Nocturnal enteral feeding and mechanical ventilation risk in intensive care unit patients: A deep Learning–Based causal inference study

Background & aims: Continuous enteral nutrition is widely used in critically ill patients, but its clinical superiority over intermittent feeding remains uncertain. A particular concern is nocturnal feeding, a common component of continuous regimens, which may elevate the risk of aspiration due to diminished airway protective reflexes during sleep. This study investigated whether nocturnal enteral feeding is associated with a higher requirement for mechanical ventilation in intensive care unit (ICU) patients. Methods: We analyzed Medical Information Mart for Intensive Care (MIMIC)-IV (n = 1551) for model development and eICU Collaborative Research Database (eICU) (n = 3394) for external validation, including those who began enteral feeding within 72 h of ICU admission. The exposure was any enteral feeding between 10 PM and 6 AM in the first 72 h; outcomes were mechanical ventilation after 72 h. We used multivariable regression models. Subgroup analyses, including tests for interaction, were performed across patient characteristics, feeding volume, and comorbidities. A deep learning-based causal inference model was developed and estimated average treatment effects, and SHapley Additive exPlanations (SHAP) analysis identified influential predictors. Results: Night-time enteral feeding was associated with higher odds of requiring mechanical ventilation, with an odds ratio (OR) of 1.88 (95 % confidence interval [CI] 1.17–3.01) in the MIMIC-IV cohort and an OR of 2.45 (95 % CI 1.01–5.95) in the eICU cohort. The deep learning–based causal inference model estimated that night-time feeding increased the probability of mechanical ventilation by 7.29 % (95 % CI 5.37–9.37) in MIMIC-IV and by 3.50 % (95 % CI 2.66–4.40) in eICU. SHAP analysis consistently identified total daily enteral feeding volume as the most influential predictor of ventilation risk. Subgroup analysis revealed that patients without cerebrovascular disease experienced higher ventilation risk with night-time feeding, whereas those with cerebrovascular disease did not. Conclusion: Among non-intubated ICU patients, night-time enteral feeding may increase the risk of mechanical ventilation, particularly at very large feeding volumes. These findings suggest reconsidering overnight continuous feeding practices and support prospective trials of circadian-aligned nutrition to reduce respiratory complications.