ABSTRACT
Background With the COVID-19 epidemic breakout in China, up to 25% of diagnosed cases are considered to be severe. To effectively predict the progression of COVID-19 via patients’ clinical features at an early stage, the prevalence of these clinical factors and their relationships with severe illness were assessed.Methods In this study, electronic databases (PubMed, Embase, Web of Science and Chinese database) were searched to obtain relevant studies including information on severe patients. Publication bias analysis, sensitivity analysis, prevalence, sensitivity, specificity, likelihood ratio, diagnosis odds ratio calculation, and visualization graphics were achieved through software Review Manager 5.3, STATA 15, Meta-DiSc 1.4 and R.Results Data of 3.547 patients from 24 studies was included in this study. The results revealed that patients with chronic respiratory system diseases (pooled positive likelihood 6.07, 95% CI: 3.12-11.82), chronic renal disease (4.79, 2.04-11.25), cardiovascular disease (3.45, 2.19-5.44), and symptoms of the onset of chest tightness (3.8, 1.44-10.05), shortness of breath (3.18, 2.24-4.51), and diarrhea (2.04, 1.38-3.04) exhibited increased probability of progressing to severe illness. C-reactive protein, ratio of neutrophils to lymphocytes, and erythrocyte sedimentation rate increased a lot in severe patients compared to non-severe. Yet it was found that clinical features including fever, cough, headache, and so on, as well as some comorbidities have little warning value.Conclusions The clinical features, and laboratory examination could be used to estimate the process of infection in COVID-19 patients. The findings contribute to the more efficient prediction of serious illness for patients with COVID-19 to reduce mortality.Systematic Review registrations Not applicable.