It was the best of times, it was the worst of times, it was the age of ‘AI’ and Tech. Tech-based solutions have made the best out of times of something as uncertain, worse, and unpredictable as COVID. With every passing day, innovations have started making it easier for humans to deal with this ‘devil’ virus which keeps on changing its forms thus wreaking havoc.
However, a glimmer of hope still shines as experts from Feinstein Institutes for Medical Research at Northwell have created an AI-based tool for assessing and predicting the respiratory failure risk in a patient within a bracket of 48 hours.
Moreover, according to the research published in the Journal of Medical Internet Research by Dr.Douglas Barnaby and Theodoros Zanos showed that the identification of the risk of respiratory failure allows patients to be prioritized for closer monitoring and critical care consultation. The study was based on the electronic health record data from a whopping 11,525 count of patients admitted to 13 Northwell hospitals in early 2020 marking the peak time of pandemic in New York. 8 percent that is 933 of these patients were placed on ventilators within the first 48 hours of admission.
The experts of Northwell used a combination of data including demographic information, labs, and vitals. Using such a combination, they fashioned three machine learning models to help monitor and score the admitted patients. Of these three, XCBoost or gradient boosted decision trees stood out as they gave an accurate prediction of an impressive 92 percent.
Since the results gave accurate predictions, it also helped to prioritize the treatment of vulnerable patients by introducing interventions at an early stage to reduce the number of deaths. The positive news is that it outperformed the Modified Early Warning Score which was also modeled to calculate the risk of respiratory failure within 48 hours. Moreover, Northwell has also decided to extend the access to this tool so that other Northwell hospitals can also improve their performance.
With the pandemic, technology and AI ushered into a new trend of evolution with novel predictive models revamping the healthcare and clinical decision-making operations. Just recently Israeli researchers calculated the hospitalization time for COVID-19 using an amalgam of demographic data, clinical information in addition to predicting the probability of in-hospital death.
AI-based interventions and advances have opened up new avenues of research for the experts which can utilize innovative algorithms for modeling of diseases, according to the Synthetic Health Data Challenge by ONC.
A Breath of Fresh Air
According to Zanos, respiratory failure is one of the most common causes of death among COVID-19 patients and so having informative, unbiased data is crucial to the process of decision-making especially in the emergency unit. He further added that AI and machine learning can be a potential game-changer as they will help identify patients whose health is deteriorating in a relatively short time.