explICU: A web-based visualization and predictive modeling toolkit for mortality in intensive care patients

Abstract

Preventing mortality in intensive care units (ICUs) has been a top priority in American hospitals. Predictive modeling has been shown to be effective in prediction of mortality based upon data from patients' past medical histories from electronic health records (EHRs). Furthermore, visualization of timeline events is imperative in the ICU setting in order to quickly identify trends in patient histories that may lead to mortality. With the increasing adoption of EHRs, a wealth of medical data is becoming increasingly available for secondary uses such as data exploration and predictive modeling. While data exploration and predictive modeling are useful for finding risk factors in ICU patients, the process is time consuming and requires a high level of computer programming ability. We propose explICU, a web service that hosts EHR data, displays timelines of patient events based upon user-specified preferences, performs predictive modeling in the back end, and displays results to the user via intuitive, interactive visualizations.

Machine learning for ovarian cancer: lasso regression-based predictive model of early mortality in patients with stage I and stage II ovarian cancer

“…While there does not exist an intuitive method for risk stratification of patients with ovarian cancer, there is strong evidence of the potential of machine learning for stratification of patients in other diseases such as heart failure (Rasmy et al 2018; Choi et al 2017; Chen et al 2019; Ng et al 2017), kidney disease (Makino et al 2019), and critical care(Chen, Su, et al 2015; Chen, Kumar, et al 2015; Katuwal and Chen 2016; Kamat et al 2018; Yu, Liu, and Nemati 2019; Johnson et al 2016). Furthermore, machine learning has been shown to be effective for readmission prediction (Rajkomar et al 2018; Desautels et al 2017; Chen, Su, et al 2015), drug adverse event prediction (Cheng and Zhao 2014).…”

A Visualization and Benchmarking Simulator for Clinical Data

“…The application has a risk prediction function for the in-hospital mortality of patients where various features such as respiratory rate, oxygen saturation and recorded events are analyzed. For previous versions of the MIMIC database, there are other tools that, in addition to visualizing patient data, include predictive models for patient mortality [11].…”

Visualizing Patient Timelines in the Intensive Care Unit

“…Chen et al present a web-based visualization tool for MIMIC-II. 12 Their visualization incorporates both a patient timeline module as well as a predictive modeling module. Like our work, the authors demonstrate how a visualization of a predictive model for mortality can be used for additional insight.…”