Machine Learning Models to Predict Severity and Mortality of COVID-19 Using Neurological Symptoms
Mona Salehi1, Amir Garakani2, Man Amanat3
1Johns Hopkins Medical School, 2Yale Medicine, 3Kennedy Krieger Institute
Objective:

This study aimed to assess the correlation of neurological symptoms with the disease severity and mortality. Furthermore, different models were proposed to predict the severity and mortality using COVID-19 symptoms including neurological manifestations.

Background:
Coronavirus disease 2019 (COVID-19) was first described after many cases of pneumonia with unknown cause were reported in Wuhan, China in December, 2019. COVID-19 can present multiple non-respiratory symptoms and it was found to be associated with increased risk of different neurological manifestations. 
Design/Methods:

This is a prospective cohort study conducted among hospitalized patients diagnosed with COVID-19 from April 7, 2020 to January 21, 2020. All manifestations were recorded and patients were examined by a neurologist to identify neurological symptoms. The severity was classified based on American Thoracic Society recommendations for community-acquired pneumonia. Data were analyzed and models were designed using IPython 7.22.0 in Jupyter notebook 6.3.0 from Anaconda.

Results:

There were 921 eligible participants (552 males and 369 females) in this study. The mean age of individuals was 61.58 ± 17.63 years. Smell and taste dysfunctions were the most common neurological symptom, reported in 619 patients (67.2%) and were associated with reduced odds of severe cases (OR: 0.63, 95%CI: 0.41 to 0.89) and death (OR: 0.61, 95%CI: 0.38 to 0.86). Myalgia (26.9%), headaches (14.8%), and dizziness (9.8%) were other common neurological symptoms. Headaches had negative correlation with severity (OR: 0.48, 95%CI: 0.34 to 0.85) and death due to COVID-19 (OR: 0.39, 95%CI: 0.77 to 0.94) but myalgia and dizziness were not correlated. The analysis of neurological and non-neurological symptoms showed that headaches and smell and taste dysfunction had the highest correlation with patients with mild severity and lower death rate.

Conclusions:
Neurological manifestations in hospitalized patients with COVID-19 can predict disease severity and mortality.             
10.1212/WNL.0000000000204025