key: cord-1002550-jvood8ur
authors: Chan, L.; Chaudhary, K.; Saha, A.; Chauhan, K.; Vaid, A.; Baweja, M.; Campbell, K.; Chun, N.; Chung, M.; Deshpande, P.; Farouk, S. S.; Kaufman, L.; Kim, T.; Koncicki, H.; Lapsia, V.; Leisman, S.; Lu, E.; Meliambro, K.; Menon, M. C.; Rein, J. L.; Sharma, S.; Tokita, J.; Uribarri, J.; Vassalotti, J. A.; Winston, J.; Mathews, K. S.; Zhao, S.; Paranjpe, I.; Somani, S.; Richter, F.; Do, R.; Miotto, R.; Lala, A.; Kia, A.; Timsina, P.; Li, L.; Danieletto, M.; Golden, E.; Glowe, P.; Zweig, M.; Singh, M.; Freeman, R.; Chen, R.; Nestler, E.; Narula, J.; Just, A. C.; Horowitz, C.; Aberg, J.; Loos, R. J.
title: Acute Kidney Injury in Hospitalized Patients with COVID-19
date: 2020-05-08
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2020.05.04.20090944
sha: 51985c50ec377af5e522f3ea71c3a8a6055bed0f
doc_id: 1002550
cord_uid: jvood8ur

Importance: Preliminary reports indicate that acute kidney injury (AKI) is common in coronavirus disease (COVID)-19 patients and is associated with worse outcomes. AKI in hospitalized COVID-19 patients in the United States is not well-described. Objective: To provide information about frequency, outcomes and recovery associated with AKI and dialysis in hospitalized COVID-19 patients. Design: Observational, retrospective study. Setting: Admitted to hospital between February 27 and April 15, 2020. Participants: Patients aged [≥]18 years with laboratory confirmed COVID-19 Exposures: AKI (peak serum creatinine increase of 0.3 mg/dL or 50% above baseline). Main Outcomes and Measures: Frequency of AKI and dialysis requirement, AKI recovery, and adjusted odds ratios (aOR) with mortality. We also trained and tested a machine learning model for predicting dialysis requirement with independent validation. Results: A total of 3,235 hospitalized patients were diagnosed with COVID-19. AKI occurred in 1406 (46%) patients overall and 280 (20%) with AKI required renal replacement therapy. The incidence of AKI (admission plus new cases) in patients admitted to the intensive care unit was 68% (553 of 815). In the entire cohort, the proportion with stages 1, 2, and 3 AKI were 35%, 20%, 45%, respectively. In those needing intensive care, the respective proportions were 20%, 17%, 63%, and 34% received acute renal replacement therapy. Independent predictors of severe AKI were chronic kidney disease, systolic blood pressure, and potassium at baseline. In-hospital mortality in patients with AKI was 41% overall and 52% in intensive care. The aOR for mortality associated with AKI was 9.6 (95% CI 7.4-12.3) overall and 20.9 (95% CI 11.7-37.3) in patients receiving intensive care. 56% of patients with AKI who were discharged alive recovered kidney function back to baseline. The area under the curve (AUC) for the machine learned predictive model using baseline features for dialysis requirement was 0.79 in a validation test. Conclusions and Relevance: AKI is common in patients hospitalized with COVID-19, associated with worse mortality, and the majority of patients that survive do not recover kidney function. A machine-learned model using admission features had good performance for dialysis prediction and could be used for resource allocation.

Preliminary reports indicate that acute kidney injury (AKI) and kidney abnormalities seem to be associated with coronavirus disease (COVID) -19 disease severity and outcomes. A recently published study that utilized autopsy specimens from 26 patients that died of COVID-19 in China,1 demonstrated that there is evidence of the invasion of severe acute respiratory distress syndrome coronavirus 2 (SARS-CoV-2) into kidney tissue, along with significant acute tubular injury, endothelial damage, as well as glomerular and vascular changes indicative of underlying diabetic or hypertensive disease.

Small studies from China, Europe and the United States thus far have reported a wide range of the incidence of AKI, ranging between 1 and 42%. The two largest studies published to date show widely disparate rates of AKI. Guan et al. reported an incidence rate of AKI of only 0.5% in 1099 patients from 552 hospitals in China.2

The only other report of over 1000 patients with COVID-19 is from Ireland, in which 3908 patients admitted to the ICU in Ireland, Wales and Northern Ireland as of April 24, had a reported incidence of AKI requiring dialysis of 22.2%, with mortality exceeding 75%.3

New York City (NYC) has been at the epicenter of the COVID-19 pandemic in the United States and worldwide and as such has had the highest number of hospitalizations. The burden of severe AKI in the setting of COVID-19 has led to widespread shortages in NYC of dialysis nurses, machines, replacement fluids and cartridges for continuous renal replacement therapy (CRRT), and dialysis.4,5 The number of AKI cases requiring acute kidney replacement therapy (KRT) has been unprecedented in modern history.

Despite the anecdotal stories regarding the incidence and severity of AKI, the epidemiology of AKI and predictive models for dialysis requirement in hospitalized patients in the United States and particularly NYC, is not well-described. Thus, the objective of this report was to describe the incidence, severity, risk factors and outcomes associated with AKI in the setting of hospitalization of COVID-19 in a major NYC Healthcare System. All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint (which this version posted May 8, 2020. .

The Mount Sinai Health System (MSHS) serves a large, racially and ethnically diverse patient population. In this study, patient data came from the electronic health records (EHR) from five major hospitals that are part of Beth Israel was not included in this cohort as they used a different EHR.

We included patients who were at least 18 years old, had a laboratory-confirmed SARS-CoV-2 infection, and were admitted to any of the aforementioned five MSHS hospitals between February 27 and 12 a.m. on April 15, 2020 (time of data freeze). A confirmed case of COVID-19 was defined by a positive reverse transcriptase polymerase chain reaction (RT-PCR) assay of a specimen collected via nasopharyngeal swab. The Mount Sinai Institutional Review Board approved this research under a broad regulatory protocol allowing for analysis of patient-level data.

We excluded patients with known end stage kidney disease (ESKD) prior to admission, patients who were hospitalized for < 48 hours, and patients who were missing laboratory and vital signs during their hospital stay (Supplemental Figure 1 and Supplemental Table 1 ).

The dataset was obtained from different sources and aggregated by the Mount Sinai COVID Informatics Center (MSCIC); (further description of MSCIC is provided in Supplementary Material). We obtained demographics, diagnosis codes (International Classification of Diseases-9/10-Clinical Modification (ICD-9/10-CM) codes) and procedures, as well as vital signs and laboratory measurements during hospitalization.

All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint (which this version posted May 8, 2020. . https://doi.org/10.1101/2020.05.04.20090944 doi: medRxiv preprint Demographics included age, sex, and language, as well as race and ethnicity in the EHR. Racial groups included White, Black or African American, Asian, Pacific Islander, Other, and Unknown. Ethnic groups included Non-Hispanic/Latino, Hispanic/Latino, or Unknown. Vital signs and laboratory values during the first 48 hours of admissions that were obtained as part of indicated clinical care were included.

We defined a pre-existing conditions as the presence of diagnosis codes associated with specific diseases as per the Elixhauser Comorbidity Software.6

AKI, the primary endpoint, was defined by at least an increase in the peak serum creatinine of 0.3 mg/dL or 50% above baseline.7 For patients with a previous serum creatinine in the 7 to 365 days prior to admission, the most recent serum creatinine value was considered the baseline creatinine. For patients without a baseline creatinine in the 7 to 365 days prior to admission, the admission creatinine was imputed based upon a Modification of Diet in Renal Disease (MDRD) estimated glomerular filtration fraction (eGFR) of 75 ml/min/1.73m as per the Kidney Disease: Improving Global Outcomes (KDIGO) AKI guidelines. 7 We assessed in-hospital mortality defined by survival status at discharge. As a subset of patients was not discharged at the time of this manuscript preparation, subgroup analysis was performed 1) AKI with and without dialysis 2) in patients that were discharged, died, and had at least 14 days, and 3) discharged alive and those who had inhospital mortality. The need for acute dialysis was ascertained by procedure codes and was cross-referenced with nursing flow sheets. For AKI recovery, we compared the last hospital creatinine with the baseline creatinine and grouped them as recovered, or with acute kidney disease (AKD) Stages 1, 2, or 3 (Supplemental Table 2 ). 8 We used local regression (loess) for smoothening creatinine values across the interval of two weeks post hospital admission to visualize trajectories by AKD category.

Baseline characteristics were reported as medians and interquartile ranges or means and standard deviations, for continuous variables. Categorical variables were summarized as counts and percentages. Patients who All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint (which this version posted May 8, 2020. . https://doi.org/10.1101/2020.05.04.20090944 doi: medRxiv preprint were still hospitalized at the time of data freeze were regarded as having a censored length of stay (LOS).

Logistic regression models adjusted for covariates were used to estimate the adjusted odds ratio for death in patients with AKI vs. without AKI. Covariates were chosen based off of univariate testing and physician input. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

A consort diagram of included patients and outcomes is depicted in Supplementary Figure 1 .

From February 27 to April 14, 2020, 3235 COVID-19 positive patients who fulfilled our inclusion and exclusion criteria were hospitalized at one of five MSHS New York City hospitals. Baseline creatinine was available prior to admission in 1196 (37%). Patient demographics, pre-existing conditions as well as vital signs and laboratory values stratified by incident in-hospital AKI are displayed in Table 1 . Patients who developed incident AKI were older and were more likely to have hypertension, congestive heart failure (CHF), diabetes mellitus (DM), and chronic kidney disease (CKD). Patients who developed incident AKI also had higher white blood cell count, lower lymphocyte percentages, higher creatinine values, and higher systolic blood pressure. Other variables were statistically significant but likely not clinically significant given small absolute differences.

AKI occurred in 1406 patients (43%) and 280 (20%) of those with AKI required dialysis. In the 815 patients admitted to the intensive care, 553 (68%) experienced AKI. Of the 1406 patients with AKI, 40% had AKI at the time of admission. The incidence of AKI in all patients and AKI in the ICU varied by the 5 MSHS locations (Table 1 and Supplementary Table 3 ). In all admissions, the proportion with stages 1, 2, and 3 AKI were 35%, 20%, 45%, respectively. In those admitted to the ICU, the respective proportions were 20%, 17%, 63%, and 34% required acute renal replacement therapy (Figure 1) . The median peak serum creatinine was 2.2 (IQR 1.6-3.7) mg/dL in those that did not receive dialysis and was 8.6 (IQR 6.5-11.4) mg/dL in those that did receive dialysis.

In those with baseline creatinine available from the prior 7 to 365 days prior to admission, the overall incidence rate of AKI was 46%, compared to an incidence rate of 42% in those in which the admission serum creatinine was imputed to match an eGFR of 75 ml/min/1.73m2.

All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Of the patients with AKI and who had an outcome (in-hospital mortality or discharge) (n=1,124), 333 (30%) had AKI that had recovered, and 791 (70%) had AKD at the time of discharge (Figure 2) . The median creatinine at time of discharge or death was 1.9 (IQR 1.1-4.5 mg/dL). In 486 patients with AKI that survived to discharge, 275 (57%) had AKI that had recovered, and 211 (43%) had AKD at the time of discharge; the median serum creatinine at discharge was 1.2 (IQR 0.9-1.6 mg/dL). Recovery from AKI varied by if the patient received dialysis or did not receive dialysis (Figure 2) . The characteristics of patients by recovery and each AKD stage is presented in Supplemental Table 4 . Creatinine trends by recovery group are presented in Figure 3 A, B and C for differing subsets of the population.

Independent predictors of incident Stage 3 AKI in patients (n=377) who developed AKI while in the hospital (n=848) were CKD (aOR 1.8, 95% CI 1.2-2.7), higher systolic blood pressure (aOR 1.01 per mm Hg, 95% CI 1-1.02), and higher potassium levels (1.3 per 1 meq/L, 95% CI 1.07-1.6) (Supplemental Figure 2 ).

Patients with AKI were more likely to have ICU admissions, mechanical ventilation, and use vasopressors administration (Supplemental Table 5 ). In-hospital mortality in patients that experienced AKI was 45% and was 7% in patients without AKI (P<0.001). In hospital mortality of patients with AKI in the ICU (52%), and non-ICU setting (41%), was markedly higher than those without AKI (ICU 9% and non-ICU 7%). After adjustment for age, gender, race, comorbidities including hypertension, congestive heart failure, diabetes mellitus, liver disease, peripheral vascular disease, CKD, lab values including white blood cell count, leukocyte percentage, hemoglobin, hematocrit, platelets, sodium, potassium, chloride, bicarbonate, BUN, creatinine, AST, ALT, alkaline phosphatase, albumin, vitals including temperature, systolic BP, diastolic BP, heart rate, respiratory rate, oxygen saturation, the adjusted OR for death was 20.9, 95% CI 11.7-37.3 for ICU-AKI vs. no AKI and the adjusted OR for death was 9.6, 95% CI 7.4 -12.3 for all patients with AKI vs. no AKI (Supplemental Table   6 ).

All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. In the training set (n=1,317 patients), the classifier had good performance with an area under the receiver operator curve (AUROC) of 0.79 for predicting AKI requiring dialysis. Performance was similar in the testing set (n=1918) with AUROC = 0.79. (Supplemental Figure 3) . The features that had a larger impact on the model included serum creatinine, age, potassium, and heart rate ( Figure 4) . All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

In over 3200 patients with COVID-19 admitted to MSHS, AKI occurred in nearly half of patients and nearly a quarter of those patients required acute dialysis. AKI was independently associated with higher mortality. Many survivors (44%) did not recover kidney function by hospital discharge, and of all patients with AKI, only 30% survived and recovered kidney function. We also trained and tested a machine learning algorithm to predict AKI requiring dialysis with good performance.

While the current COVID-19 pandemic is caused by a coronavirus like the SARS-CoV-2 outbreak in 2005, the reported incidence of AKI associated with COVID-19 disease appears to be substantially higher.10

The incidence of AKI in this study is higher than what has been reported in China and Italy and similar to the incidence in another New York City Healthcare system (Northwell).11 There are several key differences in the MSHS cohort including higher proportions of patients with comorbidities of hypertension, diabetes, and CKD which are risk factors for developing AKI.2 In our analysis, there was no association between hypertension, diabetes with several AKI, but CKD was independently associated with severe AKI. As others have reported, patients who were admitted to the ICU had a higher incidence of AKI likely reflective of more severe disease.

Currently, the exact mechanisms of AKI due to COVID-19 are unclear. A postmortem case series reported by Su et al. found significant acute tubular injury in all their patients.1 However, in another case series, in patients with normal creatinine, urinalysis abnormalities (hematuria, proteinuria, and leukocyturia) were common, and unexpected finding in acute tubular necrosis.12 Given the high incidence of AKI and lack of full recovery on discharge, identification of potential mechanisms of COVID-19 related AKI would allow for potential interventions to reduce this devastating complication.

This study is the first study in the United States to report the persistence of kidney dysfunction (lack of recovery) in survivors of COVID-associated AKI. Over 40% of patients with AKI in the setting of COVID-19 did not recover kidney function back to baseline. Given the overall severity of AKI as evidenced clinically (high peak creatinine, need for dialysis), as well as the knowledge that many patients with COVID-19 have extensive acute tubular injury on tissue examination, as well as microthrombi, along with a high prevalence of proteinuria, it is not surprising that recovery from AKI was incomplete in most patients.1 A study in a Chinese cohort reports similar findings, with less than half of patients fully recovering kidney function.13 It remains to be seen what the All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint (which this version posted May 8, 2020. . implications for post-AKI CKD, progression of prior CKD, and ESKD are for COVID-associated AKI.14-19 This will require long-term follow-up and further investigation.

Using a small set of features derived early in the hospital admission, we were able to train a ML classifier to predict the need for AKI requiring dialysis with good performance. Several features found to have a big effect on model performance included creatinine, age, and potassium. Additional features such as lymphocyte percentage were found to have more of an impact on the ML classifier than comorbidities such as hypertension. This likely reflects the association of lymphocyte percentage and disease severity as has been previously documented in COVID-19 disease or due to the loss of the protective effects of lymphocytes.20,21

Pending additional study of the utility of the ML model, this could be used for resource allocation and planning, to prevent shortages of essential supplies.

Our study should be interpreted in light of the following limitations. Over half of patients did not have a baseline creatinine values. However, we used MDRD imputation as suggested by KDIGO guidelines, and the incidence was nearly identical for those with known baseline serum creatinine vs. those that we imputed because of missing serum creatinine prior to admission.7 Recovery was assessed at time of discharge which may not reflect patients' new baseline kidney function. As the COVID-19 pandemic is still evolving, few long term follow up lab test have been performed. We did not have sufficient data on inflammatory makers (e.g., interleukin 6, ferritin, and fibrinogen) in a large proportion of patients since they were not routinely checked early in the pandemic and a majority of patients did not have these values. 22 Additionally, urinalysis was also missing in the majority of patients and we did not include in the analysis, although further studies could benefit from characterizing the urinalysis profile of COVID-19 associated AKI. Lastly, baseline medications of interest including angiotensin converting enzyme inhibitors, angiotensin receptor blockers, statins, and other nonsteroidal anti-inflammatory drugs were not included.

In conclusion, in a diverse cohort of patients with hospitalized with COVID-19 in NYC, we described a very high incidence of AKI, severe AKI requiring dialysis, and risk of death associated with AKI. We identified several predictors associated severe AKI and found that half of patients did not recover kidney function at time of discharge. Lastly, we developed a ML classifier that had good performance for the prediction of AKI requiring dialysis. The results of this study may be useful to other centers for resource planning during the List of tables and figures: Table 1 : ICD codes used for identification of ESKD cohort. ESKD was defined as having both an ESKD diagnosis code and an ESKD procedure code. Supplemental Table 2 : Definitions of acute kidney disease Supplemental Table 3 : Incidence of AKI in patients admitted to the ICU by MSHS hospital, P<0.001) Supplemental Table 4 : Patient characteristics by kidney recovery and acute kidney disease stages. Supplemental Table 5 : Outcomes in patients with AKI and without AKI and the full cohort and in patients who were discharged, had in-hospital mortality, or are admitted with at least 7 days of follow up. Supplemental Table 6 . Adjusted and Unadjusted OR of in-hospital mortality in discharged patients and in ICU admission All rights reserved. No reuse allowed without permission.

was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Renal histopathological analysis of 26 postmortem findings of patients with COVID-19 in China

Clinical Characteristics of Coronavirus Disease 2019 in China

An Overlooked, Possibly Fatal Coronavirus Crisis: A Dire Need for Kidney Dialysis

Impending Shortages of Kidney Replacement Therapy for COVID-19 Patients

KDIGO Clinical Practice Guideline for Acute Kidney Injury

Defining Early Recovery of Acute Kidney Injury

Explainable Machine Learning Model for Predicting GI Bleed Mortality in the Intensive Care Unit

Acute renal impairment in coronavirus-associated severe acute respiratory syndrome

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

Urinalysis, but not blood biochemistry, detects the early renal-impairment in patients with COVID-19

Renal Involvement and Early Prognosis in Patients with COVID-19 Pneumonia

Rarefaction of peritubular capillaries following ischemic acute renal failure: a potential factor predisposing to progressive nephropathy

Identification of persistently altered gene expression in the kidney after functional recovery from ischemic acute renal failure

Recovery from acute renal failure predisposes hypertension and secondary renal disease in response to elevated sodium

Acute kidney injury and chronic kidney disease as interconnected syndromes

Chronic kidney disease after acute kidney injury: a systematic review and meta-analysis

Urinary Biomarkers of AKI and Mortality 3 Years after Cardiac Surgery

Lymphopenia predicts disease severity of COVID-19: a descriptive and predictive study

Complete blood count in acute kidney injury prediction: a narrative review

Clinical Characteristics of Hospitalized Covid-19 Patients

Vital signs, median (IQR) Temperature (F)

Mount Sinai Hospital located in East Harlem

To all the nurses, physicians, and providers who contributed to the care of these