AI boosts accuracy in critical care nutrition monitoring

Current guidelines recommend initiating enteral nutrition (EN) or parenteral nutrition (PN) within 48 hours of admission. Yet, implementation remains inconsistent, and standard calculations for energy expenditure (EE) using predictive equations can be misleading, especially in patients with obesity, amputation, or hyperinflammatory states.

ICU mortality rates remain alarmingly high worldwide and healthcare systems stretch under the weight of complex, critically ill patients. Modern critical care faces dual burdens: rising costs and highly individualized patient needs while outdated, one-size-fits-all nutritional models continue to underperform.

A comprehensive narrative review titled “Personalized Nutrition Strategies for Patients in the Intensive Care Unit: A Narrative Review on the Future of Critical Care Nutrition” published in Nutrients (2025), spearheaded by Stoian et al., outlines an urgent call for a paradigm shift toward personalized, data-driven nutritional therapy. The study evaluates emerging tools such as indirect calorimetry, AI-driven nutrient monitoring, and genomics-integrated protocols, laying the groundwork for a new standard in ICU nutrition.

What undermines traditional ICU nutrition models?

The review identifies a critical weakness in conventional ICU nutrition practices: the failure to account for patient-specific variables and the dynamic progression of critical illness. Current guidelines recommend initiating enteral nutrition (EN) or parenteral nutrition (PN) within 48 hours of admission. Yet, implementation remains inconsistent, and standard calculations for energy expenditure (EE) using predictive equations can be misleading, especially in patients with obesity, amputation, or hyperinflammatory states.

Indirect calorimetry (IC), despite being the gold standard for estimating resting energy expenditure (REE), remains underutilized due to equipment costs and skill shortages. The reliance on static formulas rather than dynamic monitoring means that nutritional strategies may lead to overfeeding or underfeeding—both of which are associated with poor clinical outcomes. Critically ill patients frequently exhibit ICU-acquired weakness (ICUAW), characterized by muscle atrophy, prolonged recovery, and increased mortality risk. Studies cited in the review confirm that insufficient protein and energy intake correlates with increased 28-day mortality, even when calorie targets are partially met.

Additionally, hospital food quality, gastrointestinal dysfunction, and medication effects (e.g., sedatives or corticosteroids) compound malnutrition risk. Subgroups such as elderly or neurological patients face higher vulnerability, often exacerbated by swallowing disorders, edentulism, or poor mastication capabilities. The inadequacy of standard hospital diets and late nutritional screening amplify these risks.

How do emerging tools like AI, metabolomics, and genomics redefine ICU nutrition?

The study highlights groundbreaking technological advancements that enable a truly personalized approach to nutritional therapy in ICUs. AI-integrated platforms now allow for real-time tracking of patient caloric intake, metabolic markers, and nutrient absorption. These tools help minimize human error, ensure continuous monitoring, and integrate multiple variables including propofol infusions, stress-induced metabolism, and non-nutritional calorie sources.

Metabolomics, the analysis of small molecule metabolites in blood or urine, has been pivotal in identifying early markers of nutritional deficiencies. Combined with transcriptomics and proteomics, these biomarkers offer a molecular-level view of the patient’s metabolic status, allowing for tailored interventions. Studies demonstrate that disturbances in fatty acid, tryptophan, and lipid metabolism are directly linked to disease progression and survival odds in sepsis and critical illness.

Genomics and nutrigenomics offer additional precision by considering genetic predispositions to nutrient absorption, metabolic pathways, and disease susceptibility. For example, the review notes how variants in the LCT gene explain lactose tolerance in pastoral populations or how the AMY1 gene copy number correlates with starch digestion efficiency. Such insights enable clinicians to select nutrient compositions best suited to each patient’s genetic and microbiotic profile.

Despite these advances, the review cautions that many ICUs, especially in low- and middle-income countries, lack the financial infrastructure to deploy these sophisticated tools. Even in high-income settings, institutional protocols are often missing or inconsistently applied, leaving nutrition decisions solely to attending physicians without interdisciplinary collaboration.

What policies and clinical shifts are needed to operationalize personalized nutrition in ICUs?

The authors argue for an urgent reformation of ICU nutrition management, centered on multidisciplinary coordination and technological investment. Personalized nutrition cannot succeed without cohesive collaboration between ICU physicians, nutritionists, pharmacists, radiologists, IT specialists, and policy stakeholders. The review cites models like Australia’s SIMPLE framework as effective examples of systematized, interdisciplinary approaches that improve patient outcomes and reduce ICU stays.

The study proposes a structured implementation strategy: early screening within 24–48 hours, regular reassessments, investment in IC and metabolomic technologies, and legislative frameworks to integrate dietitians into ICU teams. In many hospitals, nutritionists still only contribute generic menus based on disease type (e.g., diabetes, ulcers), rarely engaging in patient-specific planning.

Furthermore, artificial intelligence can facilitate the creation of predictive models that adapt to changes in vital signs, organ function, and inflammatory markers. Such real-time algorithms not only optimize feeding schedules but also flag high-risk patients, allowing for timely nutritional interventions that could influence survival rates and hospital costs.

Barriers remain steep, costs of omics tools, lack of trained personnel, and inconsistent guideline adoption, but the authors contend that these investments will ultimately be cost-effective. Personalized nutrition not only reduces ICU-acquired complications and hospital duration, but also promotes functional recovery, minimizing the long-term burden on healthcare systems and families.