Machine Learning

Covid-19 has been claimed to have bad after-effects on recovered patients. Similarly, people with other health conditions also find it difficult to survive the virus infection. There have been reports of patients dying of Covid complications. Acute Kidney Injury is a commonly found complication among Covid-19 patients and it is treated with dialysis initiation. Recently a group of researchers from the Icahn School of Medicine at Mount Sinai Health System developed a machine learning model to alleviate this issue. The paper is titled, “ Predictive Approaches for Acute Dialysis Requirement and Death in COVID-19”, and was published in the Clinical Journal of the American Society of Nephrology. 

Acute Kidney Injury in hospitalized Covid-19 patients has been associated with increased mortality and morbidity risks. If detected early, they would need to get dialysis treatment to improve the health condition. The Mount Sinai research team developed several machine learning models to predict treatment with dialysis or death happening at different times after they are hospitalized. They chose five different hospitals from the Mount Sinai Health System and collected the data of adult patients hospitalized there with Covid-19 within the time frame of March 10 and December 26 last year. The data collected included their demographics, vital signs and lab results within 12 hours of hospitalization, and comorbidities. The time horizons used in which they recorded the death of patients happened were days 1, 3, 5, and 7 of their hospital stay. According to the paper, they created many models for testing and they were eXtreme Gradient Boosting (XGBoost) with and without imputation, LASSO, Logistic Regression, LASSO, and Random Forest. These models performed both internal and external validations. Among these machine learning models, the paper reveals that XGBoost without imputation had the highest area under the receiver curve on internal validation, the highest area under the precision-recall curve, and the highest test parameters on external validations. XGBoost outperformed all other machine learning models as the paper states. 

While speaking to HospiMedica, Dr. Akhil Vaid, MD, postdoctoral fellow at the Icahn School of medicine at Mount Sinai and one of the authors of the paper said, “For COVID-19 inpatients, this means being able to more easily identify incoming at-risk patients, while pinpointing the underlying factors that are making them better or worse. The underlying algorithm, XGBoost, excels in accuracy, speed, and other under-the-hood features that allow for easier deployment and understanding of model predictions.”

This model developed by the Mount Sina group of researchers needs external validation before it gets into the field. However, this innovation yet again proves that disruptive technology can play a pivotal role in innovating the healthcare system. With similar machine learning models, it will be easier to predict the comorbidities and complications in Covid-19 patients and treat them before it leads to fatality. Covid-19 patients are at a high risk of infections and organ dysfunctions as proved by many reports. Early recognition of such conditions can enable the healthcare systems to alleviate adverse effects on the patients and help them continuously monitor and provide special care if necessary.