AI-driven early detection of severe influenza in Jiangsu, China: a deep learning model validated through the design of multi-center clinical trials and prospective real-world deployment

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background: Influenza-related global deaths reach 650,000 annually. The current highly lethal clinical subtype of influenza is severe influenza.Aim: To develop and validate a deep learning based model for early diagnosis of severe influenza.Methods: This is a multi-centre, double-blind, multi-stage, randomised controlled clinical trial. We initially developed a framework for a 5-phase study: model development, external validation, multi-reader study, randomised controlled trial and prospective validation. The data source for the preview programme is electronic health record data from 87 hospitals in Jiangsu Province from 2019 to 2025.Significance: Our expected result is that the developed model of severe influenza can be more accurate and have a lower misdiagnosis rate than traditional clinical assessment. The pre-specified AUC was 0.18 (95% CI: 0.14-0.22), with an expected 32% reduction in misdiagnosis. The model’s performance was consistent across patients in older adults, underlying disease, and resource-poor areas. The added value of the study is that it is effective in improving early recognition of severe influenza.Ethics and dissemination: This study was approved by the Institutional Review Board of Yangzhou University Hospital (IRB No. YKL08-002). Written informed consent was obtained from all participants. The results of this study will be disseminated in the form of a conference in the Jiangsu Province area, which will facilitate the translation of clinical research results and provide a powerful decision-making tool for the precise prevention and control of severe influenza.Clinical trial number: https://clinicaltrials.gov/, identifier (ChiCTR2000028883).

This publication has 35 references indexed in Scilit: