<p><strong>Background:</strong> The coronavirus epidemic, defined as (COVID-19, firstly appeared in China (Wuhan city) in Dec. month 201 and rapidly spread worldwide. Therefore, accurate information about all clinical and non-clinical features of this epidemic has become important for societies and countries. The goal of this research is to investigate clinical, demographic, and laboratory variables of COVID-19 and their influence on patient response (death/recovery) by using robust statistical methods to solve the multicollinearity problem among independent variables. <strong>Methods:</strong> This study was based on a randomly selected sample of 200 patients using the RAND () function in Python, selected from the Zenodo database for two New York hospitals in 2022. We proposed the use of Neural Network Logistic Regression compared with Ridge Logistic Regression to solve problems in the data, such as multicollinearity. <strong>Results:</strong> In assessing the effects of variables on the clinical outcome of COVID-19 patients (death or discharge), it was found that the factors (Age, EOS PCT, LYM PCT, MONO PCT, WBC, MCHC, LUC ABS, Glucose Random, Protein Total, BASO PCT, K Bld, NEU ABS) exhibited statistically significant effects with 0.0001 &lt; p-value &lt; 0.05, whereas the remaining factors showed no statistically significant effects (p-value &gt; 0.05).<strong>Conclusion:</strong> The proposed method demonstrated high efficiency in predicting patient outcomes, achieving an accuracy of 0.85 and an area under the curve (AUC) of 0.95. In comparison, the logistic regression method achieved an accuracy of 0.825 and an AUC of 0.817.</p>