An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department.

During the coronavirus disease 2019 (COVID-19) pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3661 patients, achieves an area under the receiver operating characteristic curve (AUC) of 0.786 (95% CI: 0.745-0.830) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at New York University Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.

Increase in Ventricle Size and the Evolution of White Matter Changes on Serial Imaging in Critically Ill Patients with COVID-19.

BACKGROUND: Evolution of brain magnetic resonance imaging (MRI) findings in critically ill patients with coronavirus disease 2019 (COVID-19) is unknown. METHODS: We retrospectively reviewed 4530 critically ill patients with COVID-19 admitted to three tertiary care hospitals in New York City from March 1 to June 30, 2020 to identify patients who had more than one brain MRI. We reviewed the initial and final MRI for each patient to (1) measure the percent change in the bicaudate index and third ventricular diameter and (2) evaluate changes in the presence and severity of white matter changes. RESULTS: Twenty-one patients had two MRIs separated by a median of 22 [Interquartile range (IQR) 14-30] days. Ventricle size increased for 15 patients (71%) between scans [median bicaudate index 0.16 (IQR 0.126-0.181) initially and 0.167 (IQR 0.138-0.203) on final imaging (p < 0.001); median third ventricular diameter 6.9 mm (IQR 5.4-10.3) initially and 7.2 mm (IQR 6.4-10.8) on final imaging (p < 0.001)]. Every patient had white matter changes on the initial and final MRI; between images, they worsened for seven patients (33%) and improved for three (14%). CONCLUSIONS: On serial imaging of critically ill patients with COVID-19, ventricle size frequently increased over several weeks. White matter changes were often unchanged, but in some cases they worsened or improved, demonstrating there is likely a spectrum of pathophysiological processes responsible for these changes.

A validated, real-time prediction model for favorable outcomes in hospitalized COVID-19 patients.

The COVID-19 pandemic has challenged front-line clinical decision-making, leading to numerous published prognostic tools. However, few models have been prospectively validated and none report implementation in practice. Here, we use 3345 retrospective and 474 prospective hospitalizations to develop and validate a parsimonious model to identify patients with favorable outcomes within 96 h of a prediction, based on real-time lab values, vital signs, and oxygen support variables. In retrospective and prospective validation, the model achieves high average precision (88.6% 95% CI: [88.4-88.7] and 90.8% [90.8-90.8]) and discrimination (95.1% [95.1-95.2] and 86.8% [86.8-86.9]) respectively. We implemented and integrated the model into the EHR, achieving a positive predictive value of 93.3% with 41% sensitivity. Preliminary results suggest clinicians are adopting these scores into their clinical workflows.

Risk factors for intracerebral hemorrhage in patients with COVID-19.

Intracerebral hemorrhage (ICH) can be a devastating complication of coronavirus disease (COVID-19). We aimed to assess risk factors associated with ICH in this population. We performed a retrospective cohort study of adult patients admitted to NYU Langone Health system between March 1 and April 27 2020 with a positive nasopharyngeal swab polymerase chain reaction test result and presence of primary nontraumatic intracranial hemorrhage or hemorrhagic conversion of ischemic stroke on neuroimaging. Patients with intracranial procedures, malignancy, or vascular malformation were excluded. We used regression models to estimate odds ratios and 95% confidence intervals (OR, 95% CI) of the association between ICH and covariates. We also used regression models to determine association between ICH and mortality. Among 3824 patients admitted with COVID-19, 755 patients had neuroimaging and 416 patients were identified after exclusion criteria were applied. The mean (standard deviation) age was 69.3 (16.2), 35.8% were women, and 34.9% were on therapeutic anticoagulation. ICH occurred in 33 (7.9%) patients. Older age, non-Caucasian race, respiratory failure requiring mechanical ventilation, and therapeutic anticoagulation were associated with ICH on univariate analysis (p < 0.01 for each variable). In adjusted regression models, anticoagulation use was associated with a five-fold increased risk of ICH (OR 5.26, 95% CI 2.33-12.24, p < 0.001). ICH was associated with increased mortality (adjusted OR 2.6, 95 % CI 1.2-5.9). Anticoagulation use is associated with increased risk of ICH in patients with COVID-19. Further investigation is required to elucidate underlying mechanisms and prevention strategies in this population.

An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department

During the COVID-19 pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images, and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3,661 patients, achieves an AUC of 0.786 (95% CI: 0.742-0.827) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions, and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at NYU Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.

Cerebral Microbleeds and Leukoencephalopathy in Critically Ill Patients With COVID-19.

BACKGROUND AND PURPOSE: We conducted this study to investigate the prevalence and distribution of cerebral microbleeds and leukoencephalopathy in hospitalized patients with coronavirus disease 2019 (COVID-19) and correlate with clinical, laboratory, and functional outcomes. METHODS: We performed a retrospective chart review of 4131 COVID-19 positive adult patients who were admitted to 3 tertiary care hospitals of an academic medical center at the epicenter of the COVID-19 pandemic in New York City from March 1, 2020, to May 10, 2020, to identify patients who had magnetic resonance imaging (MRI) of the brain. We evaluated the MRIs in detail, and identified a subset of patients with leukoencephalopathy and/or cerebral microbleeds. We compared clinical, laboratory, and functional outcomes for these patients to patients who had a brain MRI that did not show these findings. RESULTS: Of 115 patients who had an MRI of the brain performed, 35 (30.4%) patients had leukoencephalopathy and/or cerebral microbleeds. Patients with leukoencephalopathy and/or cerebral microbleeds had neuroimaging performed later during the hospitalization course (27 versus 10.6 days; P<0.001), were clinically sicker at the time of brain MRI (median GCS 6 versus 14; P<0.001), and had higher peak D-dimer levels (8018±6677 versus 3183±3482; P<0.001), lower nadir platelet count (116.9±62.2 versus 158.3±76.2; P=0.03), higher peak international normalized ratio (2.2 versus 1.57; P<0.001) values when compared with patients who had a brain MRI that did not show these findings. They required longer ventilator support (34.6 versus 9.1 days; P<0.001) and were more likely to have moderate and severe acute respiratory distress syndrome score (88.6% versus 23.8%, P<0.001). These patients had longer hospitalizations (42.1 versus 20.9 days; P<0.001), overall worse functional status on discharge (mRS 5 versus 4; P=0.001), and higher mortality (20% versus 9%; P=0.144). CONCLUSIONS: The presence of leukoencephalopathy and/or cerebral microbleeds is associated with a critical illness, increased mortality, and worse functional outcome in patients with COVID-19.

Hemorrhagic stroke and anticoagulation in COVID-19.

BACKGROUND AND PURPOSE: Patients with the Coronavirus Disease of 2019 (COVID-19) are at increased risk for thrombotic events and mortality. Various anticoagulation regimens are now being considered for these patients. Anticoagulation is known to increase the risk for adverse bleeding events, of which intracranial hemorrhage (ICH) is one of the most feared. We present a retrospective study of 33 patients positive for COVID-19 with neuroimaging-documented ICH and examine anticoagulation use in this population. METHODS: Patients over the age of 18 with confirmed COVID-19 and radiographic evidence of ICH were included in this study. Evidence of hemorrhage was confirmed and categorized by a fellowship trained neuroradiologist. Electronic health records were analyzed for patient information including demographic data, medical history, hospital course, laboratory values, and medications. RESULTS: We identified 33 COVID-19 positive patients with ICH, mean age 61.6 years (range 37-83 years), 21.2% of whom were female. Parenchymal hemorrhages with mass effect and herniation occurred in 5 (15.2%) patients, with a 100% mortality rate. Of the remaining 28 patients with ICH, 7 (25%) had punctate hemorrhages, 17 (60.7%) had small- moderate size hemorrhages, and 4 (14.3%) had a large single site of hemorrhage without evidence of herniation. Almost all patients received either therapeutic dose anticoagulation (in 22 [66.7%] patients) or prophylactic dose (in 3 [9.1] patients) prior to ICH discovery. CONCLUSIONS: Anticoagulation therapy may be considered in patients with COVID-19 though the risk of ICH should be taken into account when developing a treatment regimen.

COVID-19 related neuroimaging findings: A signal of thromboembolic complications and a strong prognostic marker of poor patient outcome.

OBJECTIVE: To investigate the incidence and spectrum of neuroimaging findings and their prognostic role in hospitalized COVID-19 patients in New York City. METHODS: This is a retrospective cohort study of 3218 COVID-19 confirmed patients admitted to a major healthcare system (three hospitals) in New York City between March 1, 2020 and April 13, 2020. Clinical data were extracted from electronic medical records, and particularly data of all neurological symptoms were extracted from the imaging reports. Four neuroradiologists evaluated all neuroimaging studies for acute neuroimaging findings related to COVID-19. RESULTS: 14.1% of admitted COVID-19 patients had neuroimaging and this accounted for only 5.5% of the total imaging studies. Acute stroke was the most common finding on neuro-imaging, seen in 92.5% of patients with positive neuro-imaging studies, and present in 1.1% of hospitalized COVID-19 patients. Patients with acute large ischemic and hemorrhagic stroke had much higher mortality risk adjusted for age, BMI and hypertension compared to those COVID-19 patients without neuroimaging. (Odds Ratio 6.02 by LR; Hazard Ratio 2.28 by CRR). CONCLUSIONS: Our study demonstrates acute stroke is the most common neuroimaging finding among hospitalized COVID-19 patients. Detection of an acute stroke is a strong prognostic marker of poor outcome. Our study also highlights the fact there is limited use of neuroimaging in these patients due to multiple logistical constraints.