An official website of the United States government
Abstract Traditional methods for assessing illness severity and predicting in-hospital mortality among critically ill patients require time-consuming, error-prone calculations using static variable thresholds. These methods do not capitalize on the emerging availability of streaming electronic health record data or capture time-sensitive individual physiological patterns, a critical task in the intensive care unit. We propose a novel acuity score framework (DeepSOFA) that leverages temporal measurements and interpretable deep learning models to assess illness severity at any point during an ICU stay. We compare DeepSOFA with SOFA (Sequential Organ Failure Assessment) baseline models using the same model inputs and find that at any point during an ICU admission, DeepSOFA yields significantly more accurate predictions of in-hospital mortality. A DeepSOFA model developed in a public database and validated in a single institutional cohort had a mean AUC for the entire ICU stay of 0.90 (95% CI 0.90–0.91) compared with baseline SOFA models with mean AUC 0.79 (95% CI 0.79–0.80) and 0.85 (95% CI 0.85–0.86). Deep models are well-suited to identify ICU patients in need of life-saving interventions prior to the occurrence of an unexpected adverse event and inform shared decision-making processes among patients, providers, and families regarding goals of care and optimal resource utilization.
more »
« less
This content will become publicly available on November 23, 2026
Temporal Concept Tracing: Making Deep Learning PredictionsInterpretable and Actionable for ICU Acute Kidney InjuryPrevention
Deep learning models have demonstrated impressive accuracy in predicting acute kidney injury (AKI), a condition affecting up to 20% of ICU patients, yet their black-box nature prevents clinical adoption in high-stakes critical care settings. While existing interpretability methods like SHAP, LIME, and attention mechanisms can identify important features, they fail to capture the temporal dynamics essential for clinical decision-making, and are unable to communicate when specific risk factors become critical in a patient's trajectory. This limitation is particularly problematic in the ICU, where the timing of interventions can significantly impact patient outcomes. We present a novel interpretable framework that brings temporal awareness to deep learning predictions for AKI. Our approach introduces three key innovations: (1) a latent convolutional concept bottleneck that learns clinically meaningful patterns from ICU time-series without requiring manual concept annotation, leveraging Conv1D layers to capture localized temporal patterns like sudden physiological changes; (2) Temporal Concept Tracing (TCT), a gradient-based method that identifies not only which risk factors matter but precisely when they become critical addressing the fundamental question of temporal relevance missing from current XAI techniques; and (3) integration with MedAlpaca to generate structured, time-aware clinical explanations that translate model insights into actionable bedside guidance. We evaluate our framework on MIMIC-IV data, demonstrating that our approach performs better than existing explainability frameworks, Occlusion and LIME, in terms of the comprehensiveness score, sufficiency score, and processing time. The proposed method also better captures risk factors inflection points for patients timelines compared to conventional concept bottleneck methods, including dense layer and attention mechanism. This work represents the first comprehensive solution for interpretable temporal deep learning in critical care that addresses both the what and when of clinical risk factors. By making AKI predictions transparent and temporally contextualized, our framework bridges the gap between model accuracy and clinical utility, offering a path toward trustworthy AI deployment in time-sensitive healthcare settings.
more »
« less
- PAR ID:
- 10650154
- Publisher / Repository:
- AAAI
- Date Published:
- Journal Name:
- Proceedings of the AAAI Symposium Series
- Volume:
- 7
- Issue:
- 1
- ISSN:
- 2994-4317
- Page Range / eLocation ID:
- 448 to 455
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Patients staying in the Intensive Care Unit (ICU) have a severely disrupted circadian rhythm. Due to patients' critical medical condition, ICU physicians and nurses have to provide round-the-clock clinical care, further disrupting patients' circadian rhythm. Mistimed family visits during rest-time can also disrupt patients' circadian rhythm. Currently, such effects are only reported based on hospital visitation policies rather than the actual number of visitors and care providers in the room. To quantify visitation disruptions, we used a deep Mask R-CNN model, a deep learning framework for object instance segmentation to detect and quantify the number of individuals in the ICU unit. This study represents the first effort to automatically quantify visitations in an ICU room, which could have implications in terms of policy adjustment, as well as circadian rhythm investigation. Our model achieved precision of 0.97 and recall of 0.67, with F1 score of 0.79 for detecting disruptions in the ICU units.more » « less
-
Frasch, Martin G. (Ed.)With the wider availability of healthcare data such as Electronic Health Records (EHR), more and more data-driven based approaches have been proposed to improve the quality-of-care delivery. Predictive modeling, which aims at building computational models for predicting clinical risk, is a popular research topic in healthcare analytics. However, concerns about privacy of healthcare data may hinder the development of effective predictive models that are generalizable because this often requires rich diverse data from multiple clinical institutions. Recently, federated learning (FL) has demonstrated promise in addressing this concern. However, data heterogeneity from different local participating sites may affect prediction performance of federated models. Due to acute kidney injury (AKI) and sepsis’ high prevalence among patients admitted to intensive care units (ICU), the early prediction of these conditions based on AI is an important topic in critical care medicine. In this study, we take AKI and sepsis onset risk prediction in ICU as two examples to explore the impact of data heterogeneity in the FL framework as well as compare performances across frameworks. We built predictive models based on local, pooled, and FL frameworks using EHR data across multiple hospitals. The local framework only used data from each site itself. The pooled framework combined data from all sites. In the FL framework, each local site did not have access to other sites’ data. A model was updated locally, and its parameters were shared to a central aggregator, which was used to update the federated model’s parameters and then subsequently, shared with each site. We found models built within a FL framework outperformed local counterparts. Then, we analyzed variable importance discrepancies across sites and frameworks. Finally, we explored potential sources of the heterogeneity within the EHR data. The different distributions of demographic profiles, medication use, and site information contributed to data heterogeneity.more » « less
-
Accurate prediction and monitoring of patient health in the intensive care unit can inform shared decisions regarding appropriateness of care delivery, risk-reduction strategies, and intensive care resource use. Traditionally, algorithmic solutions for patient outcome prediction rely solely on data available from electronic health records (EHR). In this pilot study, we explore the benefits of augmenting existing EHR data with novel measurements from wrist-worn activity sensors as part of a clinical environment known as the Intelligent ICU. We implemented temporal deep learning models based on two distinct sources of patient data: (1) routinely measured vital signs from electronic health records, and (2) activity data collected from wearable sensors. As a proxy for illness severity, our models predicted whether patients leaving the intensive care unit would be successfully or unsuccessfully discharged from the hospital. We overcome the challenge of small sample size in our prospective cohort by applying deep transfer learning using EHR data from a much larger cohort of traditional ICU patients. Our experiments quantify added utility of non-traditional measurements for predicting patient health, especially when applying a transfer learning procedure to small novel Intelligent ICU cohorts of critically ill patients.more » « less
-
Background Heart failure is a leading cause of mortality and morbidity worldwide. Acute heart failure, broadly defined as rapid onset of new or worsening signs and symptoms of heart failure, often requires hospitalization and admission to the intensive care unit (ICU). This acute condition is highly heterogeneous and less well-understood as compared to chronic heart failure. The ICU, through detailed and continuously monitored patient data, provides an opportunity to retrospectively analyze decompensation and heart failure to evaluate physiological states and patient outcomes. Objective The goal of this study is to examine the prevalence of cardiovascular risk factors among those admitted to ICUs and to evaluate combinations of clinical features that are predictive of decompensation events, such as the onset of acute heart failure, using machine learning techniques. To accomplish this objective, we leveraged tele-ICU data from over 200 hospitals across the United States. Methods We evaluated the feasibility of predicting decompensation soon after ICU admission for 26,534 patients admitted without a history of heart failure with specific heart failure risk factors (ie, coronary artery disease, hypertension, and myocardial infarction) and 96,350 patients admitted without risk factors using remotely monitored laboratory, vital signs, and discrete physiological measurements. Multivariate logistic regression and random forest models were applied to predict decompensation and highlight important features from combinations of model inputs from dissimilar data. Results The most prevalent risk factor in our data set was hypertension, although most patients diagnosed with heart failure were admitted to the ICU without a risk factor. The highest heart failure prediction accuracy was 0.951, and the highest area under the receiver operating characteristic curve was 0.9503 with random forest and combined vital signs, laboratory values, and discrete physiological measurements. Random forest feature importance also highlighted combinations of several discrete physiological features and laboratory measures as most indicative of decompensation. Timeline analysis of aggregate vital signs revealed a point of diminishing returns where additional vital signs data did not continue to improve results. Conclusions Heart failure risk factors are common in tele-ICU data, although most patients that are diagnosed with heart failure later in an ICU stay presented without risk factors making a prediction of decompensation critical. Decompensation was predicted with reasonable accuracy using tele-ICU data, and optimal data extraction for time series vital signs data was identified near a 200-minute window size. Overall, results suggest combinations of laboratory measurements and vital signs are viable for early and continuous prediction of patient decompensation.more » « less
