Publications

This study examined factors associated with immune responses to SARS-CoV-2 vaccination in 463 patients with autoimmune diseases, who are at higher risk for severe infection due to their underlying disease and immunosuppressive treatments. Medication use is a major determinant of low antibody response in patients with autoimmune diseases, with most patients having inadequate humoral response to both initial and booster vaccinations when taking B cell-depleting drugs. Treatments with highest risk for low anti-spike IgG response included B cell depletion within the last year, fingolimod, and combination treatment with mycophenolate mofetil and belimumab. Low IgG responses in B cell-depleted patients with multiple sclerosis were associated with higher CD8 T cell responses. The mRNA-1273 (Moderna) vaccine was the most effective vaccine in the autoimmune population. There was minimal induction of disease flares or autoantibodies by vaccination and no significant effect of preexisting anti-type I IFN antibodies on vaccine response or breakthrough infections.

This study analyzed the dynamic clinical phenotypes of more than 35,000 COVID-19 patients admitted to New York hospitals over a two-year period (March 2020 to May 2022), encapsulating the identity of all major COVID-19 waves and three dominant virus variants (alpha, delta, omicron) into four distinct patient clusters demonstrating unique demographics, treatment profiles, and mortality outcomes. The temporal progression of these phenotypes throughout the COVID-19 pandemic demonstrated increased variability across the waves of three dominant viral variants. Four distinct clinical phenotypes remained robust in multi-site validation and were associated with different mortality rates. Although the lung phenotype with high inflammation was most prevalent at admission, the lung phenotype with low inflammation consistently prevailed thereafter. Most patients transitioned to other phenotypes as time progressed, highlighting the dynamic nature of disease progression during hospitalization.

Objectives: We aimed to correlate lung disease burden on presentation chest radiographs (CXR), quantified at the time of study interpretation, with clinical presentation in patients hospitalized with coronavirus disease 2019 (COVID-19). Material and methods: This retrospective cross-sectional study included 5833 consecutive adult patients, aged 18 and older, hospitalized with a diagnosis of COVID-19. Lung disease burden was quantified in real-time by 118 radiologists on 5833 CXRs at the time of exam interpretation with each lung annotated by the degree of lung opacity as clear (0%), mild (1-33%), moderate (34-66%), or severe (67-100%). Results: COVID-19 lung disease burden quantified in real-time on presentation CXR was characterized by demographics, comorbidities, emergency severity index, Charlson Comorbidity Index, vital signs, and lab results. An absence of opacities in COVID-19 may be associated with poor oral intake and a prerenal state as evidenced by the association of clear CXRs with a low eGFR, hypernatremia, and hypoglycemia.

Background: Clinical prediction models (CPMs) for COVID-19 may support clinical decision making. Objective: To understand attitudes toward using COVID-19 CPMs among health care providers, survivors, and surrogates in the United States and Netherlands. Methods: Qualitative study using online focus groups and interviews conducted between January 2021 and July 2021 in the United States (4 focus groups) and May to July 2021 in the Netherlands (3 focus groups and 4 interviews). Results: Many providers had reservations about CPM validity and patient-level outcome interpretation. However, survivors and surrogates indicated they would have found this information useful for decision making. Providers perceived CPMs as most useful for resource allocation, triage, research, and educational purposes. Conclusions: There is a disconnect between CPM development and clinical implementation, highlighting the need for better communication and integration of provider and patient perspectives.

Prognostic models for COVID-19 developed during the first pandemic wave were evaluated for transportability to subsequent waves and geographic settings. The study included patients presenting to the emergency department with suspected COVID-19 admitted to 12 hospitals in the New York City area and 4 large Dutch hospitals. Second-wave patients (September-December 2020) were used to evaluate models developed on first-wave patients (March-August 2020). Two prognostic models were evaluated: The Northwell COVID-19 Survival (NOCOS) model developed on NYC data and the COVID Outcome Prediction in the Emergency Department (COPE) model developed on Dutch data. Frequent updating of prognostic models is likely to be required for transportability over time and space during a dynamic pandemic.

Objective: To develop and validate self-monitoring, auto-updating prognostic models of survival for hospitalized COVID-19 patients. Methods: We analyzed demographic, laboratory, and clinical data from 34,912 hospitalized COVID-19 patients (March 2020 to May 2022) using a 2,000-patient sliding window incremented at 500-patient intervals to detect calibration drift. Results: Calibration performance drift was immediately detected with only minor fluctuations in discrimination. Dynamically updated models significantly improved overall calibration compared to static models across different waves, variants, race, and sex. Net-benefit analyses showed positive benefits. Conclusions: This is the first study to perform dynamic updating of COVID-19 prognostic survival models to correct for calibration drift. The methodology can be extended to other clinical prognostic models, improving accuracy and clinical utility.

Predicting early respiratory failure due to COVID-19 can help triage patients to higher levels of care, allocate scarce resources, and reduce morbidity and mortality. This study derived a machine learning model that predicts respiratory failure within 48 hours of admission based on data from the emergency department. Data were collected from patients with COVID-19 admitted to Northwell Health acute care hospitals between March 1 and May 11, 2020. Of 11,525 patients, 933 (8.1%) were placed on invasive mechanical ventilation within 48 hours of admission. Three predictive models were trained and validated using cross-hospital validation. The XGBoost model had the highest mean accuracy (0.919; area under the curve=0.77), outperforming the other two models as well as the Modified Early Warning Score. Important predictor variables included the type of oxygen delivery used in the emergency department, patient age, Emergency Severity Index level, respiratory rate, serum lactate, and demographic characteristics.

External validation of a previously published interpretable mortality prediction model for COVID-19 patients was conducted. The study demonstrated that the model does not perform as a triage tool based on the internal validation dataset provided by the original authors. The decision algorithm was not portable to the external validation dataset, both with unmodified and optimized parameters. Specifically, the precision was 0.48 for predicting mortality, meaning that over half of the patients that the model predicted would die actually survived. The accuracy was 0.88 and the F1 score was 0.41. These results emphasized the importance of externally validating models before their widespread adoption in actual clinical practice.

The number of cases from the coronavirus disease 2019 (COVID-19) global pandemic has overwhelmed existing medical facilities and forced clinicians, patients, and families to make pivotal decisions with limited time and information. While machine learning (ML) methods have been previously used to augment clinical decisions, there is now a demand for "Emergency ML." Throughout the patient care pathway, there are opportunities for ML-supported decisions based on collected vitals, laboratory results, medication orders, and comorbidities. This perspective highlights the utility of evidence-based prediction tools in a number of clinical settings, and how similar models can be deployed during the COVID-19 pandemic to guide hospital frontlines and healthcare administrators to make informed decisions about patient care and managing hospital volume.

A total of 5700 patients with a median age of 63 years were included in this case series, with 39.7% female. The most common comorbidities were hypertension (3026; 56.6%), obesity (1737; 41.7%), and diabetes (1808; 33.8%). At triage, 30.7% of patients were febrile, 17.3% had a respiratory rate greater than 24 breaths/min, and 27.8% received supplemental oxygen. During hospitalization, 373 patients (14.2%) were treated in intensive care unit, and 320 (12.2%) received invasive mechanical ventilation.