Demystifying COVID-19 Mortality Causes with Interpretable Data Mining (preprint)

While SARS-CoV-2 infection rates are declining, older adults remain vulnerable to severe disease with high mortality. Although there have been some studies on revealing different risk factors affecting the death of COVID-19 patients, such as bilirubin, organ failure, patient age, and underlying disease, they fail to provide a comprehensive analysis to reveal their relationships and interactive effects on the risk of death. Based on the demographic information, inspection indicators, and underlying diseases of 1917 patients (102 were dead) admitted to Xiangya Hospital over a 4-month period, we used the association rule mining method to identify the risk factors leading causes of death among elderly Omicron patients. Firstly, we used the Affinity Propagation clustering to extract key features such as blood parameters, liver function indicators, renal function indicators, coagulation function indicators, and underlying diseases affecting death from the dataset. Then, we applied the Apriori to obtain 7 groups of abnormal feature combinations with significant increments in mortality rate. The results showed a relationship between the number of abnormal feature combinations and mortality rates within different groups. For instance, patients with “C-reactive protein > 8 mg/L”, “neutrophils percentage > 75.0 %”, “lymphocytes percentage < 20 %”, and “albumin < 40 g/L” have a 2x mortality rate than the basic one. If the characteristics of “D-dimer > 0.5 mg/L” and “WBC > 9.5 * 10 9 /L” are continuously included in this foundation, the mortality rate can be increased to 3x or 4x. In addition, we also found that liver and kidney diseases significantly affect patient mortality. Given patients with liver and renal diseases associated with other abnormal features, their mortality rate can be as high as 100 %. These findings can support auxiliary diagnosis and treatment to, facilitate early intervention in patients, thereby reducing patient mortality.

Virologic and clinical characteristics for prognosis of severe COVID-19: a retrospective observational study in Wuhan, China (preprint)

Background: The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has progressed to a pandemic associated with substantial morbidity and mortality. The WHO and the United States Center for Disease Control and Prevention (CDC) have issued interim clinical guidance for management of patients with confirmed coronavirus disease (COVID-19), but there is limited data on the virologic and clinical characteristics for prognosis of severe COVID-19. Methods: A total of 50 patients with severe COVID-19 were divided into good and poor recovery groups. The dynamic viral shedding and serological characteristics of SARS-CoV-2 were explored. The risk factors associated with poor recovery and lung lesion resolutions were identified. In addition, the potential relationships among the viral shedding, the pro-inflammatory response, and lung lesion evolutions were characterized. Results: A total of 58% of the patients had poor recovery and were more likely to have a prolonged interval of viral shedding. The longest viral shedding was 57 days after symptom onset. Older age, hyperlipemia, hypoproteinemia, corticosteroid therapy, consolidation on chest computed-tomography (CT), and prolonged SARS-CoV-2 IgM positive were all associated with poor recovery. Additionally, the odds of impaired lung lesion resolutions were higher in patients with hypoproteinemia, hyperlipemia, and elevated levels of IL-4 and ferritin. Finally, viral shedding and proinflammatory responses were closely correlated with lung lesion evolutions on chest CT. Conclusions Patients with severe COVID-19 have prolonged SARS-CoV-2 infection and delayed intermittent viral shedding. Older age, hyperlipemia, hypoproteinemia, corticosteroid usage, and prolonged SARS-CoV-2 IgM positive might be utilized as predicative factors for the patients with poor recovery.