Publications by authors named "Jessica K De Freitas"
Article Synopsis
- - The study explores the effectiveness of deep learning models utilizing electrocardiogram (ECG) data to improve the specificity of screening for pulmonary embolism (PE), addressing the issue of overusing computed tomography pulmonary angiograms (CTPAs).
- - Researchers built a cohort of over 21,000 patients and developed three predictive models: one based on ECG data, one on electronic health records (EHR), and a Fusion model combining both, finding the Fusion model significantly outperformed the others in PE detection accuracy.
- - The findings suggest that integrating ECG waveforms with clinical data can enhance the specificity and overall performance in detecting PE, offering a potential improvement over traditional clinical risk scoring methods.
Article Synopsis
- Deep learning models in healthcare need large, balanced datasets to work effectively, but COVID-19 data is often imbalanced, presenting a challenge for model training.! -
- Traditional cross-entropy loss (CEL) can struggle with imbalanced data, but the study shows that using contrastive loss (CL) enhances the performance of CEL, particularly with COVID-19 electronic health records.! -
- The research demonstrates that CL models consistently perform better than CEL models in predicting patient outcomes like mortality and ICU transfers, achieving notable improvements in precision and recall metrics.!
Background Despite advances in cardiovascular disease and risk factor management, mortality from ischemic heart failure (HF) in patients with coronary artery disease (CAD) remains high. Given the partial role of genetics in HF and lack of reliable risk stratification tools, we developed and validated a polygenic risk score for HF in patients with CAD, which we term HF-PRS. Methods and Results Using summary statistics from a recent genome-wide association study for HF, we developed candidate PRSs in the Mount Sinai Bio CAD patient cohort (N=6274) by using the pruning and thresholding method and LDPred.
View Article and Find Full Text PDFRobust phenotyping of patients from electronic health records (EHRs) at scale is a challenge in clinical informatics. Here, we introduce Phe2vec, an automated framework for disease phenotyping from EHRs based on unsupervised learning and assess its effectiveness against standard rule-based algorithms from Phenotype KnowledgeBase (PheKB). Phe2vec is based on pre-computing embeddings of medical concepts and patients' clinical history.
View Article and Find Full Text PDFArticle Synopsis
- * The proposed solution introduces a heterogeneous graph model (HGM) that incorporates relational learning to better predict mortality in COVID-19 ICU patients by utilizing large EHR datasets from multiple hospitals.
- * Experimental results indicate that the HGM model, using a unique Skip-Gram relational learning strategy, significantly outperforms traditional models in accuracy and recall, achieving higher area under the receiver operating characteristic curve (auROC) across different prediction time frames.
Background: Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and the associated Coronavirus Disease 2019 (COVID-19) is a public health emergency. Acute kidney injury (AKI) is a common complication in hospitalized patients with COVID-19 although mechanisms underlying AKI are yet unclear. There may be a direct effect of SARS-CoV-2 virus on the kidney; however, there is currently no data linking SARS-CoV-2 viral load (VL) to AKI.
View Article and Find Full Text PDFArticle Synopsis
- Deep learning has become crucial in analyzing large healthcare datasets for disease classification, predictions, and decision-making in the past decade.
- Public ECG datasets have been around since the 1980s, mainly for specific cardiology issues, while private institutions now offer significantly larger databases for deep learning applications.
- This review aims to educate clinicians on deep learning basics, its current uses in ECG analysis, as well as its limitations and potential future developments.
Article Synopsis
- Machine Learning models need large and balanced datasets to effectively analyze healthcare data, but COVID-19-related data is often unbalanced, especially within electronic health records (EHR).
- Traditional methods like cross-entropy loss struggle with classification accuracy under these conditions, leading researchers to explore contrastive loss as a potential solution.
- This study, using EHR data from five hospitals, demonstrates that contrastive loss significantly improves model performance for predicting COVID-19 patient outcomes compared to cross-entropy loss, particularly when dealing with severe class imbalance.
Article Synopsis
- Machine learning models require large datasets, often limited by data silos across healthcare institutions, particularly in COVID-19 research focused on single hospitals.
- The study utilized federated learning to predict 7-day mortality in hospitalized COVID-19 patients, using data from five hospitals within the Mount Sinai Health System without aggregating sensitive patient data.
- Results showed that the LASSO model performed better at three hospitals and the multilayer perceptron (MLP) model outperformed at all five, indicating that federated learning can create effective predictive models while protecting patient privacy.
Article Synopsis
- The study aimed to analyze clinical characteristics and outcomes of hospitalized COVID-19 patients, comparing those who died in the hospital to those who were discharged alive.
- Data was collected from five hospitals in the Mount Sinai Health System for patients confirmed with COVID-19 between February and April 2020, focusing on demographics, clinical features, and mortality rates.
- Results showed that nearly half of the 2199 hospitalized patients were discharged, with a 29% overall mortality rate, higher rates of pre-existing conditions and lower lymphocyte percentages observed in patients who died compared to those who recovered.
Objective: To assess association of clinical features on COVID-19 patient outcomes.
Design: Retrospective observational study using electronic medical record data.
Setting: Five member hospitals from the Mount Sinai Health System in New York City (NYC).
Background: N-of-1 trials are single patient, multiple crossover, and comparative effectiveness experiments. Despite their rating as "level 1" evidence, they are not routinely used in clinical medicine to evaluate the effectiveness of treatments.
Objective: We explored the potential for implementing a mobile app-based n-of-1 trial platform for collaborative use by clinicians and patients to support data-driven decisions around the treatment of insomnia.
Article Synopsis
- The COVID-19 pandemic has led to significant illness and mortality worldwide, highlighting the need for better resource allocation and risk identification for patients.
- This study aimed to analyze electronic health records from COVID-19 patients in the Mount Sinai Health System to develop machine learning models predicting hospital outcomes based on patient characteristics at admission.
- Using the XGBoost algorithm, the study found strong predictive performance for in-hospital mortality and critical events, with high accuracy scores across various time frames and effective model validation across multiple hospitals.
Machine learning (ML) models require large datasets which may be siloed across different healthcare institutions. Using federated learning, a ML technique that avoids locally aggregating raw clinical data across multiple institutions, we predict mortality within seven days in hospitalized COVID-19 patients. Patient data was collected from Electronic Health Records (EHRs) from five hospitals within the Mount Sinai Health System (MSHS).
View Article and Find Full Text PDFBackground: Data on patients with coronavirus disease 2019 (COVID-19) who return to hospital after discharge are scarce. Characterization of these patients may inform post-hospitalization care.
Objective: To describe clinical characteristics of patients with COVID-19 who returned to the emergency department (ED) or required readmission within 14 days of discharge.
Article Synopsis
- - Understanding how SARS-CoV-2 interacts with the host helps develop better treatments and public health responses, particularly targeting immune pathways linked to complement and coagulation systems.
- - A study found that individuals with a history of macular degeneration or coagulation disorders are at higher risk for severe COVID-19 outcomes, regardless of age, sex, or smoking history.
- - Genetic analysis identified specific variants linked to complement and coagulation functions, suggesting that these factors influence COVID-19 severity and highlighting the need for comprehensive research methods to assess immunity and disease susceptibility.
Background: There are limited data regarding the clinical impact of coronavirus disease 2019 (COVID-19) on people living with human immunodeficiency virus (PLWH). In this study, we compared outcomes for PLWH with COVID-19 to a matched comparison group.
Methods: We identified 88 PLWH hospitalized with laboratory-confirmed COVID-19 in our hospital system in New York City between 12 March and 23 April 2020.
Article Synopsis
- The study focused on the degree of heart damage, indicated by troponin levels, in over 2,700 hospitalized COVID-19 patients at Mount Sinai Health System from late February to early April 2020.
- Results showed that 36% of these patients had elevated troponin levels, with higher rates of heart-related conditions like coronary artery disease linked to these elevations.
- The findings revealed that even minor troponin increases significantly correlated with higher mortality rates, especially in patients with existing cardiovascular diseases.
Article Synopsis
- Study analyzed myocardial injury in hospitalized COVID-19 patients at Mount Sinai Health System in NYC, focusing on troponin levels to assess heart damage.
- 18.5% of the 2,736 patients died during hospitalization, with higher troponin levels correlating to increased mortality risk.
- Results suggest that myocardial injury is common in COVID-19 patients, especially those with pre-existing cardiovascular diseases, and indicates non-ischemic heart damage related to the virus.
Background: The coronavirus 2019 (Covid-19) pandemic is a global public health crisis, with over 1.6 million cases and 95,000 deaths worldwide. Data are needed regarding the clinical course of hospitalized patients, particularly in the United States.
View Article and Find Full Text PDFArticle Synopsis
- - Understanding how SARS-CoV-2 interacts with the body can help develop better therapies and public health strategies, focusing on immune pathways related to complement and coagulation systems.
- - A study of over 11,000 patients found that pre-existing conditions related to these systems, like macular degeneration and coagulation disorders, increase risks of severe illness and death from COVID-19, independent of other factors like age or smoking.
- - Genetic analysis revealed specific genetic markers linked to immune response that could help predict COVID-19 outcomes, illustrating the importance of combining various research methods to understand disease susceptibility.
Machine Learning in Cardiology-Ensuring Clinical Impact Lives Up to the Hype.
J Cardiovasc Pharmacol Ther
September 2020
Despite substantial advances in the study, treatment, and prevention of cardiovascular disease, numerous challenges relating to optimally screening, diagnosing, and managing patients remain. Simultaneous improvements in computing power, data storage, and data analytics have led to the development of new techniques to address these challenges. One powerful tool to this end is machine learning (ML), which aims to algorithmically identify and represent structure within data.
View Article and Find Full Text PDFSleep quality has been directly linked to cognitive function, quality of life, and a variety of serious diseases across many clinical domains. Standard methods for assessing sleep involve overnight studies in hospital settings, which are uncomfortable, expensive, not representative of real sleep, and difficult to conduct on a large scale. Recently, numerous commercial digital devices have been developed that record physiological data, such as movement, heart rate, and respiratory rate, which can act as a proxy for sleep quality in lieu of standard electroencephalogram recording equipment.
View Article and Find Full Text PDFAnonymized electronic health records (EHR) are often used for biomedical research. One persistent concern with this type of research is the risk for re-identification of patients from their purportedly anonymized data. Here, we use the EHR of 731,850 de-identified patients to demonstrate that the average patient is unique from all others 98.
View Article and Find Full Text PDFPathway choice within DNA double-strand break (DSB) repair is a tightly regulated process to maintain genome integrity. RECQL4, deficient in Rothmund-Thomson Syndrome, promotes the two major DSB repair pathways, non-homologous end joining (NHEJ) and homologous recombination (HR). Here we report that RECQL4 promotes and coordinates NHEJ and HR in different cell cycle phases.
View Article and Find Full Text PDF