[Artificial intelligence-based literature data warehouse for vaccine safety].

Objective: To establish a sustainable updated literature data warehouse for global vaccine safety assessment, and provide data support for evidence-based vaccine safety assessment. Methods: Semi-automated construction and updating of a literature data warehouse were achieved through the continuous integration of standard operating steps of evidence-based reviews with artificial intelligence technologies. Following the standard procedure of a systematic literature review, the literatures about vaccine safety assessment published before November 29, 2020 were retrieved from 9 databases including OVID, Scopus, Web of Science, Cochrane Library, and ClinicalTrails.org in English and Wanfang, CNKI, VIP, and SinoMed in Chinese. Literatures were screened for two rounds in a semi-automatic manner (by artificial intelligence literature processing system and manual work) according to the inclusion/exclusion criteria. Furthermore, the literatures were classified according to the types of vaccines and adverse events. The updating strategy was established, and the literature data warehouse was updated regularly. Experts were organized to select specific vaccine safety topics and carry out special demonstration studies. Results: More than 0.41 million articles were retrieved. According to the inclusion/exclusion criteria, 23 304 articles were included after two rounds of screening. At present, we have selected and completed three prior topics as demonstration studies, including the systematic review of "DPT (diphtheria, pertussis and tetanus) vaccine and encephalopathy/encephalitis", and the classified management of literatures about allergic purpura and brachial plexus neuritis. Conclusions: The sustainable updated literature data warehouse of vaccine safety can provide high-quality research data for vaccine safety research, including evidence support for immunization related policy-making and adjustment and vaccine safety-related methodological research or clinical tool development; and further demonstration studies can provide references for building a new methodological framework system for timely and efficient completion of the evidence-based assessment of vaccine safety.