Development of a multivariable prediction model for severe COVID-19 disease: a population-based study from Hong Kong
Jiandong Zhou,Sharen Lee,Xiansong Wang,Yi Li,William Ka Kei Wu,Tong Liu,Zhidong Cao,Daniel Dajun Zeng,Keith Sai Kit Leung,Abraham Ka Chung Wai,Ian Chi Kei Wong,Bernard Man Yung Cheung,Qingpeng Zhang,Gary Tse
DOI: https://doi.org/10.1038/s41746-021-00433-4
IF: 15.2
2021-04-08
npj Digital Medicine
Abstract:Abstract Recent studies have reported numerous predictors for adverse outcomes in COVID-19 disease. However, there have been few simple clinical risk scores available for prompt risk stratification. The objective is to develop a simple risk score for predicting severe COVID-19 disease using territory-wide data based on simple clinical and laboratory variables. Consecutive patients admitted to Hong Kong’s public hospitals between 1 January and 22 August 2020 and diagnosed with COVID-19, as confirmed by RT-PCR, were included. The primary outcome was composite intensive care unit admission, need for intubation or death with follow-up until 8 September 2020. An external independent cohort from Wuhan was used for model validation. COVID-19 testing was performed in 237,493 patients and 4442 patients (median age 44.8 years old, 95% confidence interval (CI): [28.9, 60.8]); 50% males) were tested positive. Of these, 209 patients (4.8%) met the primary outcome. A risk score including the following components was derived from Cox regression: gender, age, diabetes mellitus, hypertension, atrial fibrillation, heart failure, ischemic heart disease, peripheral vascular disease, stroke, dementia, liver diseases, gastrointestinal bleeding, cancer, increases in neutrophil count, potassium, urea, creatinine, aspartate transaminase, alanine transaminase, bilirubin, D-dimer, high sensitive troponin-I, lactate dehydrogenase, activated partial thromboplastin time, prothrombin time, and C-reactive protein, as well as decreases in lymphocyte count, platelet, hematocrit, albumin, sodium, low-density lipoprotein, high-density lipoprotein, cholesterol, glucose, and base excess. The model based on test results taken on the day of admission demonstrated an excellent predictive value. Incorporation of test results on successive time points did not further improve risk prediction. The derived score system was evaluated with out-of-sample five-cross-validation (AUC: 0.86, 95% CI: 0.82–0.91) and external validation ( N = 202, AUC: 0.89, 95% CI: 0.85–0.93). A simple clinical score accurately predicted severe COVID-19 disease, even without including symptoms, blood pressure or oxygen status on presentation, or chest radiograph results.
health care sciences & services,medical informatics