Left ventricular assist devices (LVADs), such as the HeartMate 3 (HM3), are a treatment option for advanced heart failure (HF). Non-invasive monitoring of heart-pump interaction is crucial but relies on estimated parameters only. Despite the widespread clinical utilization of HM3 estimated flow rate (Qest) and pulsatility index (PI) their accuracy was not yet quantified. This study assessed the HM3 monitoring by describing underlying physical characteristics, estimation algorithms and deriving accuracy of estimates in clinical practice. Identification measurements were performed to quantify the physical characteristics and mechanisms of flow estimation. Virtual patients coupled to a hybrid hock circulatory loop were employed to evaluate estimator accuracies in realistic operating conditions. The correlation between Qest and torque-generating-current was strong (r² > 0.99, p < 0.001), confirming that estimation is based on speed, current, and viscosity. Also, the PI is directly derived from current. However, the underlying current/flow relationship is ambiguous (one current value corresponds to two flow rates). The HM3 flow estimator showed an RMSE of 1.63 L/min and a correlation coefficient of r = 0.86 (p < 0.001). The non-linear relationship obscures the systolic portion of flow as current declines during peak systole at higher flow rates. HM3 monitoring is unable to accurately represent the actual hemodynamics, particularly in the transition between full support and partial support, which is a region highly prevalent and crucial in clinical practice. It is therefore recommended to interpret values and trends of the HM3 monitoring system with caution.
© 2025. The Author(s).
