Implementation and learning curve in AI-assisted fluid management during abdominal oncologic surgery: a retrospective observational study

Comprehensive Summary

The study evaluates the implementation and learning curve associated with an artificial intelligence-based decision support system, Assisted Fluid Management (AFM), in guiding intraoperative fluid therapy during major abdominal oncologic surgery. Through a retrospective observational design, researchers analyzed data from 59 patients undergoing abdominal oncological surgery at a high-volume tertiary center. Of these 59 surgeries, 67% were open surgeries with 88% of patients being categorized as ASA III physical status. Overall, there were 404 fluid challenges recorded within this data set. The study compared two consecutive time periods following AFM adoption, examining changes in surgeon behavior, including the proportion of clinician-initiated versus AI-suggested fluid boluses, as well as the physiological effectiveness of these interventions which was determined by stroke volume response. Results demonstrated a significant shift over time toward increased reliance on AFM-suggested fluid challenges and a corresponding decrease in clinician-initiated interventions. This transition was associated with improved effectiveness of fluid administration, with greater stroke volume responses observed, particularly for AI-guided boluses. On average, the AFM AI-guided fluid challenge exhibited a higher percentage of positive responses as compared to manual clinician-led fluid challenges, where group B (July 2025 - March 2026 cohort) yielded a 20% greater positive response, where a positive response was a 6.8% or greater stroke volume increase. Overall, the findings suggest that integration of AI-assisted decision support systems can influence clinician behavior and improve the precision of intraoperative fluid management. On adoption, there was a measurable learning curve as adaptation towards AFM increased among the subsequent cohort.

Outcomes and Implications

This study highlights the potential of artificial intelligence-based decision support systems to standardize and enhance intraoperative fluid management in complex surgical settings. As seen in the second cohort, adoption of AI-guided fluid management was markedly increased and preferred over manual fluid challenges. This suggests that AI has the potential to streamline intraoperative surgical management with great acuity. In fact, AI may improve outcomes in conjunction with physician-led decision making as AFM saw a greater positive stroke volume response likelihood. The findings suggest that tools like AFM may reduce variability in practice and promote more consistent and data-driven interventions, which are particularly valuable in high-stakes environments such as oncologic surgery where fluid balance is critical. For clinicians, this emphasizes the importance of familiarity and continued interaction with AI systems to fully realize their benefits, while for healthcare systems, it highlights the need for structured implementation and training strategies to support adoption. However, the study also points to important limitations, including its retrospective design, small sample size, and lack of direct assessment of postoperative clinical outcomes, leaving uncertainty about the broader impact on patient morbidity and mortality. Nonetheless, AI-enhanced clinical decisionmaking has great potential in improving patient outcomes by leveraging large datasets and learning models to make high-stakes clinical judgements.