Machine learning based clinical decision support system for early COVID-19 mortality prediction

Abstract

The coronavirus disease 2019 (COVID-19) is an acute respiratory disease that has been classified as a pandemic by World Health Organization (WHO). The sudden spike in the number of infections and high mortality rates have put immense pressure on the public medical systems. Hence, it is crucial to identify the key factors of mortality that yield high accuracy and consistency to optimize patient treatment strategy. This study uses machine learning methods to identify a powerful combination of five features that help predict mortality with 96% accuracy: neutrophils, lymphocytes, lactate dehydrogenase (LDH), high-sensitivity C-reactive protein (hs-CRP) and age. Various machine learning algorithms have been compared to achieve a consistent high accuracy across the days that span the disease. Robust testing with three cases confirm the strong predictive performance of the proposed model. The model predicts with an accuracy of 90% as early as 16 days before the outcome. This study would help accelerate the decision making process in healthcare systems for focused medical treatments early and accurately.

COVID-19 Risk Stratification and Mortality Prediction in Hospitalized Indian Patients

“…Studies have shown that use of steroids like Dexamethasone lowered COVID-19 fatalities significantly when administered to patients who require supplemental oxygen [28–31]. We observed a relation between the usage of these drugs and the survival of patients with extreme lymphocytes and neutrophils counts, which are associated with mortality(Figure 8) [14, 15, 24, 32, 33].…”

“…Karthikeyan et.al. [15] built a neural network that predicted mortality in Wuhan cohort with an accuracy of 96.5% ± 0.6% using only five parameters, age, lymphocyte (%), neutrophil (%), LDH and CRP. The same model when tested on the Indian cohort (current dataset) predicted mortality with an accuracy of only 58%.…”

“…We also explored the differences between Wuhan and New Delhi cohorts in key biomarkers across survivors and the dead [14, 15, 24]. We choose mortality as the indicator for comparison as it does not depend on subjective labelling.…”

“…For each patient in the dataset, there were multiple lab test results recorded on different days before the outcome. We have considered each individual recorded test result as a unique data point for training and testing as has been done before [14, 15]. Each sample has a dimensionality equal to the number of unique parameters measured across all lab tests considered for the analysis.…”

“…This provided an interpretable machine learning solution with at least 90% accuracy. Karthikeyan et al [15] analysed the same dataset through comparing various machine learning algorithms. XGBoost feature selection and neural network classification yielded the best performance with the important biomarkers selected as neutrophil (%), lymphocyte (%), LDH, CRP and age.…”