CLINICAL CHARACTERISTICS OF HOSPITALIZED COVID -19 INFECTED PATIENTS IN PALESTINE (Descriptive Analytics and Data Mining Approach) رسالة ماجستير

Abstract

Background: Palestine is one of the countries that has been affected by COVID-19 pandemic. This research study aims to understand and improve the clinical and diagnostic knowledge of COVID-19 patients in Palestine using conventional statistical analysis and data mining tools to develop model can classify and predict patient future health situation. Method: Quantitative research design using descriptive analysis was carried out to obtain an insight into 132 hospitalized patients. A retrospective review of patient's medical records was conducted. Data was collected from two designated hospitals in West Bank of Palestine for a period of 19 March to 20 July 2020. The clinical data includes follow up laboratory tests, clinical observation, and treatments plan during their first two days of stay in the hospital. Results: The most common symptoms on admission were cough (51.5%), fever (41.7%), and shortness of breath (25.8%). Numerous differences were reported between severe and not severe cases, including higher White blood cells (WBC), neutrophil, Lactate dehydrogenase (LDH), ferritin, BUN, and creatine (P<.05), and lower lymphocytes percentage and SPO2 (P<.001). CRP, Ferritin level, and Monocytes tests were statistically significant in their association with severity level (P-value <0.05). Having diabetes, hypertension and Cerebrovascular Disease is significantly associated with the severity level of COVID-19. The most common treatments given for COVID- 19 patients were Antipyretic (75.0%), Supplementary drugs (55.3%), Anticoagulant (47.0%), and Antibiotic (46.2%). Four popular artificial intelligent tools were used to develop a prediction model for COVID-19 severity level; Logistic Regression (LR), Support Victor Machine (SVM), Decision Tree (DT), and Artificial Neural Networks (ANNs) with accuracy (71%, 71,5%, 71,5%, and 74%) respectively. IV Conclusion: This study was conducted in two phases; First: a comprehensive descriptive analysis of the patient's demographic, comorbidities, complaints, and laboratory findings, examined the relationship between each variable with patient's health status and the severity of the symptoms experienced (severe, non-severe), This phase showed that the most common symptoms on admission were cough (51.5%), fever (41.7%), and shortness of breath (25.8%). Numerous differences were reported between severe and non-severe cases, including higher WBC, neutrophil, LDH, ferritin, BUN, and creatinine (P<.05), and lower lymphocytes percentage and SPO2 (P<.001). Relationships between CRP and severity level on the one hand and Ferritin level, and Monocytes on the other were statistically significant (P-value <0.05). Having diabetes, hypertension and Cerebrovascular Disease was significantly associated with the higher severity level of COVID-19 (P<.001). Second: The disease severity data from the first phase with artificial intelligence (AI) algorithms were used to develop an algorithm that can predict the patients’ future condition. Four popular machine learning tools were used; Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), and Artificial Neural Networks (ANNs). Models’ accuracy levels were; 71%, 71,5%, 71,5%, and 74%, respectively, all considered high accuracy levels the models registered in spite of the limited available data used in models testing. AI can be very effective to describe behavior of COVID-19 and inform clinical decisions and treatment plans,