External Validation and Comparison of a General Ward Deterioration Index Between Diversely Different Health Systems.

OBJECTIVES: Implementing a predictive analytic model in a new clinical environment is fraught with challenges. Dataset shifts such as differences in clinical practice, new data acquisition devices, or changes in the electronic health record (EHR) implementation mean that the input data seen by a model can differ significantly from the data it was trained on. Validating models at multiple institutions is therefore critical. Here, using retrospective data, we demonstrate how Predicting Intensive Care Transfers and other UnfoReseen Events (PICTURE), a deterioration index developed at a single academic medical center, generalizes to a second institution with significantly different patient population. DESIGN: PICTURE is a deterioration index designed for the general ward, which uses structured EHR data such as laboratory values and vital signs. SETTING: The general wards of two large hospitals, one an academic medical center and the other a community hospital. SUBJECTS: The model has previously been trained and validated on a cohort of 165,018 general ward encounters from a large academic medical center. Here, we apply this model to 11,083 encounters from a separate community hospital. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The hospitals were found to have significant differences in missingness rates (> 5% difference in 9/52 features), deterioration rate (4.5% vs 2.5%), and racial makeup (20% non-White vs 49% non-White). Despite these differences, PICTURE's performance was consistent (area under the receiver operating characteristic curve [AUROC], 0.870; 95% CI, 0.861-0.878), area under the precision-recall curve (AUPRC, 0.298; 95% CI, 0.275-0.320) at the first hospital; AUROC 0.875 (0.851-0.902), AUPRC 0.339 (0.281-0.398) at the second. AUPRC was standardized to a 2.5% event rate. PICTURE also outperformed both the Epic Deterioration Index and the National Early Warning Score at both institutions. CONCLUSIONS: Important differences were observed between the two institutions, including data availability and demographic makeup. PICTURE was able to identify general ward patients at risk of deterioration at both hospitals with consistent performance (AUROC and AUPRC) and compared favorably to existing metrics.

Trans-ocular brain impedance index for assessment of cerebral autoregulation in a porcine model of cerebral hemodynamic perturbation.

Cerebrovascular autoregulation (CA) is often impaired following traumatic brain injury. Established technologies and metrics used to assess CA are invasive and conducive for measurement, but not for continuous monitoring. We developed a trans-ocular brain impedance (TOBI) method that may provide non-invasive and continuous indices to assess CA. In this study, we monitored impedance metrics such as respiratory-induced impedance amplitude changes (dz) as well as a novel impedance index (DZx), which is a moving Pearson correlation between mean arterial pressure (MAP) and dz. Yorkshire swine were instrumented to continuously record ICP, MAP, and cerebral blood flow (CBF). TOBI was recorded by placement of standard ECG electrodes on closed eyelids and connected to a data acquisition system. MAP, ICP and CBF were manipulated utilizing an intravenous vasopressor challenge. TOBI indices (dz and DZx) were compared to the hemodynamic indicators as well as pressure reactivity index (PRx). During the vasopressor challenge, dz was highly correlated with ICP, CPP, and CBF (r = < - 0.49, p < 0.0001). ICP, CPP, and CBF had a mean percent increase (standard deviation) from baseline of 29(23.2)%, 70(25)%, and 37(72.6)% respectively while dz decreased by 31(15.6)%. Receiver operator curve test showed high predictive performance of DZx when compared to PRx with area under the curve above 0.86, with high sensitivity and specificity. Impedance indices appear to track changes in PRx and hemodynamics that affect cerebral autoregulation. TOBI may be a suitable less invasive surrogate to PRx and capable of tracking cerebral autoregulation.

Demonstrating the consequences of learning missingness patterns in early warning systems for preventative health care: A novel simulation and solution.

When using tree-based methods to develop predictive analytics and early warning systems for preventive healthcare, it is important to use an appropriate imputation method to prevent learning the missingness pattern. To demonstrate this, we developed a novel simulation that generated synthetic electronic health record data using a variational autoencoder with a custom loss function, which took into account the high missing rate of electronic health data. We showed that when tree-based methods learn missingness patterns (correlated with adverse events) in electronic health record data, this leads to decreased performance if the system is used in a new setting that has different missingness patterns. Performance is worst in this scenario when the missing rate between those with and without an adverse event is the greatest. We found that randomized and Bayesian regression imputation methods mitigate the issue of learning the missingness pattern for tree-based methods. We used this information to build a novel early warning system for predicting patient deterioration in general wards and telemetry units: PICTURE (Predicting Intensive Care Transfers and other UnfoReseen Events). To develop, tune, and test PICTURE, we used labs and vital signs from electronic health records of adult patients over four years (n = 133,089 encounters). We analyzed primary outcomes of unplanned intensive care unit transfer, emergency vasoactive medication administration, cardiac arrest, and death. We compared PICTURE with existing early warning systems and logistic regression at multiple levels of granularity. When analyzing PICTURE on the testing set using all observations within a hospital encounter (event rate = 3.4%), PICTURE had an area under the receiver operating characteristic curve (AUROC) of 0.83 and an adjusted (event rate = 4%) area under the precision-recall curve (AUPR) of 0.27, while the next best tested method-regularized logistic regression-had an AUROC of 0.80 and an adjusted AUPR of 0.22. To ensure system interpretability, we applied a state-of-the-art prediction explainer that provided a ranked list of features contributing most to the prediction. Though it is currently difficult to compare machine learning-based early warning systems, a rudimentary comparison with published scores demonstrated that PICTURE is on par with state-of-the-art machine learning systems. To facilitate more robust comparisons and development of early warning systems in the future, we have released our variational autoencoder's code and weights so researchers can (a) test their models on data similar to our institution and (b) make their own synthetic datasets.

Redox Potential Correlates with Changes in Metabolite Concentrations Attributable to Pathways Active in Oxidative Stress Response in Swine Traumatic Shock.

INTRODUCTION: Oxidation-reduction (redox) reactions, and the redox potential (RP) that must be maintained for proper cell function, lie at the heart of physiologic processes in critical illness. Imbalance in RP reflects systemic oxidative stress, and whole blood RP measures have been shown to correlate with oxygen debt level over time in swine traumatic shock. We hypothesize that RP measures reflect changing concentrations of metabolites involved in oxidative stress. To test this hypothesis, we compared blood and urine RP with concentrations of multiple metabolites in a swine traumatic shock model to identify meaningful RP-metabolite relationships. METHODS: Seven swine were subjected to traumatic shock. Mixed venous (MV) RP, urine RP, and concurrent MV and urine metabolite concentrations were assessed at baseline, max O 2 Debt (80 mL/kg), end resuscitation, and 2 h post-resuscitation. RP was measured at collection via open circuit potential using nanoporous gold electrodes with Ag/AgCl reference and a ParstatMC potentiostat. Metabolite concentrations were measured by quantitative 1 H-NMR spectroscopy. MV and urine RP were compared with time-matched metabolites across all swine. LASSO regression with leave-one-out cross validation was used to determine meaningful RP/metabolite relationships. Metabolites had to maintain magnitude and direction of coefficients across 6 or more swine to be considered as having a meaningful relationship. KEGG IDs of these metabolites were uploaded into Metscape for pathway identification and evaluation for physiologic function. RESULTS: Meaningful metabolite relationships (and mean coefficients across cross-validation folds) with MV RP included: choline (-6.27), ATP (-4.39), glycine (5.93), ADP (1.84), glucose (15.96), formate (-13.09), pyruvate (6.18), and taurine (-7.18). Relationships with urine RP were: betaine (4.81), urea (4.14), glycine (-2.97), taurine (10.32), 3-hydroxyisobutyrate (-7.67), N-phenylacetylglycine, PAG (-14.52), hippurate (12.89), and formate (-5.89). These meaningful metabolites were found to scavenge extracellular peroxide (pyruvate), inhibit ROS and activate cellular antioxidant defense (taurine), act as indicators of antioxidant mobilization against oxidative stress (glycine + PAG), and reflect renal hydroxyl radical trapping (hippurate), among other activities. CONCLUSIONS: Real-time RP measures demonstrate significant relationships with metabolites attributable to metabolic pathways involved in systemic responses to oxidative stress, as well as those involved in these processes. These data support RP measures as a feasible, biologically relevant marker of oxidative stress. As a direct measure of redox state, RP may be a useful biomarker and clinical tool in guiding diagnosis and therapy in states of increased oxidative stress and may offer value as a marker for organ injury in these states as well.

Predicting Intensive Care Transfers and Other Unforeseen Events: Analytic Model Validation Study and Comparison to Existing Methods.

BACKGROUND: COVID-19 has led to an unprecedented strain on health care facilities across the United States. Accurately identifying patients at an increased risk of deterioration may help hospitals manage their resources while improving the quality of patient care. Here, we present the results of an analytical model, Predicting Intensive Care Transfers and Other Unforeseen Events (PICTURE), to identify patients at high risk for imminent intensive care unit transfer, respiratory failure, or death, with the intention to improve the prediction of deterioration due to COVID-19. OBJECTIVE: This study aims to validate the PICTURE model's ability to predict unexpected deterioration in general ward and COVID-19 patients, and to compare its performance with the Epic Deterioration Index (EDI), an existing model that has recently been assessed for use in patients with COVID-19. METHODS: The PICTURE model was trained and validated on a cohort of hospitalized non-COVID-19 patients using electronic health record data from 2014 to 2018. It was then applied to two holdout test sets: non-COVID-19 patients from 2019 and patients testing positive for COVID-19 in 2020. PICTURE results were aligned to EDI and NEWS scores for head-to-head comparison via area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve. We compared the models' ability to predict an adverse event (defined as intensive care unit transfer, mechanical ventilation use, or death). Shapley values were used to provide explanations for PICTURE predictions. RESULTS: In non-COVID-19 general ward patients, PICTURE achieved an AUROC of 0.819 (95% CI 0.805-0.834) per observation, compared to the EDI's AUROC of 0.763 (95% CI 0.746-0.781; n=21,740; P<.001). In patients testing positive for COVID-19, PICTURE again outperformed the EDI with an AUROC of 0.849 (95% CI 0.820-0.878) compared to the EDI's AUROC of 0.803 (95% CI 0.772-0.838; n=607; P<.001). The most important variables influencing PICTURE predictions in the COVID-19 cohort were a rapid respiratory rate, a high level of oxygen support, low oxygen saturation, and impaired mental status (Glasgow Coma Scale). CONCLUSIONS: The PICTURE model is more accurate in predicting adverse patient outcomes for both general ward patients and COVID-19 positive patients in our cohorts compared to the EDI. The ability to consistently anticipate these events may be especially valuable when considering potential incipient waves of COVID-19 infections. The generalizability of the model will require testing in other health care systems for validation.

Dose optimization of early high-dose valproic acid for neuroprotection in a swine cardiac arrest model.

AIM: High-dose valproic acid (VPA) improves the survival and neurologic outcomes after asphyxial cardiac arrest (CA) in rats. We characterized the pharmacokinetics, pharmacodynamics, and safety of high-dose VPA in a swine CA model to advance clinical translation. METHODS: After 8 â€‹min of untreated ventricular fibrillation CA, 20 male Yorkshire swine were resuscitated until return of spontaneous circulation (ROSC). They were block randomized to receive placebo, 75 â€‹mg/kg, 150 â€‹mg/kg, or 300 â€‹mg/kg VPA as 90-min intravenous infusion (n â€‹= â€‹5/group) beginning at ROSC. Animals were monitored for 2 additional hours then euthanized. Experimental operators were blinded to treatments. RESULTS: The mean(SD) total CA duration was 14.8(1.2) minutes. 300 â€‹mg/kg VPA animals required more adrenaline to maintain mean arterial pressure ≥80 â€‹mmHg and had worse lactic acidosis. There was a strong linear correlation between plasma free VPA Cmax and brain total VPA (r2 â€‹= â€‹0.9494; p â€‹< â€‹0.0001). VPA induced dose-dependent increases in pan- and site-specific histone H3 and H4 acetylation in the brain. Plasma free VPA Cmax is a better predictor than peripheral blood mononuclear cell histone acetylation for brain H3 and H4 acetylation (r2 â€‹= â€‹0.7189 for H3K27ac, r2 â€‹= â€‹0.7189 for pan-H3ac, and r2 â€‹= â€‹0.7554 for pan-H4ac; p â€‹< â€‹0.0001). CONCLUSIONS: Up to 150 â€‹mg/kg VPA can be safely tolerated as 90-min intravenous infusion in a swine CA model. High-dose VPA induced dose-dependent increases in brain histone H3 and H4 acetylation, which can be predicted by plasma free VPA Cmax as the pharmacodynamics biomarker for VPA target engagement after CA.

Use of resuscitative balloon occlusion of the aorta in a swine model of prolonged cardiac arrest.

AIM: We examined the use of a Resuscitative Endovascular Balloon Occlusion of the Aorta (REBOA) catheter during cardiopulmonary resuscitation (CPR) after cardiac arrest (CA) to assess its effect on haemodynamics such as coronary perfusion pressure (CPP), common carotid artery blood flow (CCA-flow) and end-tidal CO2 (PetCO2) which are associated with increased return of spontaneous circulation (ROSC). METHODS: Six male swine were instrumented to measure CPP, CCA-Flow, and PetCO2. A 7Fr REBOA was advanced into zone-1 of the aorta through the femoral artery. Ventricular fibrillation was induced and untreated for 8 min. CPR (manual then mechanical) was initiated for 24 min. Continuous infusion of adrenaline (epinephrine) was started at minute-4 of CPR. The REBOA balloon was inflated at minute-16 for 3 min and then deflated/inflated every 3 min for 3 cycles. Animals were defibrillated up to 6 times after the final cycle. Animals achieving ROSC were monitored for 25 min. RESULTS: Data showed significant differences between balloon deflation and inflation periods for CPP, CCA-Flow, and PetCO2 (p < 0.0001) with an average difference (SD) of 13.7 (2.28) mmHg, 15.5 (14.12) mL min-1 and -4 (2.76) mmHg respectively. Three animals achieved ROSC and had significantly higher mean CPP (54 vs. 18 mmHg), CCA-Flow (262 vs. 135 mL min-1) and PetCO2 (16 vs. 8 mmHg) (p < 0.0001) throughout inflation periods than No-ROSC animals. Aortic histology did not reveal any significant changes produced by balloon inflation. CONCLUSION: REBOA significantly increased CPP and CCA-Flow in this model of prolonged CA. These increases may contribute to the ability to achieve ROSC.

Novel Noninvasive Method of Cerebrovascular Blood Volume Assessment Using Brain Bioimpedance.

Cerebrovascular autoregulation (CAR) is the ability of vessels to modulate their tone in response to changes in pressure. As an auto-protective mechanism, CAR is critical in preventing secondary brain injury post-trauma. Monitoring of changes in cerebral blood volume might be valuable in evaluating CAR and response to various therapies. In this study, we utilized an ocular-brain bioimpedance interface to assess real time changes in cerebral blood volume in response to a number of physiological challenges. We hypothesize that changes in brain bioimpedance (dz) would track changes in cerebral blood volume. Anesthetized animals were instrumented for monitoring of intracranial pressure (ICP), mean arterial blood pressure, cerebral perfusion pressure (CPP) and cerebral blood flow (CBF). Bioimpedance was monitored continuously through electrocardiographic electrodes placed over the eyelids. Interventions such as hyperventilation, vasopressor administration, creation of an epidural hematoma, and systemic hemorrhage were used to manipulate levels of ICP, CPP, and CBF. The dz correlated with changes in ICP, CPP, and CBF (r = -0.72 to -0.88, p < 0.0001). The receiver operating characteristic for dz at different thresholds of CPP and CBF showed high impedance performance with area under the curve between 0.80-1.00 (p < 0.003) and sensitivity and specificity varying between 83%-100% and 70%-100%, respectively. Our preliminary tests show that brain bioimpedance as measured through the ocular-brain interface tracks changes in CPP and CBF with high precision and may prove to be valuable in the future in assessing changes in cerebral blood volume and CAR.

Dynamic Limb Bioimpedance and Inferior Vena Cava Ultrasound in Patients Undergoing Hemodialysis.

Assessment of volume status in critically ill patients poses a challenge to clinicians. Measuring changes in the inferior vena cava (IVC) diameter using ultrasound is becoming a standard tool to assess volume status. Ultrasound requires physicians with significant training and specialized expensive equipment. It would be of significant value to be able to obtain this measurement continuously without physician presence. We hypothesize that dynamic changes in limb's bioimpedance in response to respiration could be used to predict changes in IVC. Forty-six subjects were tested a hemodialysis session. Impedance was measured via electrodes placed on the arm. Simultaneously, the IVC diameter was assessed by ultrasound. Subjects were asked to breathe spontaneously and perform respiratory maneuvers using a respiratory training device. Impedance (dz) was determined and compared with change in IVC diameter (dIVC; r = 0.76, p < 0.0001). There was significant relationship between dz and dIVC (p< 0.0001). Receiver-operator curves for dz at thresholds of dIVC (20% to70%) demonstrated high predictive power with areas under the curves (0.87-0.99, p < 0.0001). This evaluation suggests that real-time dynamic changes in limb impedance are capable of tracking a wide range of dynamic dIVC. This technique might be a suitable surrogate for monitoring real-time changes in dIVC to assess intravascular volume status.