AI-based alert and decision-support system for malnutrition risk and nutritional intake monitoring using nursing documentation of eating

PI-Rambam, Dr. Irena Papier, RN, DMS, PI-Technion, Assist. Prof. Dvir Aran, PhD,

Research Team
Dr. Shay Perek, MD; Dr. Yona Vaisbuch, MD; Gila Hyams, RN, MA; Dr. Haggai Bar-Yoseph, MD, Rambam

Artificial Intelligence offers a powerful framework for advancing nutritional support during hospitalization by enabling continuous, data driven clinical insight. Based on AI tools, hospital systems can integrate oral intake records and other patient performance parameters recorded in nursing documentation, as well as laboratory values, and procedures that are often fragmented and not synthesized in real time. By transforming these data into actionable clinical signals, AI supports early identification of nutritional risk and timely clinical decision making. Nutritional risk is typically assessed routinely at hospital admission, while ongoing reassessment during the hospital stay is often limited or absent.

Nutrition

Automated analysis of nursing reports and intake monitoring allows earlier recognition of insufficient nutritional intake and reduces delays in initiating nutritional support. Predictive models can stratify patient risk, anticipate complications related to undernutrition, and support individualized nutritional strategies based on a patient-specific clinical profile.

For medical and nursing research, AI enables systematic exploration of associations between delayed or absent nutritional support and clinical outcomes. This approach supports both hypothesis generation and real-time clinical decision support, strengthening evidence-based practice and improving patient safety and outcomes during hospitalization.

Objectives and clinical need

Undernutrition affects approximately 30–40% of hospitalized patients and is frequently underrecognized, particularly when nutritional deterioration occurs after admission and worsening during hospitalization. While nutritional risk is routinely screened on admission, ongoing monitoring of food intake during hospitalization is limited and fragmented. Based on the available data, there is a clear clinical need for automated systems that continuously identify patients with declining nutritional intake and support timely, multidisciplinary nutritional interventions within routine hospital workflows.

Database

The current study is based on a single-center dataset from the electronic medical records of Rambam Health Care Campus. The dataset includes structured and unstructured nursing documentation of meal intake using semi-quantitative food intake measures, demographic and clinical variables, laboratory data, hospitalization characteristics, and free-text from clinical notes extracted from the EMR. All adult patients hospitalized for 5 or more days across various departments are included. The study is designed in two phases: Phase 1: Development of a computational tool to retrieve, integrate, and summarize nutritional intake data throughout hospitalization. Natural language processing techniques are applied to nursing free-text documentation to enrich and standardize food intake features, enabling clear visualization of intake trends for clinical use. Phase 2: Construction and validation of machine learning time series forecasting models to predict low nutritional intake several days in advance. Models are trained using demographic, clinical, nutritional, and administrative variables to identify patients at risk and generate early alerts for intervention.

Research results

The project will deliver an AI-based decision support system that transforms routinely collected nursing documentation into actionable clinical insights, enabling continuous nutritional screening, detection of declining intake, and bridging the gap between documentation and effective nutritional care during hospitalization.

This study is based on previous publications

  • Papier, I., Chermesh, I., Mashiach, T., Gruenwald, I., & Banasiewicz, T. (2025). Prevalence of the use of oral nutritional supplements among acute inpatients at risk of malnutrition and associated patient characteristics. Journal of Clinical Nursing, 34(3), 849–859. https://doi.org/10.1111/jocn.17076
  • Papier, I., Chermesh, I., Mashiach, T., & Banasiewicz, T. (2023). Evaluation of prevalence of food intake monitoring during acute hospitalization and its association with malnutrition screening scores of inpatients who were not considered for enteral or parenteral nutrition. Nutrition, 110, 112031. https://doi.org/10.1016/j.nut.2023.112031
  • Papier, I., Sagi-Dain, L., Chermesh, I., Mashiach, T., & Banasiewicz, T. (2022). Absence of oral nutritional support in low food intake inpatients is associated with an increased risk of hospital-acquired pressure injury. Clinical Nutrition ESPEN, 51, 190–198. https://doi.org/10.1016/j.clnesp.2022.09.003