PD011 - REAL-TIME AI-DRIVEN RISK STRATIFICATION TO OPTIMIZE PARENTERAL NUTRITION SAFETY: A CLINICAL IMPLEMENTATION STUDY

PD011

REAL-TIME AI-DRIVEN RISK STRATIFICATION TO OPTIMIZE PARENTERAL NUTRITION SAFETY: A CLINICAL IMPLEMENTATION STUDY

C. Y. Wang^1,*, C.-Y. Chen², Y.-L. Shih², L.-C. Sun², H.-L. Tsai¹

¹Surgery, ²Nursing, Kaohsiung Medical University Hospital, Kaohsiung, Taiwan, Province of China

Rationale: Can a real-time AI-driven risk stratification system improve early detection of clinically significant abnormalities and optimize decision-making in parenteral nutrition care?

Methods: A rule-based AI scoring system incorporating clinical expertise was embedded into the NTT workflow. Laboratory abnormalities were dynamically classified into three risk tiers: red, orange, and yellow, with a cumulative score guiding clinical prioritization. The system continuously screened PN patients and generated real-time alerts. Implementation outcomes were prospectively evaluated over a 16-day working period, focusing on intervention rate, workflow efficiency, and clinical relevance.

Results: The AI system identified 78 clinically relevant alert events. Of these, 38.4% resulted in immediate modification of PN prescriptions, demonstrating direct clinical impact. Among the remaining cases, 60.3% were appropriately managed with close monitoring, while only 1.3% were deemed unrelated to PN, indicating high specificity. The system enabled early detection of electrolyte disturbances and hepatic/renal dysfunction, significantly reducing review time and accelerating intervention. Importantly, cumulative risk scoring effectively identified patients with multiple concurrent abnormalities, facilitating early recognition of critical conditions such as refeeding syndrome and multi-organ dysfunction. This approach reduced unnecessary alerts and improved resource allocation toward high-risk patients.

Conclusion: This real-time AI-driven graded alert system represents a scalable and clinically impactful innovation in nutrition support. By transforming fragmented data into actionable risk stratification, it enhances decision-making efficiency, reduces clinician burden, and improves patient safety. 

Disclosure of Interest: None declared