Guiding Efficient, Effective, and Patient-Oriented Electrolyte Replacement in Critical Care: An Artificial Intelligence Reinforcement Learning Approach

Abstract

Both provider- and protocol-driven electrolyte replacement have been linked to the over-prescription of ubiquitous electrolytes. Here, we describe the development and retrospective validation of a data-driven clinical decision support tool that uses reinforcement learning (RL) algorithms to recommend patient-tailored electrolyte replacement policies for ICU patients. We used electronic health records (EHR) data that originated from two institutions (UPHS; MIMIC-IV). The tool uses a set of patient characteristics, such as their physiological and pharmacological state, a pre-defined set of possible repletion actions, and a set of clinical goals to present clinicians with a recommendation for the route and dose of an electrolyte. RL-driven electrolyte repletion substantially reduces the frequency of magnesium and potassium replacements (up to 60%), adjusts the timing of interventions in all three electrolytes considered (potassium, magnesium, and phosphate), and shifts them towards orally administered repletion over intravenous replacement. This shift in recommended treatment limits risk of the potentially harmful effects of over-repletion and implies monetary savings. Overall, the RL-driven electrolyte repletion recommendations reduce excess electrolyte replacements and improve the safety, precision, efficacy, and cost of each electrolyte repletion event, while showing robust performance across patient cohorts and hospital systems.

Man Versus Machine: Provider Directed Versus Precision Automated Critical Care Management in a Porcine Model of Distributive Shock

“…Specific applications include early detection of acute kidney injury, pulmonary embolism, gene expression in sepsis, and leukocyte phenotyping when trying to understand the pathophysiology of sepsis (45)(46)(47)(48)(49). Reinforcement learning has also been used to formulate ICU electrolyte replacement protocols and determine treatment decisions in sepsis (50,51).…”

Development and validation of a reinforcement learning model for ventilation control during emergence from general anesthesia

“…To the best of our knowledge, this is the first study to develop and validate the reinforcement learning model to suggest the optimal timing of controlling ventilation during anesthesia emergence in surgical patients. Previous studies have developed offline reinforcement learning models to solve complicated medical problems 11 , 14 – 17 . One study developed a reinforcement learning model to recommend various interventions, such as administering intravenous fluid and medications, to treat patients with sepsis in the intensive care unit (ICU) 14 .…”

Artificial intelligence in critical illness and its impact on patient care: a comprehensive review

“…Furthermore, recent advancements in image analysis using convolutional neural networks have enabled the evaluation of traumatic brain injury with more accuracy than manual methods when viewed on head computed tomography scans ( 6 ). In a retrospective analysis by Prasad et al ( 7 ), a reinforcement learning (RL) approach was used to develop a treatment protocol for electrolyte replacements in an ICU setting. This system provides recommendations for patient care that can be continuously updated based on the patient’s specific needs.…”

“…Optimal RL policy is reported to be able to recommend electrolyte replacements in a more targeted manner, potentially reducing the number of repletion events and the cost and time associated with unnecessary or repeat orders. Additionally, the system uses a reward and punishment system, reducing the costs and risks associated with intravenous delivery ( 7 ). This is not to underscore the value and significance of care-givers in the critical care setting; instead, it is a remarkable example of how new technologies such as AI can have a significant impact on the care of critically ill patients.…”