Predicting fluid-response, the heart of hemodynamic management: A model-based solution

Area of Science:

• Cardiovascular Physiology
• Critical Care Medicine
• Biomedical Engineering

Background:

• Intravenous fluid infusions are critical for treating circulatory shock.
• Predicting individual patient stroke volume (SV) response to fluids is challenging.
• Accurate SV prediction could optimize fluid therapy and patient outcomes.

Purpose of the Study:

• To develop and validate a model-based method for predicting SV changes in response to fluid infusions.
• To assess the model's ability to guide fluid therapy decisions in a preclinical setting.
• To evaluate different parameter identification strategies for model accuracy.

Main Methods:

• A model-based method was applied to predict SV changes in pigs (N=6) receiving fluid infusions.
• Model parameters were identified using data from the entire infusion, first 200 mL, or first 100 mL.
• Root-mean-square error (RMSE), polar plot analysis, and ROC analysis were used for validation.

Main Results:

• The model demonstrated good agreement with measured SV trends, particularly when parameters were identified using more data (SV_fl^Fit, SV_fl²⁰⁰).
• Model predictions (SV_fl²⁰⁰, SV_fl¹⁰⁰) showed moderate ability to distinguish responsive from non-responsive fluid interventions (ROC AUC 0.64-0.69).
• Parameter identification from the first 200 mL (SV_fl²⁰⁰) provided a balance between accuracy and data requirements.

Conclusions:

• The model-based method shows promise for predicting SV dynamics and guiding fluid therapy.
• Sufficient data (e.g., first 200 mL of infusion) is crucial for accurate parameter identification and prediction.
• Further research is needed to confirm clinical applicability and explore integration into closed-loop systems.