Frontiers in Public Health · Oct 31, 2026
From Multimodal Sensing to Intervention: A Critical Narrative Review of Sleep, Physical Activity and Diet
By Xuan Dong, Wenmin Li
Background: Multimodal sensing and supporting records can relate sleep, physical activity and diet within shared observation periods. Their contribution depends on how measurements inform prediction and intervention. Objective: To synthesize evidence linking multimodal measurement, prediction and intervention, distinguishing tested connections from conceptual ones. Methods: This critical narrative review used purposeful, non-exhaustive selection informed by PubMed and Web of Science searches. XD performed study selection and data extraction; WL independently verified the study selection and extracted evidence. Results: Reviewed studies documented concurrent three-domain observations and task-specific measurement and prediction. Eating-event sensor fusion increased sensitivity at a cost in precision; dietary validity depended on the target and reference method. Prediction studies used internal participant or record splits, including temporal testing, and independent-cohort evaluation. Input availability and evaluation populations shaped interpretation. Randomized trials compared adaptive goals, messaging policies, dietary strategies and supported programs. Adaptive goals better maintained activity than a fixed target; adaptive messaging improved step trajectories versus random selection among participants meeting recording requirements. Personalized dietary targeting reduced glycated hemoglobin (HbA1c) and time with sensor glucose above 140 mg/dL more than a Mediterranean diet; another trial's primary weight-loss comparison against a low-fat diet was nonsignificant. These contrasts concerned different populations and outcomes. Conclusion: The review connects measurement targets, information windows and intervention comparisons along the sensing-to-action pathway. Author-proposed frameworks organize the selected evidence around inputs, decisions and delivery conditions. Matched model comparisons, policy trials and service evaluations address predictive gain, strategy effects and sustained outcomes, harms and resource use, respectively.