Predicting acute kidney injury at hospital re-entry using high-dimensional electronic health record data

Abstract

Acute Kidney Injury (AKI), a sudden decline in kidney function, is associated with increased mortality, morbidity, length of stay, and hospital cost. Since AKI is sometimes preventable, there is great interest in prediction. Most existing studies consider all patients and therefore restrict to features available in the first hours of hospitalization. Here, the focus is instead on rehospitalized patients, a cohort in which rich longitudinal features from prior hospitalizations can be analyzed. Our objective is to provide a risk score directly at hospital re-entry. Gradient boosting, penalized logistic regression (with and without stability selection), and a recurrent neural network are trained on two years of adult inpatient EHR data (3,387 attributes for 34,505 patients who generated 90,013 training samples with 5,618 cases and 84,395 controls). Predictions are internally evaluated with 50 iterations of 5-fold grouped cross-validation with special emphasis on calibration, an analysis of which is performed at the patient as well as hospitalization level. Error is assessed with respect to diagnosis, race, age, gender, AKI identification method, and hospital utilization. In an additional experiment, the regularization penalty is severely increased to induce parsimony and interpretability. Predictors identified for rehospitalized patients are also reported with a special analysis of medications that might be modifiable risk factors. Insights from this study might be used to construct a predictive tool for AKI in rehospitalized patients. An accurate estimate of AKI risk at hospital entry might serve as a prior for an admitting provider or another predictive algorithm.

Developing Deep Learning Continuous Risk Models for Early Adverse Event Prediction in Electronic Health Records: an AKI Case Study

“…Given that we have demonstrated the effectiveness of the proposed protocol in the context of predicting the risk of future AKI, here we compare our protocol to those that were previously used to develop AKI risk models. Most prior work on AKI risk modelling with ML involved protocols that were limited to only produce models that provide predictions at a single point in time [26][27][28] , instead of a personalised, continuously updating risk score. Previously developed continuous machine learning models of AKI risk have either not demonstrated a clinically applicable level of predictive performance 29 or have focused on predictions across too short a time horizon, leaving little time for clinical assessment and intervention 30 .…”

“Beyond MELD” – Emerging strategies and technologies for improving mortality prediction, organ allocation and outcomes in liver transplantation

“…For instance, incorporation of time-variant variables, such as laboratory values and vital signs, captured in EHRs have enabled continuous prediction of the development of acute kidney injury during inpatient admissions. 70,71 Moreover, the use of longitudinal and sequential data elements gathered from EHR flowsheets, medication administrations, physician notes, and radiology reports have enabled the construction of deeplearning models to more accurately predict inhospital mortality, 30-day readmissions, and prolonged length of stay. 72 In clinical hepatology, the integration of longitudinal EHR elements, such as structured flowsheet entries, medication administration, procedure orders, vital signs, and laboratory values, has enabled dynamic calculations of the North American Consortium for the Study of End-Stage Liver Disease-ACLF and Chronic Liver Failure Consortium-ACLF prognostication scores in hospitalised patients with ACLF.…”

An introduction to machine learning for clinicians: How can machine learning augment knowledge in geriatric oncology?

“…This is easily accomplished and visualized in two or three dimensions (Fig. 1), but given a high number of variables (and therefore dimensions), more sophisticated and computationally intensive algorithms [14] and visualization methods [15] are required. Unsupervised methods are often used in an exploratory way, rather than to yield definitive conclusions, and output is highly dependent on the algorithm and hyperparameters selected.…”