A multi-cohort study on prediction of acute brain dysfunction states using selective state space models

Brandon Silva,Miguel Contreras,Sabyasachi Bandyopadhyay,Yuanfang Ren,Ziyuan Guan,Jeremy Balch,Kia Khezeli,Tezcan Ozrazgat Baslanti,Ben Shickel,Azra Bihorac,Parisa Rashidi

2024-03-12

Abstract:Assessing acute brain dysfunction (ABD), including delirium and coma in the intensive care unit (ICU), is a critical challenge due to its prevalence and severe implications for patient outcomes. Current diagnostic methods rely on infrequent clinical observations, which can only determine a patient's ABD status after onset. Our research attempts to solve these problems by harnessing Electronic Health Records (EHR) data to develop automated methods for ABD prediction for patients in the ICU. Existing models solely predict a single state (e.g., either delirium or coma), require at least 24 hours of observation data to make predictions, do not dynamically predict fluctuating ABD conditions during ICU stay (typically a one-time prediction), and use small sample size, proprietary single-hospital datasets. Our research fills these gaps in the existing literature by dynamically predicting delirium, coma, and mortality for 12-hour intervals throughout an ICU stay and validating on two public datasets. Our research also introduces the concept of dynamically predicting critical transitions from non-ABD to ABD and between different ABD states in real time, which could be clinically more informative for the hospital staff. We compared the predictive performance of two state-of-the-art neural network models, the MAMBA selective state space model and the Longformer Transformer model. Using the MAMBA model, we achieved a mean area under the receiving operator characteristic curve (AUROC) of 0.95 on outcome prediction of ABD for 12-hour intervals. The model achieves a mean AUROC of 0.79 when predicting transitions between ABD states. Our study uses a curated dataset from the University of Florida Health Shands Hospital for internal validation and two publicly available datasets, MIMIC-IV and eICU, for external validation, demonstrating robustness across ICU stays from 203 hospitals and 140,945 patients.

Artificial Intelligence,Applications,Machine Learning

What problem does this paper attempt to address?

The key problem that this paper attempts to solve is the early prediction and dynamic monitoring of acute brain dysfunction (ABD, including delirium and coma) in the intensive care unit (ICU). Specifically, the research aims to solve the following problems by using electronic health record (EHR) data to develop automated methods: 1. **Existing diagnostic methods rely on infrequent clinical observations**: These methods can only determine the patient's ABD status after ABD occurs and cannot provide early warnings. 2. **Manual observations increase the workload of ICU nurses**: This may lead to nurse fatigue and reduce the overall quality of care. 3. **Limitations of existing models**: - Most existing models only predict a single state (for example, only predict delirium or coma), rather than predicting multiple states simultaneously. - At least 24 hours of observation data are required to make a prediction, and real - time dynamic prediction cannot be made. - Usually only one prediction is made, and the state changes of patients during the entire ICU stay cannot be continuously monitored. - The use of small sample sizes and single - hospital private datasets limits the generalization ability and reproducibility of the models. To overcome these problems, the study proposes the following improvement measures: - **Dynamically predict ABD status**: Use selective state - space models (such as MAMBA) and Transformer models to dynamically predict ABD status for each 12 - hour time window. - **Predict state transitions**: Not only predict ABD status, but also predict transitions from non - ABD to ABD and between different ABD states to provide more detailed clinical information. - **Use large - scale public datasets for verification**: Conduct external verification on two large - public datasets (MIMIC - IV and eICU) to ensure the generalization and robustness of the model. - **Real - time performance evaluation**: Evaluate the real - time prediction performance of the model in the actual ICU environment to ensure its effectiveness in practical applications. Through these improvements, the study aims to improve the early prediction ability of ABD in the ICU, thereby improving patient prognosis and reducing the occurrence of negative outcomes such as self - harm, coma, and death.

A multi-cohort study on prediction of acute brain dysfunction states using selective state space models

Transformer Models for Acute Brain Dysfunction Prediction

Delirium prediction in the ICU: designing a screening tool for preventive interventions

APRICOT-Mamba: Acuity Prediction in Intensive Care Unit (ICU): Development and Validation of a Stability, Transitions, and Life-Sustaining Therapies Prediction Model

An attention based deep learning model of clinical events in the intensive care unit

A Long Short-Term Memory Ensemble Approach for Improving the Outcome Prediction in Intensive Care Unit

A Knowledge Distillation Ensemble Framework for Predicting Short and Long-term Hospitalisation Outcomes from Electronic Health Records Data

Early Prediction of Cardiac Arrest in the Intensive Care Unit Using Explainable Machine Learning: Retrospective Study

Early prediction of delirium upon intensive care unit admission: Model development, validation, and deployment

Using Machine Learning and Electronic Health Records to Identify Neuropsychiatric Risk Scores for Delirium in ICU and General Hospital Settings

Monitoring ICU Mortality Risk with A Long Short-Term Memory Recurrent Neural Network

RAIM: Recurrent Attentive and Intensive Model of Multimodal Patient Monitoring Data

Merging Clinical and EEG Biomarkers in an Elastic-Net Regression for Disorder of Consciousness Prognosis Prediction

Improving ICU Risk Predictive Models Through Automation Designed for Resiliency Against Documentation Bias

Multimodal Prediction of 3- and 12-Month Outcomes in ICU Patients with Acute Disorders of Consciousness

DeLLiriuM: A large language model for delirium prediction in the ICU using structured EHR

Predicting Intensive Care Unit Length of Stay and Mortality Using Patient Vital Signs: Machine Learning Model Development and Validation

Forecasting acute kidney injury and resource utilization in ICU patients using longitudinal, multimodal models

Prediction of In-Hospital Cardiac Arrest in the Intensive Care Unit: Machine Learning-Based Multimodal Approach

New onset delirium prediction using machine learning and long short-term memory (LSTM) in electronic health record

The development and validation of a novel deep-learning algorithm to predict in-hospital cardiac arrest in ED-ICU (emergency department-based intensive care units): a single center retrospective cohort study