Risk Stratification for ICU Delirium using Pervasive Ambient Sensing Information

2026-06-17

AI Summary

This study addresses the frequent neglect of environmental factors in early prediction of delirium in intensive care units (ICUs). It presents the first systematic validation that passively collected audio and light signals can independently predict delirium risk. Leveraging multicenter ICU data, the authors employ four efficient temporal neural networks—including convolutional architectures—to model these environmental inputs and apply SHAP for interpretability analysis. Results demonstrate that sound emerges as a key predictive feature, with convolutional models achieving an AUC of 0.80 on both acoustic-only and audio-light fused data. Notably, incorporating ambient light significantly improves short-term (within one week) risk stratification accuracy. This approach offers clinicians a non-intrusive, interpretable auxiliary tool for early delirium prediction.

Abstract

Delirium is a common and serious complication in the Intensive Care Unit (ICU), associated with increased morbidity, prolonged hospital stays, and higher healthcare costs. Despite its prevalence, early prediction and prevention remain challenging. Environmental factors such as ambient sound and light may influence the onset of delirium, yet they are often overlooked in risk assessments. In this study, we examined whether light intensity and sound pressure levels can independently predict delirium across multiple prediction horizons. We evaluated four efficient sequential neural network models on data collected from 9 ICUs across 309 patients to predict delirium for 10 prediction-window sizes. We reported feature importance and direction of influence using Shapley Additive Explanations analysis. The convolutional model achieved the strongest discrimination, with AUC = 0.80 on sound data and on combined data. Sound features were the dominant predictors overall. Integrating sound with light improved short-term ($<1$ week) prediction, with the combined model assigning the highest risk immediately after the sensing period. These findings suggest that passive ambient sensing, especially sound, can add a clinically meaningful, interpretable signal for delirium risk estimation and offer a practical pathway to enrich multimodal ICU prediction and prevention strategies.

Problem

• ICU delirium
• risk stratification
• ambient sensing
• sound pressure
• light intensity

Innovation

• ambient sensing
• delirium prediction
• sequential neural networks
• SHAP interpretability
• ICU monitoring