hospitals from the perspective of regional collaborative governance, and to explore the embedding path of medical-engineering integration tools such as information dissemination engineering, human factors engineering and health information technology within this system. Methods Taking Baoshan District of Shanghai as the research case, this study comprehensively employed methods such as literature research, policy text analysis, questionnaire surveys, and semi-structured interviews to conduct a survey of 327 medical staff from 4 district-level public hospitals (312 valid questionnaires, with an effective rate of 95.4%) within the jurisdiction. Simultaneously, in-depth interviews were conducted with 12 directors in charge and key personnel in health science popularization. Readability metrics and information entropy were used to measure the information adaptability of popular science content, and Monte Carlo simulation was adopted to test the robustness of expected outcomes. Results Baoshan District has initially established a health science popularization practical pattern featuring “district-level coordination, professional supply, community undertaking, and integrated dissemination”. However, the survey revealed that only 38.8% of the medical staff participated in science popularization-related training in the past year, and the average time spent on science popularization creation by medical staff per week was less than 1.5 hours. There were significant differences in the health science popularization capabilities among doctors, nurses, pharmacists, and public health personnel. Currently, there are problems such as scattered science popularization data, the absence of interface standards, and the lack of objective behavioral data for effect evaluation. Structural contradictions such as unstable coordination among the main bodies, inaccurate supply-demand matching, a lack of training systems, incomplete incentives, and insufficient digital-intelligent support remain prominent. Conclusion A four-in-one health science popularization talent training mechanism consisting of “organizational collaboration-stratified trainingdissemination evaluation-guarantee support” should be constructed. The compliance review and public opinion risk control modules for science popularization should be integrated. With digital and intelligent technologies such as the district-level data center, audience profiling, intelligent review and simulation training as the continuous support, an institutionalized closed-loop operation system should be formed.