Data analysis using training and testing data sets
Model building involved partitioning the data set into ‘training’ and
‘testing’ sets. We apportioned 70% of the data to train the model and
used the remaining 30% to test the model, using random selection
algorithms. Following development of the model, we analyzed a set of
test data to compare predicted versus observed results to validate the
model. The regression model generated a prediction formula as follows:
Outcome = -36.898 - 0.020 AGE + 0.894 COMORBID - 0.048 Viral
Load - 0.004 IFNα + 0.444 Fever - 0.003 IL6 + 0.271 D-Dimer + 0.000
Ferritin - 0.000 Lymphocyte Count - 0.037 O2 saturation
- 2.57034e-006 NAB
The linear coefficients of the prediction equation determined the
weights of each parameter to predict the clinical outcome.