Predicting mortality in SARS-COV-2 (COVID-19) positive patients in the inpatient setting using a novel deep neural network

Background The nextwave of COVID-19 pandemic is anticipated to be worse than the initial one and will strain the healthcare systems even more during the winter months. Our aim was to develop a novel machine learning-based model to predict mortality using the deep learning Neo-V framework. We hypothesized this novel machine learning approach could be applied to COVID-19 patients to predict mortality successfully with high accuracy. Methods We collected clinical and laboratory data prospectively on all adult patients (≥18 years of age) that were admitted in the inpatient setting at Aga Khan University Hospital between February 2020 and September 2020 with a clinical diagnosis of COVID-19 infection. Only patients with a RT-PCR (reverse polymerase chain reaction) proven COVID-19 infection and complete medical records were included in this study. A Novel 3-phase machine learning framework was developed to predict mortality in the inpatients setting. Phase 1 included variable selection that was done using univariate and multivariate Cox-regression analysis; all variables that failed the regression analysis were excluded from the machine learning phase of the study. Phase 2 involved new-variables creation and selection. Phase 3 and final phase applied deep neural networks and other traditional machine learning models like Decision Tree Model, k-nearest neighbor models, etc. The accuracy of these models were evaluated using test-set accuracy, sensitivity, specificity, positive predictive values, negative predictive values and area under the receiver-operating curves. Results After application of inclusion and exclusion criteria (n=)1214 patients were selected from a total of 1228 admitted patients. We observed that several clinical and laboratory-based variables were statistically significant for both univariate and multivariate analyses while others were not. With most significant being septic shock (hazard ratio [HR], 4.30; 95% confidence interval [CI], 2.91–6.37), supportive treatment (HR, 3.51; 95% CI, 2.01–6.14), abnormal international normalized ratio (INR) (HR, 3.24; 95% CI, 2.28–4.63), admission to the intensive care unit (ICU) (HR, 3.24; 95% CI, 2.22–4.74), treatment with invasive ventilation (HR, 3.21; 95% CI, 2.15–4.79) and laboratory lymphocytic derangement (HR, 2.79; 95% CI, 1.6–4.86). Machine learning results showed our deep neural network (DNN) (Neo-V) model outperformed all conventional machine learning models with test set accuracy of 99.53%, sensitivity of 89.87%, and specificity of 95.63%; positive predictive value, 50.00%; negative predictive value, 91.05%; and area under the receiver-operator curve of 88.5. Conclusion Our novel Deep-Neo-V model outperformed all other machine learning models. The model is easy to implement, user friendly and with high accuracy.