Patient demographic characteristics
Between 12th February and 19th February 2020, 1,077 adult COVID-19 patients were admitted to the First Hospital of Wuhan; among them, 141 (13.1%) consecutive patients aged ≥75 years were included in the present study. The median age was 81 years (IQR,78-85), ranging from 75 to 97 years, and the majority were female (84, 59.6%). The median time from disease onset to hospitalization was 10 (IQR, 6.0-14.5)days.
Comorbidities were present in 76.6% of patients, with cardiovascular disease including hypertension, coronary heart disease, and atrial fibrillation being the most common comorbidity (63.1%). Other comorbidities included endocrine disorders (25.5%, including diabetes, thyroid dysfunction), central nerve system disorders(21.3%, including stroke, Alzheimer's disease, and Parkinson's disease), chronic kidney disease (3.5%), and malignancy (5.7%). Epidemic investigation indicated that no patient had a history of direct contact with Huanan seafood market in Wuhan city.
Overall, 18 patients died (Death group) during hospitalization, corresponding to an in-hospital fatality rate of 12.8%. The cause of death was refractory shock in eight patients and acute respiratory distress syndrome (ARDS) in ten patients (sustained hypoxia despite of ventilation in five patients). By 31st March 2020, the remaining 123 patients had been discharged (Discharged group) from hospital after proper treatment. The median time from admission to discharge was 29(IQR, 25.0-36.0) days and from illness onset to death or discharge was 39.0(IQR, 31.0-48.0)days. The latter was significantly shorter in the Death group (26.5, IQR 14.8-38.8 days vs. 40.0, IQR 33.0-50.0 days in the Discharged group, P<0.0001).
The age-stratified fatality rate is provided in Figure 1 with the highest fatality rate at 27.3% in the ≥90 years subgroup. Although there seemed a trend of an increase of fatality rate by age, the comparison across all age subgroups did not achieve statistical significance (χ2=5.259，P=0.154).
Clinical presentations and laboratory tests on admission
The majority of patients (64.5%) reported being febrile since the onset of disease but only 11.3% of the population had fever (defined as an axillary temperature of >37.3°C) on arrival. Other common symptoms on admission were dry cough (53.9%), fatigue (39.0%), and expectoration (35.5%) as listed in Table 1. The incidence of all symptoms was comparable between the Death and the Discharged groups.
Table 1 Clinical Characteristics on Admission of Patients in the Discharged and Death Groups
Some patients presented with signs suggesting unstable condition including heart rate of >100bpm (9.9%), respiratory rate of >24 breaths/min (11.3%), systolic blood pressure of ≤90mmHg (1.4%), and SpO2≤90% on room air (13.5%). When comparing between groups, there were significantly more patients in the Death group with these signs on admission (Table 1).
The laboratory test results on admission are summarised in Table 2. In the Death group, there was a markedly higher percentage of patients with abnormal findings in terms of the white blood cell count, lymphocyte count, neutrophil count, platelet count, CRP, PCT, D-dimer, blood urea nitrogen, creatinine, lactic dehydrogenase (LDH), N-terminal pro-brain natriuretic peptide (NT-proBNP), and hypersensitive troponin I（hs-TnI）. All patients underwent a CT scan immediately before or after they were admitted to hospital; 37.6% had typical ground-glass opacity lesions in at least three lung lobes.
Table 2 Laboratory Findings on Admission in the Discharged and Death Groups
Overall, 38 patients (27%) were classified as severe or critical cases on admission and required immediate intensive care. These patients accounted for 77.8% of all cases of in-hospital death. This ratio was markedly higher than that in the Discharged group (24 severe/critical cases on admission of 123 discharged patients, 19.5%, P<0.0001).
Complications and treatment during hospitalization
After admission, 27.7% of the patients developed de novo arrhythmias including atrial fibrillation with rapid ventricular response, atrial flutter, atrial tachycardia, and first- or second-degree heart block, but no life-threatening arrhythmia was observed in our population. The incidence of arrhythmic occurrence in two groups was similar. Of the 123 patients who were eventually discharged, only seven (5.7%) deteriorated to ARDS during hospitalization, while in the Death group, this incidence was 55.6% (P<0.0001). Deterioration to refractory shock was seen in eight patients who all died.
Arbidol was the most frequently (91.5%) used antiviral medicine. Most patients received antibiotics (76.6%), while 20.6% received glucocorticoid therapy. Mechanical ventilation, either non-invasive or invasive (IMV), was required in 21 cases (14.9%); six of these patients were intubated (IMV).Other treatments for COVID-19 are detailed in Table 3. Patients who received antibiotics, antifungal agents, or glucocorticoids; who needed immunoglobulin or blood transfusion; and who required mechanical ventilation were more likely to have a fatal outcome.
Table 3 Complication and Treatment during hospitalizationin the Discharged and Death Groups
Risk factors of in-hospital death
We tested the demographic and laboratory variables that exhibited a borderline significant or significant difference between the Discharged and Death groups in the univariate regression analysis, with the exception of age and CRP. Because there was collinearity between the variables of WBC, lymphocyte, and neutrophil, we excluded WBC and neutrophil. So totally 16 variables were included in the multivariate regression. The results indicated that male sex (odds ratio [OR]=13.1, 95% confidence interval [CI]1.1 to 160.1, P=0.044) and three on-admission check-ups including body temperature of>37.3°C (OR=80.5, 95%CI 4.6 to 1407.6, P=0.003), SPO2 of ≤90% without additional oxygen supply (OR=70.1, 95%CI 4.6 to 1060.4, P=0.002), and NT-proBNP of >1800ng/L (OR=273.5, 95%CI 14.7 to 5104.8, P<0.0001) were independent risk factors of fatal outcome.
To predict in-hospital death for elderly COVID-19 patients, a statistical model, namely Sex, Temperature, SPO2,and NT-proBNP(STONP), was developed using logistic regression and then the ROC curve was plotted. In the present derivation cohort, the AUC of the STONP model was 0.971 (95% CI 0.928 to 0.992) with a negative predictive value of 98.4% and a positive predictive value of 77.8%. As a validation cohort was unavailable at this time, we compared the STONP model with the Mortality Probability Models II-Admission (MPM-II Adm), which has been widely used in intensive care medicine to calculate the possibility of in-hospital death. The performance of the STONP model was comparable with that of the MPM-II Adm (AUC 0.915, 95% CI 0.856 to 0.955; z statistic 1.814, P=0.0697 vs. STONP; Figure 2). A web-based tool and an App(Android system only)of the STONP model are available at this hyperlink: https://janzhou.org/covid-19/stonp.html or via the barcodes provided in the Additional file 1.
Table 4 Univariate and Multivariate Logistic Regression for Prediction of In-hospital Death