An interpretable RL framework for pre-deployment modeling in ICU hypotension management.

Computational methods from reinforcement learning have shown promise in inferring treatment strategies for hypotension management and other clinical decision-making challenges. Unfortunately, the resulting models are often difficult for clinicians to interpret, making clinical inspection and validation of these computationally derived strategies challenging in advance of deployment. In this work, we develop a general framework for identifying succinct sets of clinical contexts in which clinicians make very different treatment choices, tracing the effects of those choices, and inferring a set of recommendations for those specific contexts. By focusing on these few key decision points, our framework produces succinct, interpretable treatment strategies that can each be easily visualized and verified by clinical experts. This interrogation process allows clinicians to leverage the model's use of historical data in tandem with their own expertise to determine which recommendations are worth investigating further e.g. at the bedside. We demonstrate the value of this approach via application to hypotension management in the ICU, an area with critical implications for patient outcomes that lacks data-driven individualized treatment strategies; that said, our framework has broad implications on how to use computational methods to assist with decision-making challenges on a wide range of clinical domains.

A Method to Explore Variations of Ventilator-Associated Event Surveillance Definitions in Large Critical Care Databases in the United States.

The Centers for Disease Control has well-established surveillance programs to monitor preventable conditions in patients supported by mechanical ventilation (MV). The aim of the study was to develop a data-driven methodology to examine variations in the first tier of the ventilator-associated event surveillance definition, described as a ventilator-associated condition (VAC). Further, an interactive tool was designed to illustrate the effect of changes to the VAC surveillance definition, by applying different ventilator settings, time-intervals, demographics, and selected clinical criteria. DESIGN: Retrospective, multicenter, cross-sectional analysis. SETTING: Three hundred forty critical care units across 209 hospitals, comprising 261,910 patients in both the electronic Intensive Care Unit Clinical Research Database and Medical Information Mart for Intensive Care III databases. PATIENTS: A total of 14,517 patients undergoing MV for 4 or more days. MEASUREMENTS AND MAIN RESULTS: We designed a statistical analysis framework, complemented by a custom interactive data visualization tool to depict how changes to the VAC surveillance definition alter its prognostic performance, comparing patients with and without VAC. This methodology and tool enable comparison of three clinical outcomes (hospital mortality, hospital length-of-stay, and ICU length-of-stay) and provide the option to stratify patients by six criteria in two categories: patient population (dataset and ICU type) and clinical features (minimum Fio2, minimum positive end-expiratory pressure, early/late VAC, and worst first-day respiratory Sequential Organ Failure Assessment score). Patient population outcomes were depicted by heatmaps with mortality odds ratios. In parallel, outcomes from ventilation setting variations and clinical features were depicted with Kaplan-Meier survival curves. CONCLUSIONS: We developed a method to examine VAC using information extracted from large electronic health record databases. Building upon this framework, we developed an interactive tool to visualize and quantify the implications of variations in the VAC surveillance definition in different populations, across time and critical care settings. Data for patients with and without VAC was used to illustrate the effect of the application of this method and visualization tool.

Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care.

The aim of this work was to develop and evaluate the reinforcement learning algorithm VentAI, which is able to suggest a dynamically optimized mechanical ventilation regime for critically-ill patients. We built, validated and tested its performance on 11,943 events of volume-controlled mechanical ventilation derived from 61,532 distinct ICU admissions and tested it on an independent, secondary dataset (200,859 ICU stays; 25,086 mechanical ventilation events). A patient "data fingerprint" of 44 features was extracted as multidimensional time series in 4-hour time steps. We used a Markov decision process, including a reward system and a Q-learning approach, to find the optimized settings for positive end-expiratory pressure (PEEP), fraction of inspired oxygen (FiO2) and ideal body weight-adjusted tidal volume (Vt). The observed outcome was in-hospital or 90-day mortality. VentAI reached a significantly increased estimated performance return of 83.3 (primary dataset) and 84.1 (secondary dataset) compared to physicians' standard clinical care (51.1). The number of recommended action changes per mechanically ventilated patient constantly exceeded those of the clinicians. VentAI chose 202.9% more frequently ventilation regimes with lower Vt (5-7.5 mL/kg), but 50.8% less for regimes with higher Vt (7.5-10 mL/kg). VentAI recommended 29.3% more frequently PEEP levels of 5-7 cm H2O and 53.6% more frequently PEEP levels of 7-9 cmH2O. VentAI avoided high (>55%) FiO2 values (59.8% decrease), while preferring the range of 50-55% (140.3% increase). In conclusion, VentAI provides reproducible high performance by dynamically choosing an optimized, individualized ventilation strategy and thus might be of benefit for critically ill patients.

Intensive Care Unit Telemedicine in the Era of Big Data, Artificial Intelligence, and Computer Clinical Decision Support Systems.

This article examines the history of the telemedicine intensive care unit (tele-ICU), the current state of clinical decision support systems (CDSS) in the tele-ICU, applications of machine learning (ML) algorithms to critical care, and opportunities to integrate ML with tele-ICU CDSS. The enormous quantities of data generated by tele-ICU systems is a major driver in the development of the large, comprehensive, heterogeneous, and granular data sets necessary to develop generalizable ML CDSS algorithms, and deidentification of these data sets expands opportunities for ML CDSS research.