No abstract

No abstract

Background During a pandemic, it is important for clinicians to stratify patients and decide who receives limited medical resources. Machine learning models have been proposed to accurately predict COVID-19 disease severity. Previous studies have typically tested only one machine learning algorithm and limited performance evaluation to area under the curve analysis. To obtain the best results possible, it may be important to test different machine learning algorithms to find the best prediction model. Objective In this study, we aimed to use automated machine learning (autoML) to train various machine learning algorithms. We selected the model that best predicted patients’ chances of surviving a SARS-CoV-2 infection. In addition, we identified which variables (ie, vital signs, biomarkers, comorbidities, etc) were the most influential in generating an accurate model. Methods Data were retrospectively collected from all patients who tested positive for COVID-19 at our institution between March 1 and July 3, 2020. We collected 48 variables from each patient within 36 hours before or after the index time (ie, real-time polymerase chain reaction positivity). Patients were followed for 30 days or until death. Patients’ data were used to build 20 machine learning models with various algorithms via autoML. The performance of machine learning models was measured by analyzing the area under the precision-recall curve (AUPCR). Subsequently, we established model interpretability via Shapley additive explanation and partial dependence plots to identify and rank variables that drove model predictions. Afterward, we conducted dimensionality reduction to extract the 10 most influential variables. AutoML models were retrained by only using these 10 variables, and the output models were evaluated against the model that used 48 variables. Results Data from 4313 patients were used to develop the models. The best model that was generated by using autoML and 48 variables was the stacked ensemble model (AUPRC=0.807). The two best independent models were the gradient boost machine and extreme gradient boost models, which had an AUPRC of 0.803 and 0.793, respectively. The deep learning model (AUPRC=0.73) was substantially inferior to the other models. The 10 most influential variables for generating high-performing models were systolic and diastolic blood pressure, age, pulse oximetry level, blood urea nitrogen level, lactate dehydrogenase level, D-dimer level, troponin level, respiratory rate, and Charlson comorbidity score. After the autoML models were retrained with these 10 variables, the stacked ensemble model still had the best performance (AUPRC=0.791). Conclusions We used autoML to develop high-performing models that predicted the survival of patients with COVID-19. In addition, we identified important variables that correlated with mortality. This is proof of concept that autoML is an efficient, effective, and informative method for generating machine learning–based clinical decision support tools.

Diffuse axonal injury (DAI) is a form of brain injury that is characterized by morphologic changes to axons throughout the brain and brainstem. Previous biomechanical studies have shown that primary axonal dysfunction, ranging from minor electrophysiologic disturbances to immediate axotomy, can be related to the rate and level of axonal deformation. Some existing rodent head injury models display varying degrees of axonal injury in the forebrain and brainstem, but the extent of axonal damage in the forebrain has been limited to the contused hemisphere. This study examined whether opening the dura mater over the contralateral hemisphere could direct mechanical deformation across the sagittal midline and produce levels of strain sufficient to cause a more widespread, bilateral forebrain axonal injury following cortical impact. Intracranial deformation patterns produced by this modified cortical impact technique were examined using surrogate skull-brain models. Modeling results revealed that the presence of a contralateral craniotomy significantly reduced surrogate tissue herniation through the foramen magnum, allowed surrogate tissue movement across the sagittal midline, and resulted in an appreciable increase in the shear strain in the contralateral cortex during the impact. To evaluate the injury pattern produced using this novel technique, rat brains were subjected to rigid indentor impact injury of their left somatosensory motor cortex (1.5 mm indentation, 4.5-4.9 m/sec velocity, and 22 msec dwell time) and examined after a 2-7 day survival period. Neurofilament immunohistochemistry revealed numerous axonal retraction balls in the subcortical white matter and overlying deep cortical layers in the right hemisphere beneath the contralateral craniotomy. Retraction balls were not seen at these positions in normals, sham controls, or animals that received cortical impact without contralateral craniotomy and dural opening. The results from these physical modeling and animal experiments indicate that opening of the contralateral dura mater permits translation of sufficient mechanical deformation across the midline to produce a more widespread pattern of axonal injury in the forebrain, a pattern that is distinct from those produced by existing fluid percussion and cortical impact techniques.

A 71-year-old woman underwent routine impantable cardioverter defibrillator implantation. On a predischarge check the next day, electrical signals and thresholds were excellent and similar to those at implant. The chest X-ray was unremarkable and showed good lead position at the right ventricular apex (RVA). At a routine one-month postimplant visit, electrograms were found to be miniscule, and pacing could not be achieved. Chest X-ray and fluoroscopy suggested perforation, then this was confirmed by computed tomography scan. The tip of the lead was estimated to be within 7 mm of the surface of the skin. The system was removed surgically, and the patient continued to do well. (PACE 2008; 31:7-9) lead perforation, implantable cardioverter/defibrillator, perforation Case ReportA 71-year-old woman with nonischemic dilated cardiomyopathy (NIDCM) fulfilled criteria for implantation of an implantable cardioverter defibrillator (ICD) according to criteria derived from the sudden cardiac death in heart failure trial (SCD-HeFT).On July 2, 2007, she underwent implantation of a St. Jude Medical (SJM) system (Sylmar, CA, USA) using a model V-196 Epic Plus VR ICD and a 7001 bipolar active fixation Riata lead, which was placed in the right ventricular apex (RVA). At implant, the ventricular electrogram was 12.8 mV, impedance was 681 Ohms, and the pacing threshold was 1.0 V at 0.5 msec. Ventricular fibrillation (VF) was induced twice and terminated by the first shock (20 J on both occasions). The next day she had a predischarge check and at this, the ventricular electrogram was 6.4 mV, impedance was 430 Ohms, pacing threshold was 0.5 V at 0.5 msec, and once again, VF was terminated by a single 20 J shock. Chest X-ray (Fig. 1) showed a conventional position in the RVA with a gentle nonredundant curve as the lead passed from the atrium into the ventricle to the apex.On August 10, 2007, the patient returned for a routine postimplantation check. The ventricular electrogram was tiny (0.7 mV), impedance was 600 Ohms, and it was not possible to pace the ventricle at the highest output settings of the device. roscopy (Fig. 2) suggested perforation, and this was confirmed by computed tomography of the chest (Fig. 3). The tip of the lead was estimated to lie within 7 mm of the surface of the skin. The patient remained asymptomatic.It was decided to remove the system in the operating room. A subxiphoid incision was made in the operating room, the tip of the lead mobilized, and a purse string suture placed on the right ventricle surrounding the emerging lead. The lead was then withdrawn through the vascular system. It was decided not to reimplant at that time, but to bring the patient back at a later date. She tolerated the procedure well, and had no evidence of effusion, either preoperatively (Fig. 3) at surgery or during the postoperative period.