Background

The SARS-CoV2 pandemic has killed more than 2.7 million people as of March 2021 [1]. Infection leads to a wide clinical spectrum from asymptomatic to severe multi-organ failure and death. Kidney involvement is increasingly recognized as an important complication of SARS-CoV2 infection, resulting in proteinuria, hematuria, and acute kidney injury (AKI) [2,3,4,5]. Kidney involvement is theorized to parallel severity of disease and associated common risk factors of hypoperfusion, ischemia and nephrotoxins. However, another hypothesis for kidney sequelae is related to the virus’ affinity for the ACE2 receptor with high density in the kidney [2, 6].

SARS-CoV2-related AKI has been reported in 25–60% of those critically ill, including up to 37% of critically ill children [7,8,9,10]. AKI has been associated with worse outcomes in those with Coronavirus Disease 2019 (COVID-19), the disease caused by SARS-CoV2. Initially, adult hospitals saw a rapid rise in the need for acute dialysis during COVID-19 waves [11, 12], yet this was not seen in pediatric hospitals. Overall, children seem less susceptible to infection and severe disease, so one hypothesis proposes lower rates of AKI/dialysis needs in children is a function of disease severity. Assessment of SARS-CoV2-related AKI across the age spectrum has not previously been reported.

The purpose of this study was to evaluate the incidence and epidemiology of SARS-CoV2-related AKI across the age spectrum and determine if age is an independent risk factor for AKI development in patients hospitalized with SARS-CoV2.

Methods

Study Design & Setting

This is a secondary analysis of the observational, international, prospective Viral Infection and Respiratory Illness Universal Study (VIRUS), initiated by Society of Critical Care Medicine (SCCM) in January 2020. VIRUS seeks to ascertain a wide range of clinical and outcome characteristics of patients hospitalized with SARS-CoV2 infection. The unique aspect of this registry is it captures both critically and non-critically ill hospitalized children and adults in the same cohort facilitating comparative evaluations.

Patients included in this analysis were admitted between January 2020 and March 2021; exact admission dates are confidential and not provided to investigators. Detailed methodologies have previously been described [13]. As this was deployed as a rapid registry early in the pandemic, detailed hospital-level characteristics are not available to investigators. Briefly, 298 centers from 26 countries contribute comprehensive pediatric and adult data from hospitalized patients encompassing intensive care units (ICUs) and non-ICUs. Ethical oversight was obtained at each local center and de-identified data stored in REDCap [14].

Patient population

We evaluated all participants in the registry if they had PCR- or antibody-confirmed presence of SARS-CoV2 infection, complete age and 28-day hospital outcome data, and at least one serum creatinine value. We excluded patients with clinical suspicion but no laboratory confirmation of SARS-CoV2, current pregnancy, chronic dialysis, or chronic kidney disease (CKD) stage 5.

Potential Bias

As this is an ongoing cohort registry, rapidly deployed during an evolving global pandemic, analyses were conducted by complete case analysis methods which could introduce some biases towards the more severe cases or because of imminent deaths. Nevertheless, the major exclusions were those without creatinine values or missing 28-day hospital outcomes as we assumed these patients to have the least complete data entry and highest risk for potential data entry errors.

Outcomes

The primary outcome of interest was AKI development as defined by Kidney Disease Improving Global Outcomes (KDIGO) serum creatinine-only criteria within the first 7 days of hospitalization [15]. AKI is defined as a rise in serum creatinine ≥0.3 mg/dL or > 50% from baseline. Urine output is considered part of the KDIGO AKI definition, but the registry data was determined to be insufficient as > 60% of our cohort was missing urine output values. We also further stratified AKI into stages and receipt of dialysis. Additional outcomes of interest included hospital mortality, hospital and ICU length of stay (LOS), and hospital-related complications.

The registry did not capture baseline creatinine (Crb) values (prior to hospitalization). It is therefore standard practice to estimate Crb [15,16,17]. However, the estimation of Crb is not standardized across the age spectrum. Using KDIGO guidelines for adults (≥18 years), we estimated a Crb by assuming an eGFR of 75 ml/min/1.73m2 and back calculating a creatinine with the modification of diet in renal disease (MDRD) equation [15]. No standard international guideline for estimating a Crb in children exists. We used the validated method of assuming eGFR of 120 ml/min/1.73m2 for children 2–17 years and median normative-based eGFR-for-age in children < 2 years and back calculating creatinine with the height-independent equation [18,19,20]. For patients with CKD, we used the minimum serum creatinine within the first 7 days of hospitalization as Crb estimation.

Though these are standard assumptions in AKI research in their respective fields of adult and pediatric nephrology [15,16,17], there is no standard acceptance of estimating Crb in the transition period from adolescents to adulthood. Therefore, given the lack of standardization for estimating Crb across the age spectrum, we conducted two sensitivity analyses: [1] using the full age spectrum (FAS) equation for both adults and children that does not assume a fixed eGFR by age but instead changes across the age spectrum to overcome this limitation [21] and [2] using the minimum serum creatinine as an assumed Crb for all patients. The FAS equation is limited as it has only been validated in Caucasian populations. The assumption of minimum creatinine as a baseline is limited as it assumes all patients return to their baseline within 7 days of hospitalization. In addition, we conducted a sensitivity analysis where race was removed from the MDRD calculation for adults [22].

Exposure

Primary exposure of interest was age; it was entered as years and months (children< 5 years), years (participants 5–90 years), and limited to ‘> 90’ for those > 90 years of age for privacy. For analysis, those > 90 were classified as 95 years. Age was evaluated as a continuous variable by years and categorical variable by 5-year and 20-year age increments to explore potential non-linear associations.

Additional variables

As this was an exploratory analysis, we included a variety of additional demographic, pre-hospital, and hospital-related variables from the registry. Sex and race/ethnicity were categorical. The registry de-identified center location except whether the center was in the United States or elsewhere. CDC classifications were used for weight categorization (underweight, normal weight, overweight, obese, severely obese) using BMI data for adults ≥18 years, BMI percentiles for children 2–17 years, and weight-for-height percentiles for children < 2 years [23]. CDC does not provide pediatric classification for severely obese, so those are grouped with obese for those < 18 years. SARS-CoV2 testing was determined by local centers. Other clinical data captured included comorbidities and recent pre-hospital medications as well as inpatient medications within the first 7 days. Comorbidities, including CKD, were determined by medical chart review by local investigators.

Illness severity was categorized by variables that span the age spectrum. Severe illness was defined as a composite of received invasive mechanical ventilation, vasopressor(s) and/or inotrope(s), and/or extracorporeal membrane oxygenation (ECMO). Moderate illness was defined by ICU admission without any organ support therapies listed above. Mild illness was defined as hospitalization but without an ICU admission nor organ support therapies as listed above. As some of these therapies may be clustered within centers, we accounted for this potential in our analyses described below. More traditional markers of illness severity were captured but do not translate across pediatric and adult patients so are not the primary marker assessed in this analysis (e.g., sequential organ failure assessment (SOFA) scores for adults and pediatric risk of mortality (PRISM) scores for children).

Statistical analyses

Descriptive statistics compared demographic, pre-hospital and inpatient clinical characteristics within the first 7 days of hospitalization among those with and without AKI. Wilcoxon rank sum tests and chi-square tests were used for continuous and categorical variables, respectively. Given the large sample size which leads to highly significant p-values, Cohen’s effect size estimates were calculated for continuous variables to better express the magnitude of differences (small effect 0.1–0.3, medium effect 0.3–0.6, large effect > 0.6). Univariate risk differences (RD), odds ratios (OR), and 95% confidence intervals (CIs) were calculated for hospital mortality by AKI stage. To account for common clinical practices, clustering within centers was used via generalized estimating equations (GEE) with logistic regression models to determine if age is an independent risk factor for the development of AKI in SARS-CoV2-related hospitalizations, with adjustments for the potential confounding of sex, hypertension, diabetes mellitus, cancer, CKD, race/ethnicity, and severity of illness as defined above. Determination for potential confounders to include in models were determined by a priori clinical knowledge and directed acyclic graphs. Significance was set at an alpha-level of 0.05. Sensitivity analyses were conducted using different equations for estimating a Crb and stratifications by comorbidities and whether center was U.S.-based. All analyses were conducted in SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina).

Results

Demographics

6874 patients from 142 centers met inclusion criteria (Fig. 1). 28% of participants were from non-U.S. centers (Table 1). A total of 39.6% (n = 2719) developed AKI within the first 7 days of hospitalization; this was significantly higher among patients in ICUs (1926/4075, 47.3%) compared to non-ICUs (793/2799, 28.3%), p-value< 0.0001. Almost 60% of the cohort were admitted to the ICU (n = 4075). The median age was 60 years (range 0–95 years) and 9.0% (n = 621) were < 20 years (Table 1). Those with AKI were more likely to be older (median age 65 years) than those without AKI (median age 55 years), p-value< 0.0001 and effect size 0.45, and more likely to have comorbidities (median 3 versus 1 in those without AKI), p-value< 0.0001 and effect size 0.37. Among those < 20 years, 28% (171/621) developed AKI. Differences in AKI risk based on race/ethnicity (p-value< 0.0001) were noted. Supplementary Table 1 includes hospital-related associations with AKI.

Fig. 1
figure 1

Participant Inclusion Flow Diagram by STROBE Reporting Guidelines

Full size image
Table 1 Demographics of Participants in VIRUS Registry by AKI status
Full size table

Comparing patients excluded to those included revealed no significant difference by age, sex, or location of center (i.e., U.S.-based). However, those excluded were more likely to have no comorbidities (26%) compared to those included (20%), and only 31% of the excluded group were admitted to the ICU (compared to 59% in this analysis). As expected, those missing creatinine values were often missing other key variables; BMI data missing for 50% of excluded patients compared to 24% of patients in this analysis.

Hospital complications

Among participants with AKI (n = 2719), 64% had Stage 1, 14% Stage 2, 19% Stage 3 without dialysis, and 4% Stage 3 with dialysis (Table 2). Of the patients requiring dialysis, the median duration was 5 days (IQR 2.4–12.4) ranging from 0.2–31.8 days (duration missing for 25/104 patients). Only 7% (n = 7) of those who received dialysis in the first week were from non-U.S. centers. Across AKI stages, there was a significant increase in hospital and ICU LOS (effect sizes 0.48 and 0.38, respectively), with the greatest increase being among those receiving dialysis; hospital LOS median 31 days (IQR 22–48) for those on dialysis compared to median 6 days (IQR 4–11) for those with no AKI (p-values all < 0.0001). Significant differences across AKI stages were also seen for intubation, new home oxygen requirement on discharge, vasopressor(s)/inotrope(s) use, development of thromboses, and inpatient mortality. The absolute risk of hospital mortality increased significantly (p-values< 0.0001) for each AKI stage compared to no AKI. Overall, the OR of hospital mortality in those with AKI compared to those without AKI was 4.0 (95% CI 3.5–4.5). These associations did not change significantly when alternative Crb estimators were used.

Table 2 Hospital Complications by AKI Stages for Patients Admitted with SARS-CoV2 Infection
Full size table

Association of age with AKI risk

Figure 2 depicts a bimodal distribution of AKI risk by age with those of young adolescence (10–15 years) having a higher risk than both very young children (< 5 years) and older adolescents/young adults (15–35 years), while those over age 65 years also have a high risk of AKI. Even after adjusting for potential confounders (sex, pre-existing hypertension, diabetes mellitus, cancer, CKD, race/ethnicity, and severity of illness) there remains increased risk of AKI in a bimodal distribution (odds ratio inset in Fig. 2). This pattern of AKI distribution did not change when using alternative Crb estimators (Supplementary Figs. 1, 2, 3 and 4), including the full-age spectrum equation. The pattern of AKI distribution held when evaluating those with no comorbidities versus those with comorbidities (Fig. 3) and again when we evaluated only those in the United States (Supplementary Fig. 5). Table 3 depicts a snapshot of representative age ranges and their adjusted OR of developing AKI in these different scenarios, i.e., by different Crb estimators and in a population with no pre-existing comorbidities. The data consistently shows an almost 2.5-fold increased odds of developing SARS-CoV2-related AKI for 10–15-year-olds and for 70–75-year-olds when compared to young adults (30–35 years old).

Fig. 2
figure 2

Age Distribution of Hospitalized Patients with SARS-CoV2 who Experienced AKI within First 7 days of Hospitalization. Main figure presents percentage per age bracket who developed acute kidney injury (AKI) among all hospitalized patients and further stratified by severity of illness status. Severe illness is defined as a composite indicator of invasive ventilation, use of vasopressor(s)/inotrope(s), and/or use of extracorporeal membrane oxygenation. Moderate illness is defined as admitted to an intensive care unit but without use of above organ support measures. Mild illness is defined as patient required hospitalization but not in an intensive care unit and without use of above organ support measures. Insert presents the adjusted odds ratio (OR) with 95% confidence intervals (CI) of developing AKI within the first week of hospitalization by age bracket compared to young adults (30–35-year-olds) as the referent category. Adjusted for sex, pre-existing hypertension, diabetes mellitus, cancer, chronic kidney disease, race/ethnicity, and severity of illness. AKI defined per KDIGO guidelines

Full size image
Fig. 3
figure 3

Age Distribution of Hospitalized Patients with SARS-CoV2 who Experienced AKI within First 7 days of Hospitalization Stratified by Presence or Absence of Comorbidities. Presents percentage of hospitalized patients who developed acute kidney injury (AKI) among all hospitalized patients and further stratified by presence of any comorbidity versus no pre-existing comorbidities. AKI defined per KDIGO guidelines

Full size image
Table 3 Adjusted Odds Ratios of Developing AKI by Different Definitions/Populations
Full size table

Discussion

In a large and diverse cohort evaluating AKI in COVID-19, we found a high incidence of AKI (39.6%) and that it varies across the age spectrum with a bimodal distribution. Given our cohort’s wide age span, we demonstrate a more nuanced view of SARS-CoV2-related AKI than previous evaluations. In every context of our evaluations, there was consistently a bimodal age distribution of AKI risk with the older population and early adolescent (10–15 years) population at higher risk compared to the young adult populations. This is an interesting phenomenon as to date there are only descriptions of a linear relationship between age and COVID-19 severity and its complications [24, 25]. Other known risk factors for AKI were seen in this cohort, such as sex, pre-existing comorbidities (i.e., hypertension, diabetes mellitus, cancer), and race/ethnicity. However, even after controlling for these potential confounders, there remained an association producing a bimodal age distribution in AKI risk; a 10–15-year-old had a similar odds of AKI as a 70–75-year-old (compared to 30–35-year-olds). The bimodal distribution also persisted after controlling for severity of illness and within-center correlations, which suggests something additional is contributing to the AKI risk. This contradicts an early study on SARS-CoV2-related AKI that found illness severity to be the key risk factor for SARS-CoV2-related AKI, but it was a small study (n = 223) with results from the early waves (March–June 2020), and excluded children [26]. Interestingly this bimodal distribution differs from previous non-SARS-CoV2 AKI literature which suggests a U-shaped distribution (peaks in infancy and older adults) [27,28,29].

The differences in AKI risks across the age spectrum found here were not explained by different Crb estimators. KDIGO is a standard guideline for defining AKI, yet it lacks a standard method for estimating a Crb in children when one is not known. We therefore evaluated variety of Crb estimators in pediatric and adult populations. Yet, a bimodal distribution of AKI risk by age remained even with several sensitivity analyses, including a Crb estimator (FAS) validated across the age spectrum of 2–90 years. The FAS equation assumes a slow transitional change in eGFR from childhood into adulthood [18,19,20]. In addition, the bimodal age distribution of SARS-CoV2-related AKI development persisted when evaluating only hospitalized patients with no pre-existing comorbidities or evaluating only U.S.-based centers, suggesting that comorbidity differences nor center or country specifics do not explain the bimodal pattern.

The persistence of the bimodal pattern by age, despite multiple iterative analyses, suggests there may be something unique about SARS-CoV2 and its relationship with AKI. One could hypothesize that the propensity of the SARS-CoV2 virus to attack the endothelium could also contribute to the differences seen in the older population and their risk with AKI beyond illness severity [30], though it does not explain the higher risk in early adolescence. There may be a hormonal influence in early adolescence that makes the endothelium more prone to injury compared to younger adults, but this would not fully explain the higher AKI rates in the elderly. We postulate that the bimodal AKI distribution could perhaps be a combination of SARS-CoV2-related vasculopathy and hormonal influences. There may also be yet unknown biological mechanisms that are contributing to this bimodal pattern. For example, we could not account for the different strains or clinical spectrum of SARS-CoV2 presentations which may be an important driver of the bimodal age pattern. A recent report of 2600 hospitalized adults with SARS-CoV2 infection found similarly that high AKI rates are not fully explained by known risk factors and need further exploration [31]. Fully understanding the bimodal age distribution of SARS-CoV2-related AKI risk is even more important now as countries are seeing a shift in age distribution of SARS-CoV2 infections as children are not yet eligible worldwide for vaccinations and new variants may disproportionately affect younger populations. Further in-depth epidemiological studies and animal models may be needed to understand the biological mechanisms underpinning the age distribution in SARS-CoV2-related AKI.

Similar to other studies [7,8,9, 32], this cohort demonstrates a high rate of AKI in COVID-19 patients; among ICU patients the AKI rate was 47.3% and in non-critically ill patients was 28.3%. Only a few studies report SARS-CoV2-related AKI rates outside of ICUs [33], and our results suggest a high-percentage of non-critically ill patients are at risk.

Other literature has found that SARS-CoV2-related AKI has an increased risk of mortality [4, 8, 9, 34]. In addition to this, we report a strong relationship with mortality and other hospital complications that is proportional to AKI’s severity and seen even in non-critically ill patients and those with mild increases in serum creatinine (≥0.3 mg/dL). Very few reports thus far have explored the complications associated with the varying degrees of AKI severity [2, 10]. This is important as even the slightest degree of AKI may be associated with long-term morbidity and mortality among those hospitalized with SARS-CoV2. Interestingly, though young adolescents had higher risks of AKI compared to middle adulthood, the rates of dialysis were higher in middle adulthood (20–40 years) compared to children (< 20 years). These may be related to center practice differences or the overall small sample of dialysis needs in both of these groups in this cohort (n = 7 for 20–40 year-olds and n = 1 for < 20 year-olds).

Limitations

The VIRUS registry has been a real-time assessment of the COVID-19 pandemic, so we may have introduced bias by excluding participants missing data. However, the large sample size provides real-time insight to ongoing trends and allows comparisons across the ages. Comparing the cohort of those with and without creatinine values, we found that we likely had some selection bias toward sicker patients; however, 40% of our participants were never in the ICU. A limitation of evaluating AKI across the age spectrum is the lack of standard Crb estimators, but our results were similar when using multiple estimators, suggesting there is a true phenomenon of bimodal age distribution in SARS-CoV2-related AKI that deserves further exploration. The registry includes multiple centers and as such risks introducing bias through practice pattern differences between pediatric versus adult centers and regional variations, but we controlled for this in our analyses by accounting for clustering within centers. However, evaluating data from across multiple regions and centers allows a broader view of the epidemiology of SARS-CoV2-related AKI, which is needed to plan for more in-depth case-control or randomized clinical trials evaluating different management and treatment strategies for improved outcomes in SARS-CoV2-related AKI.

Conclusions

Patients hospitalized with SARS-CoV2 have a high risk of AKI, irrespective of illness severity. We demonstrate an interesting phenomenon of a bimodal age distribution of SARS-CoV2-related AKI risk – high in the elderly and early adolescence – that deserves more in-depth exploration as it was not explained by pre-existing comorbidities, illness severity, eGFR equations, or clustering within centers. Our study reiterates other findings that SARS-CoV2-related AKI at any stage increases patients’ morbidity and mortality. However, as the pandemic lingers, outbreaks will continue, and while younger children remain unvaccinated, it is even more important to understand if there are biological reasons or other unexplored risk factors behind this bimodal age distribution of AKI risk that may guide clinical care improvements in the management of SARS-CoV2 infections and/or provide insights into the pathophysiology of this unique virus.