Search strategy and selection criteria
The systematic review was conducted according to the protocol registered in the PROSPERO database (CRD42020176589). Following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guideline throughout this review, data were identified by searches of MEDLINE, Embase and references from relevant articles using the search terms "COVID", “SARS-CoV-2”, and "novel coronavirus" (Appendix 1). Articles published in English up to 27th March, 2020 were included. Studies with original data including original articles, short and brief communication, letters, correspondences were included. Editorials, viewpoints, infographics, commentaries, reviews, or studies without original data were excluded. Studies were also excluded if they were animal studies, modelling studies, or did not measure symptoms presentation, laboratory findings, treatment and therapeutics during hospitalization.
After the removal of duplicate records, two reviewers (CW and CHA) independently screened the eligibility criteria of study titles, abstracts and full-texts, and reference lists of the studies retrieved by the literature search. Disagreements regarding the procedures of database search, study selection and eligibility were resolved by discussion. The second and the last authors (JW and AW) verified the eligibility of included studies.
Outcomes definitions
Signs and symptoms were defined as the presentation of fever, cough, sore throat, headache, dyspnea, muscle pain, diarrhea, and rhinorrhea at the hospital admission.
Laboratory findings included a complete blood count (white blood count, neutrophil, lymphocyte, platelet count), procalcitonin, prothrombin time, urea, and serum biochemical measurements (including electrolytes, renal-function and liver-function values, creatine kinase, lactate dehydrogenase, C-reactive protein, Erythrocyte sedimentation rate), and treatment measures (i.e. antiviral therapy, antibiotics, corticosteroid therapy, mechanical ventilation, intubation, respiratory support, and renal replacement therapy). Radiological outcomes included bilateral involvement identified and pneumonia identified by chest radiograph.
Comorbidities of patients evaluated in this study were hypertension, diabetes, chronic obstructive pulmonary disease (COPD), cardiovascular disease, chronic kidney disease, liver disease and cancer.
In-patient treatment included intensive care unit admission, oxygen therapy, non-invasive ventilation, mechanical ventilation, Extracorporeal membrane oxygenation (ECMO), renal replacement therapy, and pharmacological treatment. Use of antiviral drugs (Lopinavir/ritonavirm, Ribavirin, Interferon-alpha, or Interferon-beta), antibiotic drugs, corticosteroid, and inotropes (Nor-adrenaline, Adrenaline, Vasopressin, Phenylephrine, Dopamine, or Dobutamine) were considered.
Data analysis
Three authors (CW, EHMT and CHA) extracted data using a standardized spreadsheet to record the article type, country of original, surname of first author, year of publications, sample size, demographics, comorbidities, symptoms, laboratory and radiology results, pharmacological and non-pharmacological treatments.
We aggregated estimates across 90 eligible studies to produce meta-analytic estimates using a random-effects model. For dichotomous outcomes, we estimated the proportion and its respective 95% confidence interval. For laboratory parameters as continuous outcomes, we estimated the mean and standard deviation from the median and interquartile range if the mean and standard deviation were not available from the study (12), and calculated the mean and its respective 95% confidence intervals. Random-effect models on DerSimonian and Laird method were adopted due to the significant heterogeneity, checked by the I2 statistics and the p-values. I2 statistic of <25%, 25-75% and ≥75% is considered as low, moderate, high likelihood of heterogeneity. Pooled estimates were calculated and presented by using forest plots. Publication bias was estimated by Egger’s regression test. Funnel plots of outcomes were also presented to assess publication bias.
All statistical analyses were conducted using the STATA Version 13.0 (Statacorp, College Station, TX). The random effects model was generated by the Stata packages ‘Metaprop’ for proportions (13) and ‘Metan’ for continuous variables (14).
Ethical approval
This study was approved by the Institutional Review Board of the University of Hong Kong/ Hospital Authority Hong Long West Cluster (Reference Number: UW 20-112).
Role of the funding source
There was no funding source for this study.