Application of ordinal logistic regression analysis to identify the determinants of illness severity of COVID-19 in China

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Corona Virus Disease 2019 (COVID-19) has presented an unprecedented challenge to the health-care system across the world. The current study aims to identify the determinants of illness severity of COVID-19 based on ordinal responses. A retrospective cohort of COVID-19 patients from four hospitals in three provinces in China was established, and 598 patients were included from 1 January to 8 March 2020, and divided into moderate, severe and critical illness group. Relative variables were retrieved from electronic medical records. The univariate and multivariate ordinal logistic regression models were fitted to identify the independent predictors of illness severity. The cohort included 400 (66.89%) moderate cases, 85 (14.21%) severe and 113 (18.90%) critical cases, of whom 79 died during hospitalisation as of 28 April. Patients in the age group of 70+ years (OR = 3.419, 95% CI: 1.596–7.323), age of 40–69 years (OR = 1.586, 95% CI: 0.824–3.053), hypertension (OR = 3.372, 95% CI: 2.185–5.202), ALT >50 μ/l (OR = 3.304, 95% CI: 2.107–5.180), cTnI >0.04 ng/ml (OR = 7.464, 95% CI: 4.292–12.980), myohaemoglobin>48.8 ng/ml (OR = 2.214, 95% CI: 1.42–3.453) had greater risk of developing worse severity of illness. The interval between illness onset and diagnosis (OR = 1.056, 95% CI: 1.012–1.101) and interval between illness onset and admission (OR = 1.048, 95% CI: 1.009–1.087) were independent significant predictors of illness severity. Patients of critical illness suffered from inferior survival, as compared with patients in the severe group (HR = 14.309, 95% CI: 5.585–36.659) and in the moderate group (HR = 41.021, 95% CI: 17.588–95.678). Our findings highlight that the identified determinants may help to predict the risk of developing more severe illness among COVID-19 patients and contribute to optimising arrangement of health resources.

This publication has 23 references indexed in Scilit: