The dynamics and outcomes of AKI progression during the COVID‐19 pandemic

Acute kidney injury (AKI) associated with COVID‐19 is associated with poor prognosis. This study assessed the hitherto uninvestigated impact of COVID‐19 on the progression and clinical outcomes of patients with AKI.

acute kidney injury dynamics, competing risks and multi-state models, COVID-19, death, metaanalyses Summary at a glance Data from patients with AKI were compared based on COVID status.Patients with COVID demonstrated a higher case fatality ratio in AKI Stage-3 patients.COVID positive patients also demonstrated more rapid progression of AKI and prolonged hospital stays compared with COVID negative patients.

| INTRODUCTION
Acute kidney injury (AKI) is a syndrome characterized by an abrupt decline in kidney function, encompassing structural damage to nephrons and loss of renal function. 1The consequences of AKI can be severe, with its presence portending poor outcomes, including morbidity and mortality.It is multifactorial, with sepsis, ischemia, and nephrotoxicity often being confounding risk factors.The diagnostic approach to AKI is based on an acute decrease in the glomerular filtration rate (GFR), as reflected by an acute rise in the surrogate marker, serum creatinine (sCr) levels, and/or a decline in urine output (UO) over a given time interval, as defined by the Kidney Disease Improving Global Outcomes (KDIGO) guidelines.Severe acute respiratory syndrome coronavirus-2 (SARS-COV-2) is responsible for Coronavirus 2019 disease (COVID- 19), which is associated with multiorgan failure, including AKI, despite early data suggesting otherwise. 2subsequent systematic review and meta-analysis, including data from 13 137 patients from 20 studies, demonstrated a prevalence of AKI stages 1-3 among COVID-19 positive (COVID+) patients of 17% (range 0.5%-80.3%).3 Approximately 5% of all patients required kidney replacement therapy (KRT) (range 0.8%-14.7%).The overall mortality rate among COVID+ patients with AKI was 52%, with a reported range of 7%-100%.These results show a great variability in the incidence of AKI within the context of COVID-19.
[6][7] Potential risk factors responsible for AKI evolution include demographics (age, sex, ethnicity, and body mass index) and associated comorbidities (diabetes mellitus, hypertension, pre-existing cardiovascular disease, and underlying kidney disease).Hospital outcomes are also influenced by disease management and control, such as admission to intensive care unit (ICU), ventilation strategies, nephrotoxic drug administration, and disease severity.COVID-19 acquisition during the pandemic added a layer of complexity in planning, particularly during the first wave in early 2020, in no small part due to a lack of vaccines and consensus guidelines.
A number of unanswered questions have been highlighted regarding AKI in COVID+ patients; these have been broken down into epidemiological, pathophysiological, and treatment questions.
Areas of interest for research include the role of biomarkers, identification of risk factors for non-recovery, the relative contribution of each of the myriad of associated pathophysiological mechanisms (e.g., haemodynamic factors, direct viral infection, microthrombi, etc.), and whether any disease-specific treatments materially affect renal outcomes. 8e current literature does not explore disease dynamics with reference to AKI and COVID-19, with a particular paucity of data on the progression of AKI associated with COVID-19.The COVID-19 pandemic has presented one of the greatest challenges to health-care systems worldwide owing to its rapid development.Disease progression and patient recovery from disease processes routinely exert a considerable influence on the availability of health-care resources.
This was particularly magnified during the pandemic, with the additional burden of increased staff sickness due to COVID-19, leading to a significant detrimental impact on stretched workforces. 9ICU beds are at a premium for cost, staffing, training, consumables (personal protective equipment), and other logistical and technical reasons.The scarcity of resources is likely to have had some impact on clinicians' decision-making regarding the focus of care, including opting for resuscitation or palliation. 10comfortable ethical questions have been raised regarding how best to assign resources in extreme circumstances, including whom and how to prioritise care. 11Novel approaches to AKI have been used with some success under testing conditions, such as acute peritoneal dialysis. 12Although such approaches are resourceful and commendable, the ideal approach would be to offer uncompromised optimal tailored care to all patients.
The speed and probability of progression between AKI stages, as well as the time to recovery from illness, are critically relevant to inform planning, at least from the most common scenarios during inhospital stay.
This study investigated the epidemiology of COVID-19 in patients with AKI associated with progression dynamics and death in a cohort of patients admitted to a large teaching hospital during the first wave of the COVID-19 pandemic.These data allowed increasingly complex analyses conducted on hospital outcomes, enabling insights into the dynamics of AKI progression and hospital outcome (death or discharge), stratified by patients' COVID-19 status.
A prognostic model for hospital outcomes was constructed in response to the main criticisms of previous models; in >80% of papers, patients in the COVID-19 control group were unconfirmed or even untested.Given the high prevalence of asymptomatic disease in many populations, including hospital inpatients, they cannot act as true negative control. 13All patients in this cohort were investigated using polymerase chain reaction (PCR) testing for COVID-19, the gold standard test for detecting active infection.
Additionally, a systematic review and meta-analyses of publication data collected during the same period are presented, aimed at deriving pooled estimates for the prevalence of AKI and case-fatality ratios among COVID+ patients, and to understand potential sources of heterogeneity.The latter was based on further information derived from these papers, including the geographical origin of the studies, age, gender distribution, ethnicity, and the length of the data collection period for each study.

| MATERIALS AND METHODS
Data were collected for all patients aged over 18 years admitted to St. George's Hospital, London, UK, between 13/03/20 and 13/05/20, where an AKI alert in serum creatinine was detected by the computer database system, on admission, and during their hospital stay.All patients with AKI stages 1-3 were identified by degree of change in serum creatinine levels compared with baseline creatinine values, defined by KDIGO AKI guidelines 1 and tested for COVID-19 using PCR.Patients' baseline serum creatinine levels were approximated from the historical creatinine results pre-dated for at least 12 months.
Survival-type statistical methodologies were employed to learn about the dynamics of disease progression and hospital outcomes in these patients, that is, death or discharge.A competing risk model was employed for hospital outcomes, an extension of the survival analysis paradigm, intrinsically accounting for the length of time to hospital discharge or death.Associations between death and available variables were evaluated using sub-hazard ratios (SHR).A SHR greater/smaller than 1 indicated a harmful/protective association, and p-values less than .05indicated a significant association.The uncertainty of the estimates is expressed as 95% confidence intervals (CIs).
The final multi-state modelling approach used the peak AKI stage as an intermediate state between AKI at admission and the hospital outcome (death or discharge) and allowed quantification of AKI progression dynamics and the associated impact of COVID-19.
A prognostic model was constructed for the hospital outcomes using competing risk models, and time-varying measures of discrimination between death and discharge were evaluated through Receiving Operator Curve analyses and the associated area under the curve statistic (AUC, which is a probability).The closer this probability is to 1, the better the model discriminates between patients who will have poorer or better hospital outcomes.The AUC was evaluated at a series of time points representing important percentiles of the distribution of the length of hospital stay.
A systematic review and a series of meta-analyses were performed to derive a potential pooled estimate for the risk of COVID-19 in patients with AKI, while meta-regression techniques aimed to understand the sources of discrepancies between studies.The literature has been used to investigate a series of outcomes relevant to this research: the risk of AKI in COVID+ patients, the case fatality ratio in COVID+ patients with AKI, the relative risk of death for AKI versus non-AKI in COVID+ patients, and the relative risk of death for COVID+ versus COVID-patients when available.The studies were grouped based on population demographics, including geography/ ethnicity (as in China, USA, Europe), predominant gender and age, and, in the absence of clear guidelines for disease management at the time, grouping by length of study and disease severity (requirement for intensive care) were considered.Further details on the data collection, statistical methods, and analytical strategies are outlined in the Supporting Information.

| Summary statistics
Of the 953 cases of AKI Stage 1-3 identified, 576 satisfied the inclusion criteria (Figure S1).The summary statistics of all variables used in the analyses are presented in Table 1 were admitted to the ICU, of whom 130 (76%) tested COVID+ and 68 (40%) required KRT.

Data regarding AKI progression during admission are presented in
Table S1 and Figure S2.Overall, 16.8% of patients with AKI Stage-1 The effect of the available variables on the sub-hazard ratios death versus hospital discharge -adjusted for both COVID status and acute kidney injury (AKI) peak stage.The effect of age on death decreases across peak AKI stage as indicated by the interaction between age and AKI peak stage.(Figure 1).Figures 2 and 3 and Table S2 depict the daily dynamics and cumulative incidence of hospital outcomes stratified by COVID-19 status (Figure 2) and AKI peak stage (Figure 3).Among COVID+ patients, those with AKI Stage-3 displayed either the fastest increase in the daily cumulative incidence of death or the longest hospital stay (Figure 3).A series of predictions based on these curves are presented in Table S3.For example, the probability of transition from AKI Stage-1 to Stage-3 on day 7 was 6% (3%, 9%) among COVID-patients.This  S3 and Figure 4.
The prognostic model (Table 2) is based on the most parsimonious multivariable competing risk model, with retained age, COVID-19 status, AKI peak stage, and creatinine as the strongest predictors.The model also retained a significant interaction between creatinine and AKI-peak stages, indicating that creatinine levels associate with death differently across AKI-peak stages; the risk of death increases faster with creatinine levels in AKI-1 patients than in AKI-3 patients.The discriminatory power of this model based on time-varying ROC analysis results in AUC = 0.76 (0.71-0.81) on the 29th day and AUC = 0.77 (0.72-0.82) on the 40th day, by which 95% of the patients either died or were discharged from the hospital.The AUC did not change beyond this time point (Table S4).

| Meta-analyses
A total of 48 articles fulfilled our criteria to be included in Systematic Review and meta-analyses (Figure S3).Overall risk of AKI in COVID+ patients was estimated at 0.30 (0.24, 0.37), but exhibited values between 0.005 and 0.9 in studies, resulting an I 2 value of 99.65% highlighting significant heterogeneity between studies (supplementary meta-analyses Figure 1).Further analysis suggested differences in geographical regions with highest incidence in American studies and lowest incidence in Chinese studies (p < .001),and tendency towards male gender dominance (p < .001),however, variability within regions remained high (I 2 > 90%) (supplementary meta-analyses Figures 2 and 4).Case-fatality ratio of COVID + patients ranged from 1% to 52% (supplementary meta-analyses Figure S11) with large heterogeneity within different geographical regions (supplementary meta-analyses Figure S12).
Further comprehensive analysis of meta-analyses is detailed in the Supporting Information.

| DISCUSSION
This study investigated a large cohort of patients with AKI using statistical techniques (Supporting Information) and demonstrated consistent information regarding the dynamics of hospital outcomes and AKI progression.The current KDIGO guidelines divide AKI into stages 1, 2, or 3 based on specific urine output and/or increases in the serum creatinine level and/or the requirement for KRT. 14While it is broadly accepted that each stage represents a stepwise negative prognostic milestone, there is a paucity of data quantifying how patients progress between stages.
In these data analyses, we sought to assess AKI progression in a novel fashion, thus moving away from estimating sample proportions, but also estimating the transition dynamics to more severe stages in patients with AKI.By subdividing this cohort into COVID-19 positive and negative groups, these data further benefited from the possibility of isolating and quantifying the effect of the first wave of the COVID-19 pandemic on AKI progression.
The UK Renal Registry published a national AKI audit report on the nationwide collection of AKI warning test scores from 2018. 15From this report, 18% of patients with AKI episode died within 30 days of first alert in 2018; this figure is well consistent with our estimate and its 95% confidence interval of 16% (12%-20%).Furthermore, in the same document, it has been stated that the risk of death within 30 days from hospital admission reported by the UK Renal Registry also increased with peak AKI stage: 13% for AKI-1, 29% for AKI-2 and 33% for AKI-.All these results are consistent with case-fatality within  This could represent a paradigm shift from a largely qualitative approach to a more quantitative and thus more scientific methodology.
Our approach demonstrated considerable differences in the severity and progression of AKI (as well as other patient-centred metrics) in those who were COVID+ compared with those who were negative.The devastation from what was initially considered to be a primary respiratory illness has subsequently been shown to have Insight into the dynamics of death and discharge in acute kidney injury patients stratified by COVID status.Associated estimates with these figures are given in the first line of Table S2.
F I G U R E 3 Insight into the dynamics of death and discharge in acute kidney injury (AKI) patients further stratified by COVID status and AKI stage.Associated estimates with these figures are given in the last three lines of Table S2.
immediate and longer-lasting implications for multiple organ systems in the absence of preventive measures or clear clinical management guidelines.Patient-centred outcomes, such as length of stay and survival, are particularly useful for planning health-care provision and resource allocation.This delineation was admittedly from a prevaccination cohort, and although outcomes have greatly improved following widespread vaccination 16  There was a disproportionate burden of severe AKI and severe illness exacerbated by COVID-19, as evidenced by high levels of inflammatory markers, ICU admission, and requirement for intubation and KRT.4][25][26] There was no evidence to suggest an association between patients with hypertension, cardiac comorbidities, and COVID-19, in line with the current research. 19,24,27Even large multicentre studies failed to show sufficient statistical evidence that hypertension may have an impact on outcomes following COVID-19 infection 28 and meta-analyses have also shown significant heterogeneity with regard to cardiovascular risk factors and complications among COVID+ patients. 29ese analyses did not benefit from the date of COVID-19 swab or assessment; hence, it is impossible to disentangle which event occurred first, that is, COVID infection or AKI progression.Of 48 eligible studies worldwide, in a mixture of hospital patients of different degrees of severity, attempts to derive valid pooled estimates for outcomes were hampered by a high degree of heterogeneity between these studies (I 2 > 95%) present in all analyses (Supporting Information).
Three studies 26,30,31 compared the outcomes of patients with AKI in both COVID+ and COVID-patients, allowing more detailed comparisons to be made.All three studies 26,30,31 identified male sex, non-Caucasian ethnicity, admission to the ICU for mechanical ventilation, and/or KRT as common factors among AKI patients who were COVID+ compared with COVID-patients, in agreement with this study.Other risk factors associated with COVID+ status among AKI patients include diabetes mellitus, 26,31 obesity, 26 CKD, 26 and older age. 26All three studies reported higher casefatality in patients who were COVID+ compared with COVID-.
Kohle et al. 30 found case-fatality of patients with AKI and COVID+ to be 60.5%, comparable to 52% we obtained from our cohort.
These are much higher than reported by Fisher et al. 26 (33.7%) and Moledina et al. 31 (29.6%).These differences could be attributed to differences in patient management, as these studies were conducted during a time period prior to vaccination and where there were no standardised guidelines for management of COVID-19.It is also difficult to specifically quantify the exact contribution of AKI stage to death in the presence of many other contributing risk factors such as age and co-morbidities.To our knowledge, no studies specifically analysing the extent to which mortality in this population is attributable to AKI and/or its stages Nevertheless, the evidence strongly supports that presence of AKI in the context of COVID-19 is associated with poorer prognosis.
This study investigated the hospital outcomes and dynamics of AKI during the first wave of the COVID-19 pandemic.These data demonstrated a more rapid progression to severe forms of AKI or death, higher case fatality ratios, and longer lengths of stay among COVID+ patients who suffered AKI compared with COVID-patients.This study highlighted a several-fold increased requirement for dialysis, use of general and high-dependency hospital beds, and enhanced use of specialized staff to manage COVID+ AKI patients.This will be particularly pertinent when considering the potential impact of novel COVID-19 variants, especially among unvaccinated patients with AKI associated with similar infections in the future, and will aid tailored planning for this vulnerable patient group.
the COVID-group of our cohort stratified by AKI peak stage, and within or very close range of their 95% confidence intervals.From a clinician's perspective, these data highlight that progression between AKI stages during the first wave could be exacerbated by COVID-19 infection, with severe or fatal consequences.While clinicians may look at trends in creatinine, we believe that our statistical modelling offers insight into predicting potential deterioration, with greater accuracy, based on evidence informed by the available data.
It was also impossible to assess whether the patients attended the hospital with COVID-19 infection or acquired it during their hospital stay.For example, if the date of the PCR results was known, the patients would have been classified as COVID+ (within 48 h after admission) or acquired during hospitalization (COVID+ after 48 h).The dates would have been irrelevant to the onset of infection but would have offered further insights into the intertwined dynamics between AKI and COVID-19. .
The estimated acute kidney injury (AKI) transition probabilities describing the dynamics of AKI evolution during the first 40 hospital days stay from admission stratified by COVID status.Transition probabilities values on day 7, 14 and peak values and the day since admission they have been achieved are given in TableS3.
Three increasingly complex statistical techniques (Supporting Information) investigating COVID-19 infection and hospital outcomes provided consistent results and yet increasingly detailed information regarding the dynamics of hospital outcomes and AKI progression to more severe stages and/or death.We only used AKI upon admission, AKI peak stage, and hospital outcome, but this approach canF I G U R E 4