Intermediate Respiratory Care Units (IRCUs) are vital during crises like COVID-19. This study evaluated clinical outcomes and operational dynamics of a new Spanish IRCU with specialized staffing. A prospective cohort study (April-August 2021) included 249 adult patients with COVID-19 respiratory failure (UHVN IRCU, Granada). Data on demographics, Non-Invasive Ventilation (NIV), length of stay (LOS), and outcomes (ICU transfer, exitus, recovery) were analyzed. Patient flow was simulated using a data-calibrated deterministic compartmental model (Ordinary Differential Equations, ODEs) that represented state transitions, and an empirical LOS-based stochastic convolution model that incorporated admission variability. The median age was 51; 31% of patients required NIV. NIV patients were older (median 61 vs 42, p<0.001). Overall, 8% needed ICU transfer; 3% experienced in-IRCU exitus. Notably, no ICU transfers or deaths occurred among 172 non-NIV patients. Of 77 high-risk NIV patients, 68% recovered in IRCU without ICU escalation. The ODE model, based on transition rates between patient states, reflected aggregate outcomes. Both modeling approaches demonstrated system strain during admission surges (partially mitigated by simulated care efficiency improvements via parameter modulation) and yielded consistent peak occupancy estimates. This IRCU, with specialized staffing, effectively managed severe COVID-19. High recovery rates, especially for NIV patients, potentially eased ICU pressure. Dynamic modeling confirmed surge vulnerability but highlighted the benefits of care efficiency from modulated transition parameters. Findings underscore positive outcomes in this IRCU model and support such units in pandemic response.
Noisy training labels can hurt model performance. Most approaches that aim to address label noise assume label noise is independent from the input features. In practice, however, label noise is often feature or \textit{instance-dependent}, and therefore biased (i.e., some instances are more likely to be mislabeled than others). E.g., in clinical care, female patients are more likely to be under-diagnosed for cardiovascular disease compared to male patients. Approaches that ignore this dependence can produce models with poor discriminative performance, and in many healthcare settings, can exacerbate issues around health disparities. In light of these limitations, we propose a two-stage approach to learn in the presence instance-dependent label noise. Our approach utilizes \textit{\anchor points}, a small subset of data for which we know the observed and ground truth labels. On several tasks, our approach leads to consistent improvements over the state-of-the-art in discriminative performance (AUROC) while mitigating bias (area under the equalized odds curve, AUEOC). For example, when predicting acute respiratory failure onset on the MIMIC-III dataset, our approach achieves a harmonic mean (AUROC and AUEOC) of 0.84 (SD [standard deviation] 0.01) while that of the next best baseline is 0.81 (SD 0.01). Overall, our approach improves accuracy while mitigating potential bias compared to existing approaches in the presence of instance-dependent label noise.
Determining clinically relevant physiological states from multivariate time series data with missing values is essential for providing appropriate treatment for acute conditions such as Traumatic Brain Injury (TBI), respiratory failure, and heart failure. Utilizing non-temporal clustering or data imputation and aggregation techniques may lead to loss of valuable information and biased analyses. In our study, we apply the SLAC-Time algorithm, an innovative self-supervision-based approach that maintains data integrity by avoiding imputation or aggregation, offering a more useful representation of acute patient states. By using SLAC-Time to cluster data in a large research dataset, we identified three distinct TBI physiological states and their specific feature profiles. We employed various clustering evaluation metrics and incorporated input from a clinical domain expert to validate and interpret the identified physiological states. Further, we discovered how specific clinical events and interventions can influence patient states and state transitions.
Models that can predict the occurrence of events ahead of time with low false-alarm rates are critical to the acceptance of decision support systems in the medical community. This challenging task is typically treated as a simple binary classification, ignoring temporal dependencies between samples, whereas we propose to exploit this structure. We first introduce a common theoretical framework unifying dynamic survival analysis and early event prediction. Following an analysis of objectives from both fields, we propose Temporal Label Smoothing (TLS), a simpler, yet best-performing method that preserves prediction monotonicity over time. By focusing the objective on areas with a stronger predictive signal, TLS improves performance over all baselines on two large-scale benchmark tasks. Gains are particularly notable along clinically relevant measures, such as event recall at low false-alarm rates. TLS reduces the number of missed events by up to a factor of two over previously used approaches in early event prediction.
The recent success of machine learning methods applied to time series collected from Intensive Care Units (ICU) exposes the lack of standardized machine learning benchmarks for developing and comparing such methods. While raw datasets, such as MIMIC-IV or eICU, can be freely accessed on Physionet, the choice of tasks and pre-processing is often chosen ad-hoc for each publication, limiting comparability across publications. In this work, we aim to improve this situation by providing a benchmark covering a large spectrum of ICU-related tasks. Using the HiRID dataset, we define multiple clinically relevant tasks in collaboration with clinicians. In addition, we provide a reproducible end-to-end pipeline to construct both data and labels. Finally, we provide an in-depth analysis of current state-of-the-art sequence modeling methods, highlighting some limitations of deep learning approaches for this type of data. With this benchmark, we hope to give the research community the possibility of a fair comparison of their work.
Objective: When patients develop acute respiratory failure, accurately identifying the underlying etiology is essential for determining the best treatment. However, differentiating between common medical diagnoses can be challenging in clinical practice. Machine learning models could improve medical diagnosis by aiding in the diagnostic evaluation of these patients. Materials and Methods: Machine learning models were trained to predict the common causes of acute respiratory failure (pneumonia, heart failure, and/or COPD). Models were trained using chest radiographs and clinical data from the electronic health record (EHR) and applied to an internal and external cohort. Results: The internal cohort of 1,618 patients included 508 (31%) with pneumonia, 363 (22%) with heart failure, and 137 (8%) with COPD based on physician chart review. A model combining chest radiographs and EHR data outperformed models based on each modality alone. Models had similar or better performance compared to a randomly selected physician reviewer. For pneumonia, the combined model area under the receiver operating characteristic curve (AUROC) was 0.79 (0.77-0.79), image model AUROC was 0.74 (0.72-0.75), and EHR model AUROC was 0.74 (0.70-0.76). For heart failure, combined: 0.83 (0.77-0.84), image: 0.80 (0.71-0.81), and EHR: 0.79 (0.75-0.82). For COPD, combined: AUROC = 0.88 (0.83-0.91), image: 0.83 (0.77-0.89), and EHR: 0.80 (0.76-0.84). In the external cohort, performance was consistent for heart failure and increased for COPD, but declined slightly for pneumonia. Conclusions: Machine learning models combining chest radiographs and EHR data can accurately differentiate between common causes of acute respiratory failure. Further work is needed to determine how these models could act as a diagnostic aid to clinicians in clinical settings.
Recognition of anomalous events is a challenging but critical task in many scientific and industrial fields, especially when the properties of anomalies are unknown. In this paper, we introduce a new anomaly concept called "unicorn" or unique event and present a new, model-free, unsupervised detection algorithm to detect unicorns. The key component of the new algorithm is the Temporal Outlier Factor (TOF) to measure the uniqueness of events in continuous data sets from dynamic systems. The concept of unique events differs significantly from traditional outliers in many aspects: while repetitive outliers are no longer unique events, a unique event is not necessarily an outlier; it does not necessarily fall out from the distribution of normal activity. The performance of our algorithm was examined in recognizing unique events on different types of simulated data sets with anomalies and it was compared with the Local Outlier Factor (LOF) and discord discovery algorithms. TOF had superior performance compared to LOF and discord algorithms even in recognizing traditional outliers and it also recognized unique events that those did not. The benefits of the unicorn concept and the new detection method were illustrated by example data sets from very different scientific fields. Our algorithm successfully recognized unique events in those cases where they were already known such as the gravitational waves of a binary black hole merger on LIGO detector data and the signs of respiratory failure on ECG data series. Furthermore, unique events were found on the LIBOR data set of the last 30 years.
Logistic regression (LR) is widely used in clinical prediction because it is simple to deploy and easy to interpret. Nevertheless, being a linear model, LR has limited expressive capability and often has unsatisfactory performance. Generalized additive models (GAMs) extend the linear model with transformations of input features, though feature interaction is not allowed for all GAM variants. In this paper, we propose a factored generalized additive model (F-GAM) to preserve the model interpretability for targeted features while allowing a rich model for interaction with features fixed within the individual. We evaluate F-GAM on prediction of two targets, postoperative acute kidney injury and acute respiratory failure, from a single-center database. We find superior model performance of F-GAM in terms of AUPRC and AUROC compared to several other GAM implementations, random forests, support vector machine, and a deep neural network. We find that the model interpretability is good with results with high face validity.
Recurrent neural networks (RNNs) are commonly applied to clinical time-series data with the goal of learning patient risk stratification models. Their effectiveness is due, in part, to their use of parameter sharing over time (i.e., cells are repeated hence the name recurrent). We hypothesize, however, that this trait also contributes to the increased difficulty such models have with learning relationships that change over time. Conditional shift, i.e., changes in the relationship between the input X and the output y, arises when risk factors associated with the event of interest change over the course of a patient admission. While in theory, RNNs and gated RNNs (e.g., LSTMs) in particular should be capable of learning time-varying relationships, when training data are limited, such models often fail to accurately capture these dynamics. We illustrate the advantages and disadvantages of complete parameter sharing (RNNs) by comparing an LSTM with shared parameters to a sequential architecture with time-varying parameters on prediction tasks involving three clinically-relevant outcomes: acute respiratory failure (ARF), shock, and in-hospital mortality. In experiments using synthetic data, we demonstrate how parameter sharing in LSTMs leads to worse performance in the presence of conditional shift. To improve upon the dichotomy between complete parameter sharing and no parameter sharing, we propose a novel RNN formulation based on a mixture model in which we relax parameter sharing over time. The proposed method outperforms standard LSTMs and other state-of-the-art baselines across all tasks. In settings with limited data, relaxed parameter sharing can lead to improved patient risk stratification performance.