The impact of recency and adequacy of historical information on sepsis predictions using machine learning

Abstract

Sepsis is a major public and global health concern. Every hour of delay in detecting sepsis significantly increases the risk of death, highlighting the importance of accurately predicting sepsis in a timely manner. A growing body of literature has examined developing new or improving the existing machine learning (ML) approaches for timely and accurate predictions of sepsis. This study contributes to this literature by providing clear insights regarding the role of the recency and adequacy of historical information in predicting sepsis using ML. To this end, we implemented a deep learning model using a bidirectional long short-term memory (BiLSTM) algorithm and compared it with six other ML algorithms based on numerous combinations of the prediction horizons (to capture information recency) and observation windows (to capture information adequacy) using different measures of predictive performance. Our results indicated that the BiLSTM algorithm outperforms all other ML algorithms and provides a great separability of the predicted risk of sepsis among septic versus non-septic patients. Moreover, decreasing the prediction horizon (in favor of information recency) always boosts the predictive performance; however, the impact of expanding the observation window (in favor of information adequacy) depends on the prediction horizon and the purpose of prediction. More specifically, when the prediction is responsive to the positive label (i.e., Sepsis), increasing historical data improves the predictive performance when the prediction horizon is short-moderate.

Predicting Blood Pressure After Nitroglycerin Infusion Dose Titration in Critical Care Units

“…The feature importance plot for the best-performing model showed that information about previous blood pressures was more important for predicting posttitration blood pressure than nitroglycerin dosing information. These results are expected because prior and average values of the prediction target (in our case, systolic blood pressure) are known to be informative for modeling tasks 19,20 . It should be noted that permutation importance methods reveal how important particular features are to a model, not the predictive feature itself 11 .…”

“…These results are expected because prior and average values of the prediction target (in our case, systolic blood pressure) are known to be informative for modeling tasks. 19,20 It should be noted that permutation importance methods reveal how important particular features are to a model, not the predictive feature itself. 11 As such, although the features related to nitroglycerin dose titration seemed to contribute only minimally for the best-performing model in our analyses, it does not mean this information is not helpful at all for the task of systolic blood pressure prediction during nitroglycerin infusion.…”

Identification and Prediction of Clinical Phenotypes in Hospitalized Patients With COVID-19: Machine Learning From Medical Records

“…Moreover, although we used observations recorded within the first 6-hour window following admission to derive our predictive model, multisite studies have shown that the mean length of stay for patients with COVID-19 requiring ICU-level care ranges from 12 to 19 days. This suggests that predictive models trained using data over longer intervals (eg, recorded within 24 hours following admission, decreasing the prediction horizon [ 93 ]) or updating the prediction longitudinally [ 94 ] may lead to clinically useful models with improved prediction performance [ 95 ].…”

Deep Residual Networks and Bayesian Data Priors in the Survival Prediction and Classification

“…Sepsis and septic shock affect millions of people worldwide each year, with one-third to one-sixth of them dying [2]. Every hour delay in detection of sepsis significantly increases the risk of death [3]. Physicianpatient interactions significantly affect mortality, which decreases with increased medical manipulation.…”

“…With the rise of deep learning, a series of deep learning-based methods have been developed for early prediction of sepsis or risk prediction. An event embedding and temporal encoding model based on long and short-term memory to model clinical time series of sepsis patients and improve prediction performance [11]; convolutional neural network to predict 28-day survival of sepsis patients by 35 blood detection variables [11]; BiLSTM for predicting the probability of sepsis [12]; DSPA method using convolutional neural network to extract features using random forest algorithm to predict SOFA scores of sepsis patients [13]. Deep learning models are very prone to gradient disappearance during backpropagation over the network and also very prone to the problem of gradient explosion when the initial weights of the network are too large.…”