How Mount Sinai, NYU predictive models performed during early pandemic

Recently published results on the performance of two predictive analytics models developed during New York City's spring COVID-19 wave show the tools have potential to help address the winter's anticipated wave of cases, but still need fine-tuning to estimate a patient's complete trajectory.

Researchers at New York City-based New York University developed a machine learning model to help physicians prioritize care for some COVID-19 patients and form discharge plans for others. It analyzed the medical records of thousands of New York COVID-19 patients, using each patient's vital signs, oxygen requirements and recent laboratory results to determine if they would have good or bad outcomes in the next four days.

Results published Oct. 6 in NPJ Digital Medicine reveal that the tool could identify whether a hospitalized COVID-19 patient would have a favorable outcome with 90 percent precision. Since it began testing in May, the tool helped estimate COVID-19 patient outcomes more than a half million times. 

However, the tool only identified 41 percent of all the patients who exhibited a good outcome within the four-day window, meaning hospital resources could have been prioritized in a more optimal way.

Another team of New York City researchers, this one based out of Mount Sinai Health System, designed a COVID-19 mortality predictive model to accurately and cost-effectively aid clinical staff in assessing COVID-19 patients' risk of death. Using what the research team called "the largest clinical dataset to date," they analyzed data from 5,051 Mount Sinai COVID-19 patients by deploying machine learning algorithms that focused on three clinical features: age, minimum oxygen saturation over the span of the medical encounter and type of encounter (inpatient, outpatient or telehealth). 

The study's results, published in the October 2020 issue of The Lancet, show that the predictive model produced a vital sign that can be easily integrated into clinical staff's workflows, allowing them to continually assess COVID-19 patients' needs. Age and blood oxygen level were the most telling factors, the researchers reported.

More articles on data analytics:
7 recently launched contact tracing efforts
Illinois kept details on nearly 2,600 COVID-19 outbreaks confidential
Florida lawmaker warns state's COVID-19 death rates are flawed

7 recently launched contact tracing efforts

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.


Featured Webinars

Featured Whitepapers