生成式人工智能辅助决策对住院肺炎患儿预后的影响

Abstract: Objective To investigate the association between guardians’ use of generative artificial intelligence (AI) and timeliness of medical consultation and inpatient outcomes in children hospitalized with pneumonia. Methods A retrospective analysis was conducted on 320 pediatric patients aged 7 to 12 years hospitalized for pneumonia from October 2024 to March 2025. The guardians were divided into the AI group (n= 158) and the non-AI group (n= 162) based on whether they actively used AI tools such as DeepSeek. We compared length of stay (LOS), timely admission within ≤48 hours after symptom onset, and time from defervescence to discharge between groups. Multivariate linear regression analysis was performed to determine the independent effect of AI use on hospital stay duration after adjusting for age, sex, and admission severity. Results The timely consultation rate within 48 hours was significantly higher in the AI group compared to the non-AI group (75.0% vs. 29.4%, P＜0.05). The median Los was 9 days in the AI group was 11 days in the non-AI group. The length of hospital stay in the AI group was shorter than that in the non-AI group (P＜0.05). After adjustment, AI use remained independently associated with shorter LOS (β=-1.696, P= 0.007), whereas admission severity was the strongest positive predictor (β= 2.629, P＜0.05). Conclusion Guardians’ use of generative AI is associated with improved timeliness of care and shorter LOS, suggesting potential value of reliable AI tools in optimizing pediatric pneumonia management.