Abstract: SA-PO0089
Precision Plates: Artificial Intelligence-Enhanced Nutrition for Patients with ESKD
Session Information
- Artificial Intelligence: Kidney Imaging, Pathology, and Precision Phenotyping
October 24, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Artificial Intelligence, Digital Health, and Data Science
- 300 Artificial Intelligence, Digital Health, and Data Science
Authors
- Reagan, Mckenzie, University of Nebraska Medical Center College of Medicine, Omaha, Nebraska, United States
- Finch, Alexis, University of Nebraska Medical Center College of Medicine, Omaha, Nebraska, United States
- Sauvage, Elizabeth, University of Nebraska Medical Center, Omaha, Nebraska, United States
- Coe, Susan, Fresenius Kidney Care, Omaha, Nebraska, United States
- Hawkins, Jay L., University of Nebraska Medical Center, Omaha, Nebraska, United States
- Ravipati, Prasanth, University of Nebraska Medical Center, Omaha, Nebraska, United States
Background
Many hemodialysis (HD) patients fail to achieve target phosphorus due to dietary challenges. In addition to traditional dietary education, structured meal planning has been shown to improve adherence to dietary recommendations. This quality improvement project utilized artificial intelligence (AI) to help develop patient-specific meal plans that consider socioeconomic context and preferences with the goal of improving phosphorous control.
Methods
The project was conducted at an in-center HD unit with 28 patients over a 6-month period. Using the AI platform OpenEvidenceR, preliminary meal plans consisting of 3 meals and 2 snacks per day were created to three weekly budgets ($60, $100, and $140) to reflect differences in food accessibility. Plans were reviewed, edited, and finalized by a registered dietitian. A standardized prompt was used: "Create a one-week meal plan including three meals and two snacks a day for a patient on hemodialysis with hyperphosphatemia with a budget of X." All 3 plans were provided and patients were encouraged to select the plan best suited to their needs. At 3 months, feedback was collected from participants to generate further personalized meal plans (e.g., vegetarian). Phosphorous, albumin, and fluid gains were trended per dialysis unit protocol.
Results
Among the 28 patients, 15 had persistent hyperphosphatemia (>6 mg/dL). Over the observed period, 5 (33%) demonstrated consistent improvement in serum phosphorus (3 consecutive down-trending measurements or phosphorus less than 6 mg/dL). Including all patients, there was no meaningful change in volume management or albumin parameters at the dialysis unit level. At initiation, all patients were appreciative and interested in the meal plans, but that did not translate to consistent use. Sixteen patients (57.1%), noted "intermittent" or "no use" of the plans at the 3-month check-in. In terms of feedback, patients valued the greater variety of cuisine, protein options, personalization, and reduced decision fatigue. Concerns were raised regarding potassium content of certain foods and gastrointestinal intolerance with certain items.
Conclusion
In this small cohort, AI-enhanced nutrition improved phosphorus control in a meaningful number of patients. In addition, the project was well received by patients. Given the positive feedback and success of several patients, AI powered meal planning may be an effective strategy in select individuals.