Hemodialysis (HD) is a necessary treatment for end-stage kidney disease (ESKD) patients in order to prevent cardiovascular morbidity and mortality that may be related to the hemodynamic effects of rapid ultrafiltration. Despite significant advances in HD technology, only half of ESKD patients treated with HD survive more than 3 years. Fluid management remains one of the most challenging aspects of HD care, with serious implications for morbidity and mortality.
In this paper, we develop a novel algorithm to design real time optimal, robust ultrafiltration rates based on actual HD data to identifying the parameters of a fluid volume model of an individual patient during HD. Our design achieves, if exists, an optimal ultrafiltration profile for the identified nominal model under maximum ultrafiltration and hematocrit constraints and guarantees that these constraints are satisfied over a pre-defined set of parameter uncertainty. We demonstrate the robust performance of our algorithm through a combination of clinical data and simulations.
