COVID-19 mortality prediction using ensemble learning and grey wolf optimization.

COVID-19 is now often moderate and self-recovering, but in a significant proportion of individuals, it is severe and deadly. Determining whether individuals are at high risk for serious disease or death is crucial for making appropriate treatment decisions. We propose a computational method to estimate the mortality risk for patients with COVID-19. To develop the model, 4,711 reported cases confirmed as SARS-CoV-2 infections were used for model development. Our computational method was developed using ensemble learning in combination with a genetic algorithm. The best-performing ensemble model achieves an AUCROC (area under the receiver operating characteristic curve) value of 0.7802. The best ensemble model was developed using only 10 features, which means it requires less medical information so that the diagnostic cost may be reduced while the prognostic time may be improved. The results demonstrate the robustness of the used method as well as the efficiency of the combination of machine learning and genetic algorithms in developing the ensemble model.

Short-term Moderate-dose Corticosteroid Plus Immunoglobulin Effectively Reverses Covid-19 Patients Who Have Failed Low-dose Therapy (preprint)

Background: The coronavirus disease-19 (COVID-19) has spread globally with more than 6,750,000 people infected, and nearly 400,000 patients died. Currently, we are in an urgent need for effective treatment strategy to control the clinical deterioration of COVID-19 patients.Methods: The clinical data of 10 COVID-19 patients receiving short-term moderate-dose corticosteroid (160mg/d) plus immunoglobulin (20g/d) were studied in the North Yard of The First Hospital of Changsha, Hunan from January 17th to February 27th, 2020. Epidemiological, clinical, laboratory, radiological findings were analyzed. Results: After treatment with combination of low-dose corticosteroid (40-80mg/d) and immunoglobulin (10g/d), patients’ lymphocyte count (0.88±0.34 vs 0.59±0.18, P<0.05), oxygenation index including SPO2 (94.90±2.51 vs 90.50±5.91, P<0.05) and PaO2/FiO2 (321.36±136.91 vs 129.30±64.97, P<0.05) were significantly lower than pre-treatment, and CT showed that the pulmonary lesion deteriorated in all patients. While after treatment of short-term moderate-dose corticosteroid plus immunoglobulin, patients’ APACHE Ⅱ score (9.10±6.15 vs 5.50±9.01, P<0.05), body temperature (37.59±1.16 vs 36.46±0.25, P<0.05), lymphocyte count (0.59±0.18 vs 1.36±0.51, P<0.05), Lactate dehydrogenase (419.24±251.31 vs 257.40±177.88, P<0.05), and C-reactive protein (49.94±26.21 vs 14.58±15.25, P<0.05) significantly improved compared with post-treatment with low-dose therapy. In addition, oxygenation index including SPO2 (90.50±5.91 vs 97.50±1.18, P<0.05), PaO2 (60.47±14.53 vs 99.07±34.31, P<0.05), and PaO2/FiO2 (129.30±64.97 vs 340.86±146.72, P<0.05) significant improved. Furthermore, CT showed that pulmonary lesions obviously improved in 7 patients. After systematic therapy, 4 out of 10 COVID-19 patients recovered and discharged.Conclusions: Short-term moderate-dose corticosteroid plus immunoglobulin is effective for reversing the continued deterioration of COVID-19 patients who failed to respond to the low-dose therapy.

Screening COVID-19 with Chest CT in Preference to RT-PCR: Based on Mortality Analysis (preprint)

Corona Virus Disease 2019 (COVID-19) became the outbreak of infectious diseases emergency worldwide. It remains unknown whether the RT-PCR test results was associated with the prognosis of COVID-19 patients or not. In this study, a total number of 495 patients with typical chest CT feature and symptom consistent with COVID-19 were retrospectively included from Jan 23, 2020 to Feb 26, 2020. 186 (37.58%), 32 (6.46%) and 277 (55.96%) COVID-19 patients had initial positive, suspected and negative RT-PCR results, respectively. The mean age was 58.55 years and 254(51.3%) were older than 60 years. 60.00% (297/495), 22.02% (109/495) and 17.98% (89/495) of COVID-19 patients were common, severe and critically type, respectively. There were no significant differences of age, gender, time from onset to hospitalization and severity classification in the patients with initial positive and negative RT-PCR result. The mortality rate of patients with positive and negative were 7.14% and 7.94%. Patients with initial negative or initial positive RT-PCR results had no significant difference of mortality rate (c2=4.079, p=0.130). The number of patients with lymphocyte ratios under the normal level was significantly larger in patients with initial negative RT-PCR results (59/92) compared with the patients with initial positive result (86/167), p=0.033. COVID-19 patients with positive or negative RT-PCR results had no significant difference in severity and mortality. Chest CT may be a more effective tool to screen COVID-19 in preference to RT-PCR.

A multi-centered, retrospective, descriptive study on 107 dead patients with COVID-19 (preprint)

Since the emergence of Corona Virus Disease 2019 (COVID-19) in Wuhan city, Hubei Province, China, it has caused thousands of deaths. As the ongoing outbreak of COVID-19 around the world, the number of deaths will definitely continue to increase. We aimed to further describe the clinical characteristics of dead cases with COVID-19 through a large sample and multi-centered study and to find some clinical predictors for the deterioration of COVID-19 during the process.   Methods One hundred and seven patients (16 patients from Lei Shen-Shan Hospital, 54 patients from Seventh Hospital of Wuhan and 37 patients from Zhongnan Hospital of Wuhan University) with COVID-19 were enrolled in our research from Jan 22 to Feb 29, 2020. The demographic, clinical, radiological, laboratory and treatment data of all cases were analysed.   Results Of the 107 dead patients with COVID-19, 71 (66.4%) were male and 36 (33.6%) were female. The mean age of the patients was 71.2 ± 12.1 years. 82 (76.6%) of patients had chronic diseases. The mean duration from admission to death was 9 (IQR,5-14) days. Respiratory functional damage was the most common one followed by heart and kidney. Hematuria was found in 36(33.6%) patients. 89(83.2%) patients’ albumin levels were decreased. 68(63.6%) patients had anemia. concerning laboratory results, 55 (69.6%) and 56 (70.1%) patients have the elevated white blood cells and elevated Neutrophils during the process; only 43 (54.4%) have the decreased Lymphocytes; The values of platelets and haemoglobin decreased in 64(81.0%) and 58 (73.4%) patients. Alanine aminotransferase and aspartate aminotransferase elevated in near half of patients, while almost 80% of patients have the decreased albumin. The elevated blood urea nitrogen and cystatin C were manifested in about 70% of patients. Procalcitonin was elevated in 38 (71.7%) patients.   Conclusions In conclusion, the older men with chronic diseases are more likely to die from COVID-19. Apart from that, more attention should be pay on timely treatment, coinfections, malnutrition, and dysfunction of kidney and coagulation. The rising values (white blood cell, blood urea nitrogen, cystatin C, PCT and PT) and the decreased values (PLT, Hb and albumin) maybe meaningful for predict the poor prognosis.