HEMODYNAMIC OPTIMIZATION THROUGH BIOMECHANICAL MODELING IN POST-CARDIAC SURGERY PATIENTS

Abstract

The causes of hemodynamic instability after cardiac surgery are multifactorial and change over time, which makes the bedside optimization of the fluid, vasoactive and ventilator choice more difficult.  Biomimical cardiovascular models can be patient-specific and can be used to support the decision-making process by predicting physiological reactions to standard treatment.  An ICU cohort quantitative, model-directed workflow was built on the basis of invasive hemodynamic and perfusion measures at a given time interval.  Each patient was calibrated using a closed loop 0D cardiovascular model, which was assessed on RMSE and agreement.  We compared the predictability of the effects of standardized clinical perturbations on mean arterial pressure ( ΔMAP ) and cardiac index ( ΔCI ). We also examined such secondary outcomes as lactate clearance, vasopressor free, ICU duration of stay and acute kidney injury (AKI).  The admission MAP in the analytic cohort (n=168) of this study was 69 (IQR 6074) mmHg, cardiac index 2.09 (1.70 2.38) L/min/m 2, and lactate level 1.9 (1.429) mmol/L, and 96.4% of individuals in the first 6 hours were treated with vasopressor.  Calibration of the model produced a MAP RMSE of 5.0 (4.0-6.0), and a CI RMSE of 0.22 (0.16-0.30) L/min/m 2.  Correlation in the case of ΔMAP was r=0.81 and in ΔCI, the correlation was r=0.77. The average absolute error (MAE) was 2.6 mmHg ( 0 -MAP) and 0.12 L/min/m 2 ( 0 -CI). 85.7% of the 0 -MAP predictions fell within ±5 mmHg and 89.3% of the 0 -CI predictions fell within ±0.25 L/min/m 2.  The MAP increased to 79 (70–90) mmHg, cardiac index increased to 2.32 (1.872.71) L/min/m 2, the level of lactate was reduced to 1.2 (0.81.6) mmol/L and 72.6% of individuals had a lactate clearance rate of 30 or more. The AKI rate was 10.1%.  Adjusted analyses were performed to show that admission lactate (OR 1.28 per +1 mmol/L) and low urine output (<0.7 mL/kg/h; OR 2.05) are associated with AKI. Conversely, adhering to the best advice provided by the model was associated with reduced risks of chronic instability (OR 0.62).  A patient-specific closed-loop biomechanical modeling pipeline demonstrated good calibration and clinically meaningful prediction of short-term hemodynamic response reflecting its feasibility as a quantitative decision-support system to be used to optimize post-cardiac surgery.  Further hypothetical validation with real-life intervention monitoring and multiscale physiological description (e.g. multiscale vascular and valve mechanics) is required.