Evaluation of Dynamics of Renal Recovery in COVID-19 Patients with Acute Kidney Injury Admitted to Intensive Care Unit: An Observational Study (COVID-AKI Study)

ABSTRACT

Aim and background: The prevalence of acute kidney injury (AKI) among COVID-19 patients admitted to ICU was 46%. There is a paucity of data on renal recovery in a cohort of patients with AKI. Since COVID-19 is considered a public health issue, the estimates from this study might help in prognostication and health resource management. Objective: To evaluate the predictors and dynamics of renal recovery in critically ill COVID-19 patients with AKI. To study the duration and magnitude of AKI, the proportion of patients dependent on dialysis at hospital discharge, and mortality among COVID-AKI patients. Materials and methods: A single-centre, observational study was conducted in a mixed adult ICU from March 1, 2020, to February 1, 2021. COVID-19 patients who presented with or developed AKI as per KDIGO criteria within 7 days of ICU admission were included. Baseline characteristics, hemodynamic parameters, and renal recovery kinetics were captured till the discharge of the patient. Patients were followed up till 90 days post-discharge. Logistic regression with best subset selection was performed with renal recovery as an outcome (recovery is defined as attaining AKI stage 0 by KDIGO definition or 33% reduction of serum creatine from baseline) and APACHE II, rapidity of onset and progression of AKI, the magnitude of AKI, inflammatory markers, comorbidities, and P/F ratios as predictor variables. There were no multicollinearities, influential observations. Penalized-likelihood criteria (AIC and BIC models) were applied and a model with the lowest AIC or BIC was considered as the best fit to predict nonrecovery from AKI. Results: 200 patients' data were analysed, of which 67 patients recovered from AKI. Of the 67 patients, 16, 9, and 10 patients had transient AKI (<48 hours), persistent AKI (2-7 days), and AKD (7-90 days), respectively. Dialysis was required for 136 patients. The average duration for recovery from AKI was 7.4 days. The best fit model with the lowest BIC that predicted nonrecovery from AKI were the combination of APACHE II, day onset of AKI, and magnitude of AKI. Results of logistic regression showed admission APACHE II, day onset of AKI, and magnitude of AKI were statistically significant in predicting non-recovery from AKI [OR 1.1 (p < 0.001;95% CI 1.06-1.16), OR 1.6 (p = 0.001;1.24-2.24), and OR 2.9 (p < 0.001;2.03-4.36), respectively]. This model had sufficient discrimination with AUC 0.86 and was well calibrated [Hosmer-Lemeshow (HL) chi2, p = 0.06]. Overall mortality among COVID-AKI patients was 84%. Two patients were dependent on dialysis at hospital discharge. Upon follow-up of 31 survivors for 90 days, four deaths were recorded. Conclusion: In our study, a higher APACHE II score at admission, the longer time interval between admission to the onset of AKI and the higher magnitude of AKI during ICU stay predicted poor renal recovery. A significant proportion of our patients require dialysis support and this poses a challenge on hospital resources and financial burden to the family. We observed higher mortality among COVID-19 patients with AKI compared to those with AKI not associated with COVID-19.