The study developed an artificial intelligence algorithm to predict the need for endotracheal intubation (EI) in neonates up to 3 hours in advance in the neonatal intensive care unit (NICU). The model uses multimodal deep learning to analyze numerical clinical data and vital signs time series from the previous 1–3 hours. The goal is to enable proactive intervention planning and prevent EI delays that may cause adverse effects. Internal validation showed accuracy 0.9579 and AUC 0.9323, external validation accuracy 0.9411 and AUC 0.9336. The model predicts the need for EI up to 72 hours in advance in 1-hour intervals with high reliability and generalizability. This tool has wide applicability in NICU clinical practice for efficient care.