We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Prediction of 28 Days Mortality Using Tree-structured Survival Analysis in COVID-19 Patients.
- Authors
Sohyun Bae; Yoonjung Kim; Soyoon Hwang; Ki Tae Kwon; Hyun-Ha Chang; Shin-Woo Kim
- Abstract
배경 The COVID-19 pandemic has resulted more than 600,000 deaths in July, 2020 and has shown a global pandemic. We aimed to construct a prediction algorithm, to predict the prognosis in COVID-19 inpatients. 방법 We reviewed medical records of patients who were admitted with PCR-confirmed COVID-19 in Kyungpook National University Hospital and Kyungpook National University Chilgok Hospital between February 18 and July 26, 2020. The 160 patients were included in the analyses. The logistic regression and Cox regression model was used for adjustment of the parameters, and the tree-structured survival analysis of the relevant variables was used to find predictors independently associated with 28-day mortality. The statistical analyses were performed using R statistics ver. 3.1. 결과 One-hundred sixty patients (100 from Kungpook National University and 60 from Kyungpook National University Chilgok Hospital) were included in this study, of whom 34 (21.2%) had mechanical ventilator treatment and 47 (29.4%) were died within 28-days of admission. The median age was 68.0 years; 74 (46.2%) were female. One-hundred twenty (80%) patients had comorbidities. Female and usage of steroid decreased significantly the 28-days mortality and shock, fever, O2 treatment, cardiovascular disease, neurologic disease, National Early Warning Score (NEWS) score >3, CRP >5.0 ㎎/㎗ showed the significant risk factor of the 28-days mortality by univariate analysis. Cox regression analysis showed increasing odds of 28-days death associated with age >70 (OR 2.68, 95% CI 1.30-5.52, p=0.0075) and CRP >5.0 ㎎/㎗ (8.62, 2.07-35.87; p=0.0031), and showed decreasing odds of 28-days death associated with usage of steroid (0.36, 0.17-0.80; p=0.031) and female (0.56, 0.28-1.09, p=0.088, not significant). Among the parameters independently associated with 28-day mortality, CRP >5.0 ㎎/㎗ and age over 70 were identified as critical nodes in the tree-structured survival analysis (Figure 1). 결론 Early and prominent predictors of clinical outcomes are needed to improve the clinical outcomes in the COVID-19 patients. This study shows a readily applicable parameters to determine the 28 days mortality with COVID-19 and suggests that the high CRP level and old age are important for predicting the clinical outcomes in patients with COVID-19.
- Subjects
COVID-19; FORECASTING; COVID-19 pandemic; SURVIVAL analysis (Biometry); PUBLIC hospitals
- Publication
Infection & Chemotherapy, 2020, Vol 52, pS323
- ISSN
2093-2340
- Publication type
Article