Clinical outcomes of hospitalized COVID-19 patients with renal injury: a multi-hospital observational study from Wuhan

Renal injury is common in patients with coronavirus disease 2019 (COVID‐19). We aimed to determine the relationship of estimated glomerular filtration rate (eGFR) and acute kidney injury (AKI) with the characteristics, progression, and prognosis of COVID-19 in-patients. We retrospectively reviewed 1851 COVID-19 patients admitted to 3 hospitals in Wuhan, China. Clinical, laboratory, radiological, treatment, complication, and outcome data were analyzed. Patients were stratified according to levels of eGFR (≥ 90 vs. 60–89 vs. < 60 mL/min/1.73 m2). The risk of reaching the composite endpoint—intensive care unit admission, invasive ventilation, or death—was compared. On admission, 25.5% patients had renal impairment (eGFR < 90 mL/min/1.73 m2), but only 2.6% patients had chronic kidney disease (CKD). The overall in-hospital AKI incidence was 6.7%. Severe illness and comorbidities (hypertension, diabetes, CKD, and cardiovascular/cerebrovascular diseases) were more common among patients with low eGFR (< 90 mL/min/1.73 m2). Despite the more frequent use of intensive oxygen therapy, continuous blood purification, and glucocorticoid treatment, the prognosis of these patients was unsatisfactory, with the incidence of the composite endpoint (15.4% vs. 19.6% vs. 54.5%; P = 0.000) and complications (AKI, respiratory failure, cardiac injury, coagulation disorders, sepsis, etc.) increasing with decreasing eGFR. Kaplan–Meier survival analysis revealed that patients with eGFR < 90 mL/min/1.73 m2 or AKI had significantly escalated risks of reaching the composite endpoint. Multivariate regression analysis showed that renal insufficiency (eGFR < 60 mL/min/1.73 m2) on admission and in-hospital AKI independently predicted poor prognosis among COVID-19 in-patients. And renal impairment on admission was a greater predictor of poor prognosis in non-elderly patients than that in elderly patients. Early and continuous renal-function monitoring and early AKI diagnosis are necessary to predict and prevent the progression of COVID-19.


Results
Patient characteristics. A total of 1851 patients hospitalized with confirmed COVID-19 from 3 hospitals (1412 patients from Taikang Tongji Hospital, 335 patients from Huo Shen Shan Hospital, and 104 patients from Renmin Hospital of Wuhan University) were included in this study. To elucidate the clinical implications of renal dysfunction in COVID-19 patients, we stratified the patients into three groups according to their baseline eGFR as follows: 1379 patients (74.5%), ≥ 90 mL/min/1.73 m 2 (high eGFR group, HEG, normal renal function); 360 patients (19.4%, ), 60-89 mL/min/1.73 m 2 (medium eGFR group, MEG, mild renal impairment); and 112 patients (6.1%), ˂60 mL/min/1.73 m 2 (low eGFR group, LEG, moderate/severe renal function impairment). The characteristics of the 1851 COVID-19 patients are presented in Table 1. The median age was 62 years (IQR, 51-70 years), and 889 (48.0%) patients were male. Severe cases accounted for 37.9% of the study population. The most common symptoms on admission were fever (1099 patients, 59.3%) and cough (1044 patients, 56.4%). Asymptomatic patients accounted for 2.1% of the study population (Supplementary Table 1). The median duration from symptom onset to admission was 25 days (IQR, 14-38 days). Comorbidities were present in half the , patients in the MEG and LEG were older, had a higher prevalence of severe illness, were more likely to experience shortness of breath and chest distress/palpitations, and were less likely to have fever and sputum production. Moreover, comorbidities, including hypertension, diabetes, coronary heart disease, cerebrovascular disease, and CKD, were present more often among patients in the MEG and LEG.
Laboratory and radiological findings. Table 2 shows the laboratory and radiological findings on admission. In the overall study population, the median values of all laboratory indicators were within the normal range. Of the 1463 CT scans that were performed on admission, 87.6% revealed abnormal results.
Compared with patients in the HEG, patients in the MEG and LEG showed significantly elevated cardiac injury indicators (high-sensitivity cardiac troponin I, myoglobin, and lactate dehydrogenase), liver injury indicators (alanine and aspartate aminotransferases), inflammation-related indicators (C-reactive protein, interleukin-6, and procalcitonin), and coagulation function indicators (prothrombin time, fibrinogen, and d-dimer), and significantly decreased platelet count, hemoglobin, and albumin. Unlike patients in the HEG, patients in the MEG and LEG tended to have bilateral pneumonia rather than subpleural or unilateral pneumonia. Treatment, complications, and clinical outcomes. The treatment, complications, and in-hospital clinical outcomes of the patients are summarized in Table 3. Of the 1851 patients, 58.9% patients required oxygen support in the hospital. The most intense level was recorded, including low-flow (46.7%) and high-flow nasal cannula (HFNC, 6.2%) oxygen inhalation, noninvasive mechanical ventilation (NMV, 1.9%), and invasive mechanical ventilation (IMV, 4.1%). Only 1.1% patients required continuous renal replacement therapy (CRRT). Extracorporeal membrane oxygenation was used in 2 patients. Antiviral therapy was the most common treatment (64.7%), followed by antibiotic therapy (41.4%) and glucocorticoids (11.9%). Renin-angiotensin system (RAS) inhibitors and diuretics were administered to 10.9% and 10.2% patients, respectively. In addition, 23 (1.3%), 44 (2.5%), and 16 (0.9%) patients were treated with tocilizumab, umbilical cord mesenchymal stem cells, and convalescent plasma, respectively. Coagulation disorder (66.2%) was the most frequent complication, followed by anemia, hypoproteinemia, electrolyte disturbances, acute cardiac injury, respiratory failure, acidosis, and ARDS. A total of 1690 patients were followed up for a median of 16 days, during which 90.7% (1532) patients were discharged, 4.9% (83) patients died, and 4.4% (75) patients remained hospitalized. The remaining patients (161) were lost to follow-up.
Compared with the HEG, the MEG and LEG were associated with more frequent use of HFNC, NMW, IMV, and CRRT. The use of antibiotics, glucocorticoids, RAS inhibitors, and diuretics was also significantly higher in the MEG and LEG than in the HEG. The frequency of complications was higher in the MEG and LEG than in the HEG. In addition to secondary infection, other complications were more common among patients in the MEG and LEG than among those in the HEG; these included respiratory failure, ARDS, acute heart failure, acute cardiac injury, coagulation disorders, sepsis, shock, anemia, hypoproteinemia, electrolyte disturbances, and acidosis.
Renal injury and in-hospital prognosis. On admission, 4.7% and 13.5% patients had elevated SCr and BUN, respectively; 30.5% patients had hematuria, and relatively fewer patients (19.1%) had proteinuria ( Table 2). Only 5.3% and 1.9% patients had 2 + -3 + hematuria and proteinuria, respectively. Compared with patients in the HEG, patients in the MEG and LEG had elevated renal dysfunction indicators (SCr and BUN). In addition, hematuria and proteinuria at presentation were more common in the MEG and LEG than in the HEG ( Table 2). During hospitalization, the peak SCr increased gradually with decreasing eGFR ( Table 2). The incidence of AKI in the entire cohort was 6.7% according to KDIGO criteria (Table 3). Stage 1 AKI was present in 61.4% of the patients with AKI, while stages 2 and 3 were reached in 22.8% and 15.8% patients, respectively. The incidence of in-hospital AKI was higher in the MEG and LEG than in the HEG (Table 3).
Two models of multivariate analysis were calculated. The first (Model 1) was designed to address whether eGFR predicted outcomes independently of disease severity, preexisting disorders, and baseline clinical data. The second model (Model 2) was designed to address whether eGFR was an independent predictor of outcomes even when follow-up data during hospitalization were taken into account; this model included disease severity, preexisting disorders, and baseline and follow-up data. Model 1 showed a significantly higher risk of the composite endpoint (reference: HEG, ≥ 90 mL/min/1.73 m 2 ; MEG, 60-89 mL/min/1.73 m 2 : HR = 1.538, 95% CI: 0.978-2.421, P = 0.063; and LEG, < 60 mL/min/1.73 m 2 : HR = 1.953, 95% CI: 1.174-3.250, P = 0.01; Table 5 Table 6) in the LEG than in the HEG and MEG. In addition, we found that eGFR was an independent predictor of outcomes in both 18-64 (non-elderly) and > 65 years old (elderly) groups. Patients aged 18 to 64 years had a 4.868-fold higher risk of the composite endpoint in the LEG group than those in the HEG group. The risk of the composite endpoints in LEG group was 2.379 times higher than that in HEG group in patients older than 65. The results indicated that renal impairment on admission was a greater predictor of poor prognosis in non-elderly patients than that in elderly patients (  Table 5). The results of the unadjusted analysis are presented in Tables 5 and 6. Overall, the findings of the unadjusted and adjusted analyses were not materially altered. A directed acyclic graph (DAG) showed the association between risk factors and the poor prognosis of COVID-19 based on the model variables in Cox regression (Fig. 3).

Discussion
In this multi-hospital observational study, we found that renal function was closely related to the prognosis of hospitalized COVID-19 patients. More than 25% of the patients had impaired renal function, with an eGFR lower than 90 mL/min/1.73 m 2 . Patients with abnormal eGFR (< 90 mL/min/1.73 m 2 ) are more susceptible to severe COVID-19 infection. Although patients with impaired renal function received more intensive oxygen therapy, continuous blood purification, and glucocorticoid treatment, their prognosis was unsatisfactory, with a higher incidence of the composite endpoint (ICU admission, IMV, or death) and complications (AKI, respiratory failure, ARDS, acute cardiac injury, coagulation disorders, sepsis, shock, anemia, hypoproteinemia, electrolyte disturbances, and acidosis). Multivariate regression analysis showed that both eGFR < 60 mL/min/1.73 m 2 on admission and AKI occurrence during hospitalization were independent risk factors for poor in-hospital prognosis. We found that eGFR was an independent predictor of outcomes in both 18-64 (non-elderly) and > 65 years old (elderly) groups. The results showed that patients aged 18 to 64 years had a 4.868-fold higher risk of the composite endpoint in the LEG group than those in the HEG group. The risk of the composite endpoints in LEG group was 2.379 times higher than that in HEG group in patients older than 65 years. These results indicated that renal impairment on admission was a greater predictor of poor prognosis in non-elderly patients than that in elderly patients. So we drew the DAG diagram by taking age, eGFR and the compound endpoint as the main factors. Diabetes and hypertension can aggravate the progression of CKD and also lead to poor prognosis in patients with COVID-19 19,20 . So diabetes and hypertension are important confounding factors. Moreover, CKD  www.nature.com/scientificreports/ patients had a higher risk of severe COVID-19 21 and complications such as AKI 22 . In addition, our data showed that low eGFR was associated to the severe disease and the occurrence of complications. Therefore, eGFR may influence the outcome of patients with COVID-19 by regulating these two factors. Serum SCr and BUN are commonly used to evaluate renal function, however, their levels are also affected by age, race, gender and body type. As the best overall index of kidney function by fully adjusting the effect of other factors on serum SCr, eGFR is the only criterion for staging CKD and determining long-term renal outcomes 23 . The normal eGFR is ≥ 90 mL /min/1.73m 2 ; an eGFR of 60-89 mL/min/1.73 m 2 indicates mild renal impairment; and an eGFR < 60 mL/min/1.73 m 2 indicates moderate renal impairment; an eGFR < 15 mL /min/1.73m 2 indicates renal failure and requires dialysis or renal transplantation. A study of a multicenter registry found that 30% of COVID-19 patients had kidney dysfunction upon admission (eGFR < 60 mL/min/1.73 m 2 ), which was associated with higher in-hospital mortality 24 . It's important to note that eGFR at the admission may be reduced due to long duration of COVID-19 (25 days) in our study, but that this situation is similar for all the subjects. Our study showed that the incidence of complications, the composite endpoint (15.4% vs. 19.6% vs. 54.5%, P = 0.000), and mortality (2.4% vs. 7.4% vs. 27.7%, P = 0.000) increased gradually with decreasing eGFR. The risk of reaching the composite endpoint among patients in the MEG and LEG was 2.1 and 5.6 times, respectively, of the risk   www.nature.com/scientificreports/ among those in the HEG (P = 0.000). This suggests that patients with renal function impairment (eGFR < 90 mL/ min/1.73 m 2 ) on admission had significantly worse in-hospital outcomes, although the significance was lost in the MEG (eGFR: 60-89 mL/min/1.73 m 2 ) after adjustments for baseline and follow-up clinical data using the Cox regression model. Our results indicated that patients with an eGFR of 60-89 mL/min/1.73 m 2 had already escalated risks of poor prognosis, emphasizing the need for early and continuous monitoring of renal function. The prevalence of CKD in China is 10.8% 25 . However, in our study, only 2.6% of patients with COVID-19 had a history of CKD. We also found that 19.1% and 30.5% patients had proteinuria and hematuria, respectively. Thus, CKD may be underestimated in these patients on admission. Patients with kidney injury, especially those with GFR < 60 mL/min/1.73 m 2 , were less likely to develop fever, with an incidence of less than 50%. Despite a low incidence of clinical symptoms, patients with eGFR < 90 mL/min/1.73 m 2 had a higher incidence of complications, the composite endpoint, and mortality. This may be due to low immunity (reduced lymphocyte count) 26,27 , coagulation disorders (decreased platelet count, prolonged prothrombin time, and increased d-dimer) 28,29 , poor nutrition (anemia and hypoproteinemia), and persistent inflammatory states (high procalcitonin, interleukin-6, and C-reactive protein) 30 in patients with impaired renal function (eGFR < 90 mL/min/1.73 m 2 ). In addition, anorexia on admission is a risk factor for poor prognosis (Supplement Table 2), which may be associated with subsequent malnutrition.
Acute kidney injury (AKI), as a common complication of COVID-19, is usually related to disease mortality 31 . A postmortem patient series found significant acute tubular injury in all patients who had died of COVID-19 32 . The mechanisms of kidney injury in SARS-CoV-2 infection include direct viral injury via the angiotensin-converting enzyme 2 receptor, which is highly expressed in the kidneys 33 , an imbalanced renin-angiotensin-aldosterone system 34,35 and release of proinflammatory cytokines elicited by the viral infection and microvascular thrombosis 36 . We found that patients with eGFR < 60 mL/min/1.73 m 2 were more likely to develop AKI during Table 6. Multivariate Cox regression analysis of associations of baseline eGFR with the mortality in patients with COVID-19. a adjusted for disease severity, hypertension, diabetes, coronary artery heart disease, cerebrovascular disease, leukocyte count, lymphocyte count, CRP, IL-6, and procalcitonin. COVID-19, coronavirus disease 2019; eGFR, estimated glomerular filtration rate; HR, hazard ratio; CI, confidence interval.  www.nature.com/scientificreports/ hospitalization, with an incidence of 17%. Furthermore, 10% of patients with eGFR < 60 mL/min/1.73 m 2 received CRRT. A study in New York has reported a high incidence of AKI (37%-46%) among COVID-19 patients 32 and 19% of patients with AKI required dialysis, and half of them died in the hospital. We found that AKI was an independent risk factor in patients with COVID-19 after adjustments. Therefore, regular monitoring of renal function and timely diagnosis of AKI are conducive to the treatment of COVID-19 patients. Our study has some limitations. First, the incidence of CKD and AKI may be underestimated in some patients due to the lack of baseline medical records, late admission, and lack of renal function examination after admission. Second, a small proportion of patients were still in the hospital, and their outcomes were unknown at the time of the data cutoff, which might lead to the underestimation of the endpoint events. Third, due to different diagnostic paradigms in different hospitals, not all laboratory tests were performed in all patients, which led to some missing data. Last but not the least, there was no direct evidence of renal damage caused by the virus in the urine or kidney tissue.
In conclusion, impaired renal function on admission and the occurrence of AKI during hospitalization are independent predictors of poor prognosis among hospitalized COVID-19 patients. Therefore, early and continuous monitoring of renal function and early diagnosis of AKI are necessary interventions to predict and prevent the progression of COVID-19.

Methods
Study design and participants. This retrospective, multicenter study included 3 cohorts of 1851 adult inpatients (≥ 18 years) with confirmed COVID-19 pneumonia who were hospitalized in 3 hospitals that are designated care centers for patients with emerging infectious diseases in Wuhan (Tongji Taikang Hospital, Huo Shen Shan Hospital, and Renmin Hospital of Wuhan University) between February 3, 2020 and April 10, 2020. All patients were diagnosed with COVID-19 pneumonia according to WHO interim guidance 37 . Confirmed cases denoted patients with positive findings on high-throughput sequencing or real-time reverse-transcription polymerase chain reaction assays of nasal and pharyngeal swab specimens 38 . 28 Patients without renal function tests were excluded. Participants were followed up until discharge or in-hospital demise. This study was approved by the Research Ethics Commission of each participating site. Data collection was in accordance with the review board and therefore with all valid guidelines. Written informed consent was waived by the Ethics Commissions of the designated hospitals for emerging infectious disease.
Data collection and definitions. Demographic characteristics, clinical data (symptoms, comorbidities, treatments, complications, and outcomes data), laboratory findings, and chest computed tomography (CT) findings were retrieved from electronic medical records by 4 investigators (F.X., Y.L., X.D.L, Y.Y., and Y.L). The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration Eq. 39 . AKI was defined as an increase in SCr of ≥ 26.5 µmol/L within 48 h or a 50% increase in SCr from the baseline within 7 days, according to the Kidney Disease-Improving Global Outcomes (KDIGO) criteria 40 . Baseline SCr was defined as the SCr value on admission. The date of AKI onset was defined as the earliest day on which a SCr change meeting the KDIGO criteria was recorded. The stage of AKI was determined using the peak SCr level after AKI detection, with increases of 1.5-1.9, 2.0-2.9, and ≥ 3 times the baseline value being defined as AKI stages 1, 2, and 3, respectively. All cases were diagnosed and classified according to Interim Guidelines for COVID-19 of China (6th edition) provided by the National Health Commission of China. Clinical manifestations consist of four categories, mild, moderate, severe and critical. Mild cases were defifined as: (a) mild symptoms and (b) no abnormity on chest CT. Moderate cases were defifined as: (a) mild symptoms and (b) abnormalities on chest CT. Severe cases were defifined as either: (a) respiratory rate > 30 breaths/min, or (ii) oxygen saturation 93%, or (iii) PaO2/FiO2 ratio 300 mmHg. Critical cases were defined as including one criterion as follows: shock, respiratory failure requiring mechanical ventilation, organ failure requiring admission to ICU. Acute respiratory failure was defined as a decrease in oxygen saturation (< 92%) while breathing room air with severe respiratory distress or hypoxemia (partial oxygen pressure < 60 mm Hg) and/or requirement of invasive/noninvasive mechanical ventilation. Shock and acute respiratory distress syndrome (ARDS) were defined in accordance with WHO interim guidance 41 . Acute cardiac injury was diagnosed if serum levels of cardiac biomarkers (e.g., high-sensitivity cardiac troponin I) were above the 99 th percentile upper reference limit, or if new abnormalities were observed on electrocardiography and echocardiography 42 . Acute heart failure was defined as the clinical syndrome characterized by typical symptoms (e.g., breathlessness, ankle swelling, and fatigue) that may be accompanied by signs (e.g., elevated jugular venous pressure, pulmonary crackles, and peripheral edema) caused by a structural and/or functional cardiac abnormality 42 . Sepsis was defined according to Sepsis 3.0 or SIRS (Systemic Inflammatory Response Syndrome). Patients with septic shock can be identified with a clinical construct of sepsis with persisting hypotension requiring vasopressors to maintain MAP ≥ 65 mmHg and having a serum lactate level > 2 mmol/L despite adequate volume resuscitation 43 . It should be noted that the SIRS criteria may underestimate the real incidence of sepsis. Secondary infection was diagnosed when patients showed clinical symptoms or signs of pneumonia or bacteremia, and a positive culture of a new pathogen was obtained from lower respiratory tract specimens (qualified sputum, endotracheal aspirate, or bronchoalveolar lavage fluid) or blood samples after admission 5 . Anemia was defined as a hemoglobin level of < 120 g/L for male patients or < 110 g/L for female patients. Hypoproteinemia was defined as a blood albumin level of < 35 g/L.
Statistical analysis. Categorical variables were expressed as counts and percentages, and continuous variables were expressed as medians with interquartile ranges (IQRs). Linear regression was used to perform collinearity analysis on variables related to outcomes. The Kruskal-Wallis test was applied to continuous variables, and the chi-square test and Fisher exact test were used for categorical variables, as appropriate. The composite