Estimating the probability of early afterdepolarizations and predicting arrhythmic risk associated with long QT syndrome type 1 mutations.

Early afterdepolarizations (EADs) are action potential (AP) repolarization abnormalities that can trigger lethal arrhythmias. Simulations using biophysically detailed cardiac myocyte models can reveal how model parameters influence the probability of these cellular arrhythmias; however, such analyses can pose a huge computational burden. We have previously developed a highly simplified approach in which logistic regression models (LRMs) map parameters of complex cell models to the probability of ectopic beats. Here, we extend this approach to predict the probability of EADs (P(EAD)) as a mechanistic metric of arrhythmic risk. We use the LRM to investigate how changes in parameters of the slow-activating delayed rectifier current (IKs) affect P(EAD) for 17 different long QT syndrome type 1 (LQTS1) mutations. In this LQTS1 clinical arrhythmic risk prediction task, we compared P(EAD) for these 17 mutations with two other recently published model-based arrhythmia risk metrics (AP morphology metric across populations of myocyte models and transmural repolarization prolongation based on a one-dimensional [1D] tissue-level model). These model-based risk metrics yield similar prediction performance; however, each fails to stratify clinical risk for a significant number of the 17 studied LQTS1 mutations. Nevertheless, an interpretable ensemble model using multivariate linear regression built by combining all of these model-based risk metrics successfully predicts the clinical risk of 17 mutations. These results illustrate the potential of computational approaches in arrhythmia risk prediction.

A data-driven model for early prediction of need for invasive mechanical ventilation in pediatric intensive care unit patients.

RATIONALE: Acute respiratory failure is a life-threatening clinical outcome in critically ill pediatric patients. In severe cases, patients can require mechanical ventilation (MV) for survival. Early recognition of these patients can potentially help clinicians alter the clinical course and lead to improved outcomes. OBJECTIVES: To build a data-driven model for early prediction of the need for mechanical ventilation in pediatric intensive care unit (PICU) patients. METHODS: The study consists of a single-center retrospective observational study on a cohort of 13,651 PICU patients admitted between 1/01/2010 and 5/15/2018 with a prevalence of 8.06% for MV due to respiratory failure. XGBoost (extreme gradient boosting) and a convolutional neural network (CNN) using medication history were used to develop a prediction model that could yield a time-varying "risk-score"-a continuous probability of whether a patient will receive MV-and an ideal global threshold was calculated from the receiver operating characteristics (ROC) curve. The early prediction point (EPP) was the first time the risk-score surpassed the optimal threshold, and the interval between the EPP and the start of the MV was the early warning period (EWT). Spectral clustering identified patient groups based on risk-score trajectories after EPP. RESULTS: A clinical and medication history-based model achieved a 0.89 area under the ROC curve (AUROC), 0.6 sensitivity, 0.95 specificity, 0.55 positive predictive value (PPV), and 0.95 negative predictive value (NPV). Early warning time (EWT) median [inter-quartile range] of this model was 9.9[4.2-69.2] hours. Clustering risk-score trajectories within a six-hour window after the early prediction point (EPP) established three patient groups, with the highest risk group's PPV being 0.92. CONCLUSIONS: This study uses a unique method to extract and apply medication history information, such as time-varying variables, to identify patients who may need mechanical ventilation for respiratory failure and provide an early warning period to avert it.

Predicting respiratory decompensation in mechanically ventilated adult ICU patients.

Introduction: Mechanical ventilation is a life-saving treatment in the Intensive Care Unit (ICU), but often causes patients to be at risk of further respiratory complication. We created a statistical model utilizing electronic health record and physiologic vitals data to predict the Center for Disease Control and Prevention (CDC) defined Ventilator Associated Complications (VACs). Further, we evaluated the effect of data temporal resolution and feature generation method choice on the accuracy of such a constructed model. Methods: We constructed a random forest model to predict occurrence of VACs using health records and chart events from adult patients in the Medical Information Mart for Intensive Care III (MIMIC-III) database. We trained the machine learning models on two patient populations of 1921 and 464 based on low and high frequency data availability. Model features were generated using both basic statistical summaries and tsfresh, a python library that generates a large number of derived time-series features. Classification to determine whether a patient will experience VAC one hour after 35 h of ventilation was performed using a random forest classifier. Two different sample spaces conditioned on five varying feature extraction techniques were evaluated to identify the most optimal selection of features resulting in the best VAC discrimination. Each dataset was assessed using K-folds cross-validation (k = 10), giving average area under the receiver operating characteristic curves (AUROCs) and accuracies. Results: After feature selection, hyperparameter tuning, and feature extraction, the best performing model used automatically generated features on high frequency data and achieved an average AUROC of 0.83 ± 0.11 and an average accuracy of 0.69 ± 0.10. Discussion: Results show the potential viability of predicting VACs using machine learning, and indicate that higher-resolution data and the larger feature set generated by tsfresh yield better AUROCs compared to lower-resolution data and manual statistical features.

Factors associated with resistance to SARS-CoV-2 infection discovered using large-scale medical record data and machine learning.

There have been over 621 million cases of COVID-19 worldwide with over 6.5 million deaths. Despite the high secondary attack rate of COVID-19 in shared households, some exposed individuals do not contract the virus. In addition, little is known about whether the occurrence of COVID-19 resistance differs among people by health characteristics as stored in the electronic health records (EHR). In this retrospective analysis, we develop a statistical model to predict COVID-19 resistance in 8,536 individuals with prior COVID-19 exposure using demographics, diagnostic codes, outpatient medication orders, and count of Elixhauser comorbidities in EHR data from the COVID-19 Precision Medicine Platform Registry. Cluster analyses identified 5 patterns of diagnostic codes that distinguished resistant from non-resistant patients in our study population. In addition, our models showed modest performance in predicting COVID-19 resistance (best performing model AUROC = 0.61). Monte Carlo simulations conducted indicated that the AUROC results are statistically significant (p < 0.001) for the testing set. We hope to validate the features found to be associated with resistance/non-resistance through more advanced association studies.

A pilot study to predict cardiac arrest in the pediatric intensive care unit.

BACKGROUND: Cardiac arrest is a leading cause of mortality prior to discharge for children admitted to the pediatric intensive care unit. To address this problem, we used machine learning to predict cardiac arrest up to three hours in advance. METHODS: Our data consists of 240 Hz ECG waveform data, 0.5 Hz physiological time series data, medications, and demographics from 1,145 patients in the pediatric intensive care unit at the Johns Hopkins Hospital, 15 of whom experienced a cardiac arrest. The data were divided into training, validating, and testing sets, and features were generated every five minutes. 23 heart rate variability (HRV) metrics were determined from ECG waveforms. 96 summary statistics were calculated for 12 vital signs, such as respiratory rate and blood pressure. Medications were classified into 42 therapeutic drug classes. Binary features were generated to indicate the administration of these different drugs. Next, six machine learning models were evaluated: logistic regression, support vector machine, random forest, XGBoost, LightGBM, and a soft voting ensemble. RESULTS: XGBoost performed the best, with 0.971 auROC, 0.797 auPRC, 99.5% sensitivity, and 69.6% specificity on an independent test set. CONCLUSION: We have created high-performing models that identify signatures of in-hospital cardiac arrest (IHCA) that may not be evident to clinicians. These signatures include a combination of heart rate variability metrics, vital signs data, and therapeutic drug classes. These machine learning models can predict IHCA up to three hours prior to onset with high performance, allowing clinicians to intervene earlier, improving patient outcomes.

Bone Morphology and Vascular Supply of Pedicled Distal Radius Bone Using Nano-Computed Tomography.

BACKGROUND: The goal of this study was to use nano-computed tomography to describe the intraosseous vascularity and structural characteristics of commonly used distal radius vascularized bone grafts for treatment of scaphoid nonunion. METHODS: We obtained 8 fresh frozen human cadaver forearm specimens for infusion of barium contrast. Specimens were scanned and segmented to quantify the vascular volume and trabecular density within 3 common graft regions, including 1, 2 intercompartmental supraretinacular artery (1,2 ICSRA), fourth extensor compartment artery (4 ECA), and volar carpal artery (VCA), as well as thirds of the scaphoid. Outcomes also included mean and maximum cortical thickness and number of cortical perforators. Single-specimen analyses were also performed comparing vascularity and trabecular density of each graft with scaphoid regions of a single specimen. Statistical analysis was performed using analysis of variance with post hoc Tukey testing when P value was less than .05. RESULTS: There was no significant difference between groups in the mean percent vascularity (P = .76). The ratio of trabecular bone in each graft to scaphoid thirds was less than 1. The mean cortical thickness (0.79 mm, 95% confidence interval [CI], 0.66-0.93 mm) and maximum cortical thickness (1.45 mm, 95% CI, 1.27-1.63 mm) of VCA grafts were both significantly greater than those of 4 ECA and 1,2 ICSRA (P < .001). CONCLUSIONS: There were no differences between vascular density of the 3 grafts and the scaphoid. Pedicled distal radius bone grafts have similar vascularity but morphometric differences such as cortical thickness and trabecular density which have unclear clinical implications.

Deep Learning Model for Static Ocular Torsion Detection Using Synthetically Generated Fundus Images.

Purpose: The objective of the study is to develop deep learning models using synthetic fundus images to assess the direction (intorsion versus extorsion) and amount (physiologic versus pathologic) of static ocular torsion. Static ocular torsion assessment is an important clinical tool for classifying vertical ocular misalignment; however, current methods are time-intensive with steep learning curves for frontline providers. Methods: We used a dataset (n = 276) of right eye fundus images. The disc-foveal angle was calculated using ImageJ to generate synthetic images via image rotation. Using synthetic datasets (n = 12,740 images per model) and transfer learning (the reuse of a pretrained deep learning model on a new task), we developed a binary classifier (intorsion versus extorsion) and a multiclass classifier (physiologic versus pathologic intorsion and extorsion). Model performance was evaluated on unseen synthetic and nonsynthetic data. Results: On the synthetic dataset, the binary classifier had an accuracy and area under the receiver operating characteristic curve (AUROC) of 0.92 and 0.98, respectively, whereas the multiclass classifier had an accuracy and AUROC of 0.77 and 0.94, respectively. The binary classifier generalized well on the nonsynthetic data (accuracy = 0.94; AUROC = 1.00). Conclusions: The direction of static ocular torsion can be detected from synthetic fundus images using deep learning methods, which is key to differentiate between vestibular misalignment (skew deviation) and ocular muscle misalignment (superior oblique palsies). Translational Relevance: Given the robust performance of our models on real fundus images, similar strategies can be adopted for deep learning research in rare neuro-ophthalmologic diseases with limited datasets.

Predicting Intensive Care Delirium with Machine Learning: Model Development and External Validation.

BACKGROUND: Delirium poses significant risks to patients, but countermeasures can be taken to mitigate negative outcomes. Accurately forecasting delirium in intensive care unit (ICU) patients could guide proactive intervention. Our primary objective was to predict ICU delirium by applying machine learning to clinical and physiologic data routinely collected in electronic health records. METHODS: Two prediction models were trained and tested using a multicenter database (years of data collection 2014 to 2015), and externally validated on two single-center databases (2001 to 2012 and 2008 to 2019). The primary outcome variable was delirium defined as a positive Confusion Assessment Method for the ICU screen, or an Intensive Care Delirium Screening Checklist of 4 or greater. The first model, named "24-hour model," used data from the 24 h after ICU admission to predict delirium any time afterward. The second model designated "dynamic model," predicted the onset of delirium up to 12 h in advance. Model performance was compared with a widely cited reference model. RESULTS: For the 24-h model, delirium was identified in 2,536 of 18,305 (13.9%), 768 of 5,299 (14.5%), and 5,955 of 36,194 (11.9%) of patient stays, respectively, in the development sample and two validation samples. For the 12-h lead time dynamic model, delirium was identified in 3,791 of 22,234 (17.0%), 994 of 6,166 (16.1%), and 5,955 of 28,440 (20.9%) patient stays, respectively. Mean area under the receiver operating characteristics curve (AUC) (95% CI) for the first 24-h model was 0.785 (0.769 to 0.801), significantly higher than the modified reference model with AUC of 0.730 (0.704 to 0.757). The dynamic model had a mean AUC of 0.845 (0.831 to 0.859) when predicting delirium 12 h in advance. Calibration was similar in both models (mean Brier Score [95% CI] 0.102 [0.097 to 0.108] and 0.111 [0.106 to 0.116]). Model discrimination and calibration were maintained when tested on the validation datasets. CONCLUSIONS: Machine learning models trained with routinely collected electronic health record data accurately predict ICU delirium, supporting dynamic time-sensitive forecasting.

Biomechanical Comparison of a Novel Multiplanar, Perpendicular Whipstitch With the Krackow Stitch and Standard Commercial Whipstitch.

Background: Using alternating orthogonal suture throws with the looped whipstitch technique may allow enhanced suture fixation. Hypothesis: It was hypothesized that this novel multiplanar, perpendicular looped whipstitch (MP) technique would have improved biomechanical properties compared with the standard looped whipstitch (WS) and Krackow stitch (KS). Study Design: Controlled laboratory study. Methods: A total of 30 cadaveric tibialis anterior tendons were randomly assigned into 3 groups of 10. Tendons were secured to a custom clamp, and the other end was sutured using 1 of 3 techniques: the KS, WS, or novel MP. The MP was performed with alternating orthogonal throws starting right to left, then front to back, left to right, and back to front. Each technique used 4 passes of No. 2 FiberWire spaced 5 mm apart and ending 10 mm from the tendon end. Tendons were preloaded to 5 N, pretensioned to 50 N at 100 mm/min for 3 cycles, returned to 5 N for 1 minute, cycled from 5 to 100 N at 200 mm/min for 100 cycles, and then loaded to failure at 20 mm/min. Elongation was recorded after pretensioning and cycling and was measured both across the suture-tendon interface and from the base of the suture-tendon interface to markings on the suture limbs (construct elongation). One-way analyses of variance were performed, with Bonferroni post hoc analysis when appropriate. Results: There were no differences in cross-sectional area or stiffness among the 3 techniques. The ultimate load for WS (183.33 ± 57.44 N) was less compared with both MP (270.76 ± 39.36 N) and KS (298.90 ± 25.94 N) (P ≤ .001 for both). There was less construct elongation for KS compared with WS and MP for total displacement, measured from pretensioning to the end of cycling (P < .001). All 3 techniques saw a decrease in length (shortening) at the suture-tendon interface during testing. There was more shortening at the suture-tendon interface for WS compared with KS (P = .006). Conclusion: The KS appears superior, as it maximized strength while minimizing construct elongation or graft shortening. The ultimate load of the MP technique was greater than that of the standard technique but not significantly different from that of the KS technique. Clinical Relevance: The KS is preferred. If using a WS, multiplanar, perpendicular passes should be considered.

aEYE: A deep learning system for video nystagmus detection.

Background: Nystagmus identification and interpretation is challenging for non-experts who lack specific training in neuro-ophthalmology or neuro-otology. This challenge is magnified when the task is performed via telemedicine. Deep learning models have not been heavily studied in video-based eye movement detection. Methods: We developed, trained, and validated a deep-learning system (aEYE) to classify video recordings as normal or bearing at least two consecutive beats of nystagmus. The videos were retrospectively collected from a subset of the monocular (right eye) video-oculography (VOG) recording used in the Acute Video-oculography for Vertigo in Emergency Rooms for Rapid Triage (AVERT) clinical trial (#NCT02483429). Our model was derived from a preliminary dataset representing about 10% of the total AVERT videos (n = 435). The videos were trimmed into 10-sec clips sampled at 60 Hz with a resolution of 240 × 320 pixels. We then created 8 variations of the videos by altering the sampling rates (i.e., 30 Hz and 15 Hz) and image resolution (i.e., 60 × 80 pixels and 15 × 20 pixels). The dataset was labeled as "nystagmus" or "no nystagmus" by one expert provider. We then used a filtered image-based motion classification approach to develop aEYE. The model's performance at detecting nystagmus was calculated by using the area under the receiver-operating characteristic curve (AUROC), sensitivity, specificity, and accuracy. Results: An ensemble between the ResNet-soft voting and the VGG-hard voting models had the best performing metrics. The AUROC, sensitivity, specificity, and accuracy were 0.86, 88.4, 74.2, and 82.7%, respectively. Our validated folds had an average AUROC, sensitivity, specificity, and accuracy of 0.86, 80.3, 80.9, and 80.4%, respectively. Models created from the compressed videos decreased in accuracy as image sampling rate decreased from 60 Hz to 15 Hz. There was only minimal change in the accuracy of nystagmus detection when decreasing image resolution and keeping sampling rate constant. Conclusion: Deep learning is useful in detecting nystagmus in 60 Hz video recordings as well as videos with lower image resolutions and sampling rates, making it a potentially useful tool to aid future automated eye-movement enabled neurologic diagnosis.

Computational signatures for post-cardiac arrest trajectory prediction: Importance of early physiological time series.

BACKGROUND: There is an unmet need for timely and reliable prediction of post-cardiac arrest (CA) clinical trajectories. We hypothesized that physiological time series (PTS) data recorded on the first day of intensive care would contribute significantly to discrimination of outcomes at discharge. PATIENTS AND METHODS: Adult patients in the multicenter eICU database who were mechanically ventilated after resuscitation from out-of-hospital CA were included. Outcomes of interest were survival, neurological status based on Glasgow motor subscore (mGCS) and surrogate functional status based on discharge location (DL), at hospital discharge. Three machine learning predictive models were trained, one with features from the electronic health records (EHR), the second using features derived from PTS collected in the first 24 h after ICU admission (PTS24), and the third combining PTS24 and EHR. Model performances were compared, and the best performing model was externally validated in the MIMIC-III dataset. RESULTS: Data from 2216 admissions were included in the analysis. Discrimination of prediction models combining EHR and PTS24 features was higher than models using either EHR or PTS24 for prediction of survival (AUROC 0.83, 0.82 and 0.79 respectively), neurological outcome (0.87, 0.86 and 0.79 respectively), and DL (0.80, 0.78 and 0.76 respectively). External validation in MIMIC-III (n = 86) produced similar model performance. Feature analysis suggested prognostic significance of previously unknown EHR and PTS24 variables. CONCLUSION: These results indicate that physiological data recorded in the early phase after CA resuscitation contain signatures that are linked to post-CA outcome. Additionally, they attest to the effectiveness of ML for post-CA predictive modeling.

SWIFT: A deep learning approach to prediction of hypoxemic events in critically-Ill patients using SpO2 waveform prediction.

Hypoxemia is a significant driver of mortality and poor clinical outcomes in conditions such as brain injury and cardiac arrest in critically ill patients, including COVID-19 patients. Given the host of negative clinical outcomes attributed to hypoxemia, identifying patients likely to experience hypoxemia would offer valuable opportunities for early and thus more effective intervention. We present SWIFT (SpO2 Waveform ICU Forecasting Technique), a deep learning model that predicts blood oxygen saturation (SpO2) waveforms 5 and 30 minutes in the future using only prior SpO2 values as inputs. When tested on novel data, SWIFT predicts more than 80% and 60% of hypoxemic events in critically ill and COVID-19 patients, respectively. SWIFT also predicts SpO2 waveforms with average MSE below .0007. SWIFT predicts both occurrence and magnitude of potential hypoxemic events 30 minutes in the future, allowing it to be used to inform clinical interventions, patient triaging, and optimal resource allocation. SWIFT may be used in clinical decision support systems to inform the management of critically ill patients during the COVID-19 pandemic and beyond.

Predicting Flow Rate Escalation for Pediatric Patients on High Flow Nasal Cannula Using Machine Learning.

Background: High flow nasal cannula (HFNC) is commonly used as non-invasive respiratory support in critically ill children. There are limited data to inform consensus on optimal device parameters, determinants of successful patient response, and indications for escalation of support. Clinical scores, such as the respiratory rate-oxygenation (ROX) index, have been described as a means to predict HFNC non-response, but are limited to evaluating for escalations to invasive mechanical ventilation (MV). In the presence of apparent HFNC non-response, a clinician may choose to increase the HFNC flow rate to hypothetically prevent further respiratory deterioration, transition to an alternative non-invasive interface, or intubation for MV. To date, no models have been assessed to predict subsequent escalations of HFNC flow rates after HFNC initiation. Objective: To evaluate the abilities of tree-based machine learning algorithms to predict HFNC flow rate escalations. Methods: We performed a retrospective, cohort study assessing children admitted for acute respiratory failure under 24 months of age placed on HFNC in the Johns Hopkins Children's Center pediatric intensive care unit from January 2019 through January 2020. We excluded encounters with gaps in recorded clinical data, encounters in which MV treatment occurred prior to HFNC, and cases electively intubated in the operating room. The primary study outcome was discriminatory capacity of generated machine learning algorithms to predict HFNC flow rate escalations as compared to each other and ROX indices using area under the receiver operating characteristic (AUROC) analyses. In an exploratory fashion, model feature importance rankings were assessed by comparing Shapley values. Results: Our gradient boosting model with a time window of 8 h and lead time of 1 h before HFNC flow rate escalation achieved an AUROC with a 95% confidence interval of 0.810 ± 0.003. In comparison, the ROX index achieved an AUROC of 0.525 ± 0.000. Conclusion: In this single-center, retrospective cohort study assessing children under 24 months of age receiving HFNC for acute respiratory failure, tree-based machine learning models outperformed the ROX index in predicting subsequent flow rate escalations. Further validation studies are needed to ensure generalizability for bedside application.

Estimating ectopic beat probability with simplified statistical models that account for experimental uncertainty.

Ectopic beats (EBs) are cellular arrhythmias that can trigger lethal arrhythmias. Simulations using biophysically-detailed cardiac myocyte models can reveal how model parameters influence the probability of these cellular arrhythmias, however such analyses can pose a huge computational burden. Here, we develop a simplified approach in which logistic regression models (LRMs) are used to define a mapping between the parameters of complex cell models and the probability of EBs (P(EB)). As an example, in this study, we build an LRM for P(EB) as a function of the initial value of diastolic cytosolic Ca2+ concentration ([Ca2+]iini), the initial state of sarcoplasmic reticulum (SR) Ca2+ load ([Ca2+]SRini), and kinetic parameters of the inward rectifier K+ current (IK1) and ryanodine receptor (RyR). This approach, which we refer to as arrhythmia sensitivity analysis, allows for evaluation of the relationship between these arrhythmic event probabilities and their associated parameters. This LRM is also used to demonstrate how uncertainties in experimentally measured values determine the uncertainty in P(EB). In a study of the role of [Ca2+]SRini uncertainty, we show a special property of the uncertainty in P(EB), where with increasing [Ca2+]SRini uncertainty, P(EB) uncertainty first increases and then decreases. Lastly, we demonstrate that IK1 suppression, at the level that occurs in heart failure myocytes, increases P(EB).

Early Prediction of Multiple Organ Dysfunction in the Pediatric Intensive Care Unit.

Objective: The objective of the study is to build models for early prediction of risk for developing multiple organ dysfunction (MOD) in pediatric intensive care unit (PICU) patients. Design: The design of the study is a retrospective observational cohort study. Setting: The setting of the study is at a single academic PICU at the Johns Hopkins Hospital, Baltimore, MD. Patients: The patients included in the study were <18 years of age admitted to the PICU between July 2014 and October 2015. Measurements and main results: Organ dysfunction labels were generated every minute from preceding 24-h time windows using the International Pediatric Sepsis Consensus Conference (IPSCC) and Proulx et al. MOD criteria. Early MOD prediction models were built using four machine learning methods: random forest, XGBoost, GLMBoost, and Lasso-GLM. An optimal threshold learned from training data was used to detect high-risk alert events (HRAs). The early prediction models from all methods achieved an area under the receiver operating characteristics curve ≥0.91 for both IPSCC and Proulx criteria. The best performance in terms of maximum F1-score was achieved with random forest (sensitivity: 0.72, positive predictive value: 0.70, F1-score: 0.71) and XGBoost (sensitivity: 0.8, positive predictive value: 0.81, F1-score: 0.81) for IPSCC and Proulx criteria, respectively. The median early warning time was 22.7 h for random forest and 37 h for XGBoost models for IPSCC and Proulx criteria, respectively. Applying spectral clustering on risk-score trajectories over 24 h following early warning provided a high-risk group with ≥0.93 positive predictive value. Conclusions: Early predictions from risk-based patient monitoring could provide more than 22 h of lead time for MOD onset, with ≥0.93 positive predictive value for a high-risk group identified pre-MOD.

Prediction of Impending Septic Shock in Children With Sepsis.

OBJECTIVES: Sepsis and septic shock are leading causes of in-hospital mortality. Timely treatment is crucial in improving patient outcome, yet treatment delays remain common. Early prediction of those patients with sepsis who will progress to its most severe form, septic shock, can increase the actionable window for interventions. We aim to extend a time-evolving risk score, previously developed in adult patients, to predict pediatric sepsis patients who are likely to develop septic shock before its onset, and to determine whether or not these risk scores stratify into groups with distinct temporal evolution once this prediction is made. DESIGN: Retrospective cohort study. SETTING: Academic medical center from July 1, 2016, to December 11, 2020. PATIENTS: Six-thousand one-hundred sixty-one patients under 18 admitted to the Johns Hopkins Hospital PICU. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: We trained risk models to predict impending transition into septic shock and compute time-evolving risk scores representative of a patient's probability of developing septic shock. We obtain early prediction performance of 0.90 area under the receiver operating curve, 43% overall positive predictive value, patient-specific positive predictive value as high as 62%, and an 8.9-hour median early warning time using Sepsis-3 labels based on age-adjusted Sequential Organ Failure Assessment score. Using spectral clustering, we stratified pediatric sepsis patients into two clusters differing in septic shock prevalence, mortality, and proportion of patients adequately fluid resuscitated. CONCLUSIONS: We demonstrate the applicability of our methodology for early prediction and stratification for risk of septic shock in pediatric sepsis patients. Through analyses of risk score evolution over time, we corroborate our past finding of an abrupt transition preceding onset of septic shock in children and are able to stratify pediatric sepsis patients using their risk score trajectories into low and high-risk categories.

SWIFT: A Deep Learning Approach to Prediction of Hypoxemic Events in Critically-Ill Patients Using SpO 2 Waveform Prediction.

Hypoxemia is a significant driver of mortality and poor clinical outcomes in conditions such as brain injury and cardiac arrest in critically ill patients, including COVID-19 patients. Given the host of negative clinical outcomes attributed to hypoxemia, identifying patients likely to experience hypoxemia would offer valuable opportunities for early and thus more effective intervention. We present SWIFT (SpO 2 W aveform I CU F orecasting T echnique), a deep learning model that predicts blood oxygen saturation (SpO 2 ) waveforms 5 and 30 minutes in the future using only prior SpO 2 values as inputs. When tested on novel data, SWIFT predicts more than 80% and 60% of hypoxemic events in critically ill and COVID-19 patients, respectively. SWIFT also predicts SpO 2 waveforms with average MSE below .0007. SWIFT provides information on both occurrence and magnitude of potential hypoxemic events 30 minutes in advance, allowing it to be used to inform clinical interventions, patient triaging, and optimal resource allocation. SWIFT may be used in clinical decision support systems to inform the management of critically ill patients during the COVID-19 pandemic and beyond.

Novel Lineage-Tracing System to Identify Site-Specific Ectopic Bone Precursor Cells.

Heterotopic ossification (HO) is a form of pathological cell-fate change of mesenchymal stem/precursor cells (MSCs) that occurs following traumatic injury, limiting range of motion in extremities and causing pain. MSCs have been shown to differentiate to form bone; however, their lineage and aberrant processes after trauma are not well understood. Utilizing a well-established mouse HO model and inducible lineage-tracing mouse (Hoxa11-CreERT2;ROSA26-LSL-TdTomato), we found that Hoxa11-lineage cells represent HO progenitors specifically in the zeugopod. Bioinformatic single-cell transcriptomic and epigenomic analyses showed Hoxa11-lineage cells are regionally restricted mesenchymal cells that, after injury, gain the potential to undergo differentiation toward chondrocytes, osteoblasts, and adipocytes. This study identifies Hoxa11-lineage cells as zeugopod-specific ectopic bone progenitors and elucidates the fate specification and multipotency that mesenchymal cells acquire after injury. Furthermore, this highlights homeobox patterning genes as useful tools to trace region-specific progenitors and enable location-specific gene deletion.

Spectral clustering of risk score trajectories stratifies sepsis patients by clinical outcome and interventions received.

Sepsis is not a monolithic disease, but a loose collection of symptoms with diverse outcomes. Thus, stratification and subtyping of sepsis patients is of great importance. We examine the temporal evolution of patient state using our previously-published method for computing risk of transition from sepsis into septic shock. Risk trajectories diverge into four clusters following early prediction of septic shock, stratifying by outcome: the highest-risk and lowest-risk groups have a 76.5% and 10.4% prevalence of septic shock, and 43% and 18% mortality, respectively. These clusters differ also in treatments received and median time to shock onset. Analyses reveal the existence of a rapid (30-60 min) transition in risk at the time of threshold crossing. We hypothesize that this transition occurs as a result of the failure of compensatory biological systems to cope with infection, resulting in a bifurcation of low to high risk. Such a collapse, we believe, represents the true onset of septic shock. Thus, this rapid elevation in risk represents a potential new data-driven definition of septic shock.

Immobilization after injury alters extracellular matrix and stem cell fate.

Cells sense the extracellular environment and mechanical stimuli and translate these signals into intracellular responses through mechanotransduction, which alters cell maintenance, proliferation, and differentiation. Here we use a mouse model of trauma-induced heterotopic ossification (HO) to examine how cell-extrinsic forces impact mesenchymal progenitor cell (MPC) fate. After injury, single-cell (sc) RNA sequencing of the injury site reveals an early increase in MPC genes associated with pathways of cell adhesion and ECM-receptor interactions, and MPC trajectories to cartilage and bone. Immunostaining uncovers active mechanotransduction after injury with increased focal adhesion kinase signaling and nuclear translocation of transcriptional coactivator TAZ, inhibition of which mitigates HO. Similarly, joint immobilization decreases mechanotransductive signaling, and completely inhibits HO. Joint immobilization decreases collagen alignment and increases adipogenesis. Further, scRNA sequencing of the HO site after injury with or without immobilization identifies gene signatures in mobile MPCs correlating with osteogenesis, and signatures from immobile MPCs with adipogenesis. scATAC-seq in these same MPCs confirm that in mobile MPCs, chromatin regions around osteogenic genes are open, whereas in immobile MPCs, regions around adipogenic genes are open. Together these data suggest that joint immobilization after injury results in decreased ECM alignment, altered MPC mechanotransduction, and changes in genomic architecture favoring adipogenesis over osteogenesis, resulting in decreased formation of HO.