Acute kidney injury (AKI) affects up to half of critically ill patients admitted to intensive care units (ICU). In patients with AKI and hemodynamic instability, continuous renal replacement therapy (CRRT) is the preferred dialysis modality for solute and volume control. ICU mortality in this vulnerable population is high (~75%) but kidney recovery occurs in up to two-thirds of survivors. Fluid overload is a potentially modifiable risk factor associated with these outcomes. However, there are currently no universally accepted approaches for predicting kidney recovery, survival or individual response to fluid removal during CRRT. Due to recent advances in computer science and widespread big data usage, deep learning (DL) has emerged as a valuable approach. DL allows construction of risk prediction models using time-series data that incorporate thousands of variables and dynamic changes in these variables derived from multi-dimensional sources and not only static values of these variables. We propose to develop and validate innovative DL approaches to dynamically predict these outcomes using multi-modal data from electronic health records and CRRT machines. We demonstrated superiority of DL models without a-priori variable selection compared to optimized logistic regression (C-Statistic of 0.72 vs. 0.62) for prediction of RRT liberation. We also showed that mortality prediction improved by incorporating changes in clinical data within 6-hour intervals after CRRT initiation. In addition, we identified distinctive mortality risk according to quintiles of achieved net ultrafiltration rates, after adjustment by patient?s weight, duration of CRRT, and other clinical parameters: OR 8.0 (95% CI: 2.7-25.1) when the highest quintile (>36 ml/kg/day) was compared to the lowest quintile (<13 ml/kg/day). We hypothesize that innovative DL approaches integrating time- series data will generate accurate and generalizable risk prediction models that can impact CRRT delivery. We will utilize a multi-institutional dataset that encompasses clinical data and CRRT programmatic and therapy data (CRRTnet registry, n=1500 patients) for model development and an independent multi-institutional dataset for validation (n=1500 patients) to: 1) continuously predict short-term (7-day) and medium-term (28-day) liberation from RRT due to kidney recovery; 2) continuously predict 24-hour mortality; and 3) identify and validate sub- phenotypes of patients with AKI on CRRT with differing achieved net ultrafiltration rates. This innovative research will assist 1) the development of novel clinical decision support platforms for guiding informed CRRT delivery and promoting kidney recovery; 2) the identification of sub-phenotypes of patients that can benefit from precision- medicine approaches to fluid removal during CRRT; and 3) the design of interventional studies focusing on fluid removal during CRRT to impact patient-centered outcomes.
Continuous renal replacement therapy (CRRT) is the preferred dialysis treatment for critically ill patients with acute kidney injury and hemodynamic instability, yet mortality is high (~75%). Presently, there are no universally accepted approaches for predicting kidney recovery, survival or individual response to fluid removal during CRRT. We propose to develop and validate innovative deep learning approaches to dynamically predict these outcomes, which could guide CRRT decision-making including intensification, de-escalation, and enrich clinical trials focusing on fluid removal during CRRT.
