Recognition and management of acute kidney injury in children: The ISN 0by25 Global Snapshot study

Background In low and middle-income countries, reliable data on the epidemiology of childhood acute kidney injury (AKI) is lacking. The Global Snapshot, conducted by the ISN “0by25” AKI initiative, was a world-wide cross-sectional, observational study to evaluate AKI in hospitalized patients. Here we report the pediatric results of this study. Patients and methods We prospectively collected data on children who met the Kidney Disease Improving Global Outcomes AKI criteria during a 10-week window in late 2014. AKI risk factors, etiological factors, management and outcomes were recorded using standardized forms and protocols. Countries were classified according to their 2014 gross national income (GNI) per person into high-income countries (HIC), upper-middle income countries (UMIC) and low and low-middle income countries (LLMIC). Need for renal replacement therapy, mortality, and renal recovery were assessed 7 days after AKI diagnosis or at hospital discharge, whichever came first. Results 92 centers from 41 countries collected data on 354 pediatric AKI patients; 53% of the children developed AKI while hospitalized and 47% in the community. The most common etiological factors for AKI differed across GNI categories as well as between patients with community-acquired vs. hospital-acquired AKI. Children from HIC were younger, and larger proportion of AKI in this group were due to post-surgical complications vs. other etiologies when compared to other income categories. In patients with hypotension as the cause of AKI, the adjusted risk of death was almost 10-fold higher compared to patients without hypotension as an etiological factor for AKI development. Mortality was similar within AKI stages in HIC and UMIC. In LLMIC, patients with the highest AKI level of severity had higher mortality than patients in higher income categories. Patients from LLMIC and UMIC had a 57-fold and 11 fold higher adjusted risk of death, respectively, compared to patients from HIC. Conclusion In resource-limited countries, pediatric AKI-associated mortality is disproportionately higher when compared to high-resource areas, especially among patients with more severe AKI.


Introduction
In the developed world, the use of standard AKI definitions [1][2][3] has led to an improved understanding of the epidemiology and outcomes of this illness in both adult [4,5] and pediatric [6,7] populations; however, little data is available on the epidemiology of pediatric AKI in resource-limited regions of the world [8]. The scarcity of pediatric reports from these regions impairs the ability to improve AKI prevention and recognition; essential factors needed to decrease the incidence and improve the outcomes of pediatric AKI [8][9][10][11][12][13][14][15].
The "0by25" initiative of the International Society of Nephrology (ISN) has the ambitious goal of reaching zero preventable deaths from AKI across the world by 2025 [16]. The AKI Global Snapshot (GSN) [17] was an initial attempt to understand the incidence, causes, treatment and outcomes of AKI around the world, focusing on lower and lower-middle income group countries, which are underrepresented in the majority of AKI studies. In this report, we hypothesize that the causes and outcomes of pediatric AKI are dissimilar in diverse geographic and economic regions of the world.

Patients and methods
The GSN study design has been described previously [17]. Briefly, this was a prospective observational study conducted worldwide via recruitment of health care volunteers across nephrology societies, meetings, a dedicated website (www.0by25.org), electronic communications and medical society membership. These physicians collected local AKI data, which was transmitted to a data-coordinating center utilizing a standardized electronic form. The University of California, San Diego, Human Research Protection Program approved this study. Participating providers obtained institutional review board approval from their institutions in accordance with local ethics regulations. Providers selected an index day and then screened suspected cases of AKI within a three-day window before or after the index day. Screened patients were followed to identify those that met pediatric KDIGO criteria for AKI. Detailed data was collected for eligible patients on the cause, treatment, and outcome of AKI until the time of hospital discharge, death or otherwise reaching the 7th day of follow-up (S1 Data Elements). Exclusion criteria included patients with chronic kidney disease (CKD) stage 5, known endstage renal disease (ESRD) on dialysis, or with a functional renal transplant.
The most recent creatinine value prior to the development of AKI was considered the baseline creatinine level. If no baseline creatinine value was available, then the first creatinine level available during hospital admission was considered the reference creatinine value. We applied modified AKI KDIGO criteria [1,18,19] and defined confirmed AKI as an increase or decrease in serum creatinine of !0.3 mg/dl and/or an increase in creatinine of ! 50% from the reference value within a minimum of three days from the index day or a prior known baseline value. When urine volume was available, we considered confirmed AKI if urine output was <0.5 ml/kg/hour for six or more hours. We used the pRIFLE criteria to classify AKI severity [20] and the Schwartz formula [21] to estimate glomerular filtration rate (eGFR) in patients under 18 years of age. Urine output criterion is reported as urine volume adjusted by body weight over a 24-hour period. Countries were divided into High (HIC); Upper Middle (UMIC) and Low Middle and Low (LLMIC) income groups based on their 2014 gross national income (GNI) per person. Community-acquired AKI was defined as an episode of AKI when the initial event occurred outside of the hospital setting and the patient was subsequently admitted to the hospital with a diagnosis of AKI. Hospital acquired AKI was defined as an episode of AKI due to a kidney insult that occurred in hospitalized patients who developed de novo AKI during their hospital stay [22]. Outcomes included renal replacement therapy (RRT) requirement, renal recovery and mortality, and were assessed either at seven days or at last observation, if later than 7 days. When known, pre-AKI creatinine values and past medical history were used to determine if CKD was present at baseline. Renal recovery determination was based on the last available serum creatinine (SCr) and defined as "complete" if SCr was equal to or lower than at baseline or reference; "partial" if lower than the value at diagnosis but still higher than baseline or reference, and "no-recovery" if the SCr levels did not decrease and/or the patient remained on dialysis.
Data was entered into a secure web based platform (KEEP.distributedhealthlabs.org and keep.isn.org) that was accessible on the ISN 0 by 25 AKI website (www.0by25.org) [23]. We present continuous variables as mean (standard deviation, SD) or median (interquartile range, IQR), as appropriate. Normality was tested by the Kolmogorov-Smirnov test. The association between levels in each categorical covariate and three GNI levels was evaluated using the Chisquared test. Normality was tested with the Kolmogorov-Smirnov test. Skewed distributions are described with medians and interquartile ranges and were compared with using the Kruskal-Wallis test.
Missing data for height were imputed using the 50 th percentile height based on age and sex from the 2006 WHO Multicenter Growth Reference Study child growth standards. (http:// www.who.int/childgrowth/standards/technical_report/en/). Logistic regression was used to identify risk factors that may have contributed to mortality. In logistic regression, cases that are missing variables in the proposed model are dropped from the analysis, leaving only complete cases. For the model creation, GNI and age were fixed and every other risk factors were conditional on these two covariates. The likelihood ratio test (LRT) was used to compare the model with the additional risk factor to the previous model. We used a P value < 0.10 to indicate whether the additional risk factor was added. Analysis was conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 3.2.1.
centers were in cities with a population of more than five million inhabitants (29, 32.6%). RRT was available in almost all (87, 97.8%) centers.
There were differences in the etiology of AKI in HIC compared to UMIC and LLMIC. Chief factors associated with AKI in HIC were hypotension (30%), post-surgical complications (27%) and dehydration (26%). In contrast, dehydration was the most common etiologic factor in LLMIC (43.5%) and UMIC (30.6%) ( Table 1). Infection, nephrotoxic medications and primary kidney diseases were more common AKI etiologies in LLMIC than in UMIC or HIC countries.

AKI diagnosis and severity
The diagnosis of AKI was made exclusively by SCr criteria in 141 (39.8%); exclusively by urine output in 112 (31.6%); and by both criteria in 101(28.5%) patients. In HIC, SCr levels at diagnosis of AKI were lower than levels from other GNI categories. The first available estimated GFR during hospital admission was lower in UMIC and LLMIC patients than HIC patients (Table 2). Similarly, urine output was lower at diagnosis in LLMIC and UMIC patients compared to HIC patients. Most patients in LLMIC (94%), UMIC (85%) and HIC (73%) were diagnosed at the "Failure" (highest AKI) stage.  Table). Among the dialyzed patients, 18 were from HIC, 29 from UMIC, and 27 from LLMIC. Solute control was the primary indication for RRT, followed by fluid overload, and electrolyte or acid/base disturbances (Table 3). RRT modalities included peritoneal dialysis, intermittent hemodialysis, and continuous RR; the latter modality was not used in any LLMIC patient. Estimated GFR at the start of RRT was significantly lower in LLMIC patients than in those from the other two income categories ( Table 2).
RRT was indicated but not begun in 7 (9.2%) of LLMIC, 6(14.0%) of UMIC and 2(1.3%) of HIC patients (P <0.0001). Futility was cited as the chief cause for withholding RRT in HIC, but in LLMIC, two patients were not dialyzed due to financial constraints and 3 due to lack of trained staff.

Outcomes
Renal recovery. Renal recovery at seven days was complete in 128 (36.6%) patients and partial in 117 (33%) children. No renal recovery was seen in 58 (16%) patients; however, no data on renal recovery was available in 42 (12%) children. Non-recovery rates were significantly lower in HIC patients (12, 6.9%) than in UMIC (22, 30%) or LLMIC patients (24, 22%) ( Table 4). Recovery rates from community-and hospital-acquired AKI were similar. Complete renal recovery was not different among patients reaching maximum "Risk" or "Injury" levels of AKI severity. Non-recovery was more frequent in "Failure" patients from UMIC and LLMIC. Of dialyzed patients, 33 (44%) did not recover renal function by the day of last observation. Patients not requiring RRT recovered function more frequently than those requiring RRT (205, 75.6% versus 33, 44%, respectively; P <0.001) ( Table 4). In the univariate analysis, non-recovery (complete or partial) was associated with hypotension and systemic disease, oliguria at diagnosis and higher baseline SCr.
In UMIC patients, hospital mortality was significantly higher than community acquired AKI, but was not different in HIC and LLMIC patients (Fig 2). In both ICU and non-ICU settings, more deaths occurred in LLMIC patients (39.4% and 10.1% respectively) than in UMIC (8, 32% and 1, 2.1%) and HIC patients (2, 2.85% and 0%; P <0.0001). Mortality in patients receiving RRT was higher across all GNI categories, and reached 25% in LLMIC, compared to 10% in HIC and 13.7% in UMIC (p = 0.03) (Fig 2). No significant differences were found in mortality frequency by maximum AKI stage in HIC and UMIC patients. In LLMIC, the mortality was higher in patients reaching maximum "Failure" stage.
In the univariate analysis, risk factors associated with mortality included lower baseline eGFR, hypotension and GNI category. Pre-existing chronic heart disease, cardiac factors causing AKI and primary kidney diseases were associated with a lower risk for mortality. In multivariable logistic analysis, only GNI category and hypotension remained significantly associated with mortality ( Table 5). The adjusted risk of death among patients from LLMIC was 57-fold higher than in HIC patients, and 11-fold higher in patients from UMIC compared to HIC patients. In patients with hypotension as the cause of AKI, the adjusted risk of death was almost 10-fold higher compared to patients without hypotension as an etiological factor for AKI development (Table 5).

Discussion
The Global Snapshot is the first worldwide prospective cross-sectional study of AKI resulting in hospitalization that was designed to capture data from different health care settings across six continents. A prior publication provided data for all participants in the GSN [23]; this article focuses on the pediatric patients represented in this study. The provenance of this pediatric cohort is mostly from North America and South Asia, but all other geographic areas were represented including Africa, an area where little information was heretofore available [24] (Fig 1). In our sample, patients were mainly diagnosed in large urban centers and managed by nephrologists, which is in stark contrast to the prevalent situation in LLMIC, where nephrologists are scarce or absent, and where AKI occurs in the community and rarely reaches large urban centers [13,14,16,17,[24][25][26]. A unique aspect of our sample was that almost half of the patients were considered to have developed AKI in the community setting. Acute kidney injury occurred in different circumstances in HIC compared to UMIC and LLMIC. Patients from HIC were younger (predominantly neonates and infants), their baseline SCr was more readily available, and their AKI was predominantly attributed to post-operative cardiac complication. Conversely, in LLMIC, AKI commonly developed in the community and baseline renal function data was less frequently available. In LLMIC, the high proportion of community-acquired AKI is likely associated with public health infrastructure-related problems such as poor sanitation, endemic infections and waterborne gastroenteritis, suggesting that primary prevention, early risk assessment, and training of first-contact care providers may decrease the incidence and improve the outcomes of these preventable AKI cases [13,14,16,26].
As expected in this young cohort, the frequency of comorbid conditions was low across the GNI categories except for anemia, which was present in 24% of the patients in LLMIC. The anemia seen in LLMIC patients may be a marker for prevalent malnutrition, chronic infection and/or endemic malarial and parasitic disease, which in turn might impact the incidence and outcome of AKI in this population [27][28][29][30][31][32][33].
Age was not associated with mortality, suggesting that settings and etiologies might be stronger determinants of the dramatic differences in outcome when stratified by GNI. Interestingly, the LLMIC cohort showed a sizable number of patients with primary kidney diseases such as acute glomerulonephritis, illustrating differing characteristics and etiologies of AKI in the developing world; this observation has been reported in other developing country cohorts [24,34,35]. Conversely, a more frequent diagnosis of primary kidney diseases may be the result of indication bias, as renal biopsies were more commonly done in UMIC and LLMIC than in HIC.
Urine output, an important warning sign of impending kidney dysfunction [36], was available in only half of the patients on the day of AKI diagnosis. The high prevalence of neonates, in whom accurate urine output quantification is challenging, may be part of the reason for this lack of data. When available, urine output volume adjusted by weight was significantly lower at diagnosis in LLMIC and UMIC compared to HIC. The importance of urine output assessment in children has been recently demonstrated in an epidemiologic study involving more than 32 ICUs worldwide [15]. In this study, the use of SCr alone would have resulted in missing 67% of AKI cases. Our multivariable analysis for kidney recovery emphasizes the importance of urine output assessment, and also demonstrated that patients with both oliguria and elevated serum creatinine levels had more severe AKI and were less likely to recover renal function completely [36,37].
In the overall GSN cohort, a similar proportion of children and adults (80%) from low and low middle-income countries developed AKI in the community [17]. The etiological factors leading to AKI development were similar: mainly dehydration and hypotension. However, in LLMIC, almost all children were diagnosed at the most severe pRIFLE stage (Failure), as compared with 58% of adults who were diagnosed at the equivalent KDIGO Stage 3 [17]. These differences in time to recognition as well as the increased mortality rate among children demonstrate the urgent need to improve timely access to care centers, and to identify the systematic problems causing such delays [17]. Commonly missed or delayed AKI recognition reflects a lack of awareness of the possibility of kidney disease as well as a lack of resources for diagnosis and management [14,26,38]. In these resource-limited regions, where serum creatinine tests are often unavailable or unaffordable, a lack of appreciation of the severity of the kidney injury likely leads to late referrals to higher levels of care. The higher frequency of community acquired AKI in LLMIC raises the concern that in those limited-resource areas, an even larger number of community cases may have never reached the urban centers and likely remain unaccounted for [14,16,24,35]. Education directed at increasing the awareness of the impact of AKI, the provision of point of care tests, the development and implementation of regionally appropriate kidney care guidelines and the facilitation of prompt referral should all contribute to improvements in the recognition, diagnosis, initial management and monitoring of kidney function [8,14,22,26].
Although by GSN design, patient follow-up was short (seven days), in that limited time window, in LLMIC we were able to demonstrate a much higher mortality rate associated with community-acquired AKI than that observed in HIC. This important finding again reflects the vulnerability of pediatric AKI patients in those regions of the world (Fig 2). Multiple reasons likely contributed to the higher mortality rate, including severe infections, late presentation, delayed recognition and diagnosis, and delayed [14], incomplete or unavailable RRT, all of which have been shown to lead to poor outcomes in developing countries [15][16][17]24,25,35,[39][40][41][42][43].
Many LLMIC countries fail to offer RRT and financial constraints limit the duration of treatment, leading to severely increased mortality. In those countries, improving AKI awareness and promoting risk stratification, earlier diagnosis and treatment should significantly improve outcomes and spare children from the need for more severe, costlier and commonly unaffordable treatments, and from death [22,40,44,45].

Strengths and limitations
The GSN pediatric snapshot is a survey of a convenience sample that may over-represent certain demographic and socioeconomic factors. Firstly, more than 50% of the patients reported were from North America or South Asia. Secondly, urban areas were more likely to report AKI cases than rural areas. Thirdly, heterogeneous age ranges in different country income categories limit our understanding of the overall picture of AKI in these settings. Given that most of the data in GSN was reported by nephrologists, and the actual scarcity or absence of nephrologists in LLMIC, it is very likely that less severe cases of AKI, seen in the community and treated by non-specialists, were not included in the sample.
Outcomes were assessed at seven days or at hospital discharge. This time frame is not long enough to determine the full impact of AKI in mortality or the overall recovery from AKI. In LLMIC, true baseline SCr values were infrequently available. These limitations make it difficult to accurately assess the apparent severity of AKI on presentation, and complicates the distinction between late AKI and AKI superimposed on CKD. On the other hand, the better recovery demonstrated among LLMIC children suggests that the higher serum creatinine levels at hospital admission may actually be associated with a late presentation (as demonstrated by others) [14], again emphasizing the importance of early AKI recognition in achieving better outcomes [22]. This problem requires study in the field and in predominantly rural communities; such studies are being currently conducted in Bolivia, Mali and Nepal, as the International Society of Nephrology "0 by 25" Pilot Study [46].
Notwithstanding these limitations, this snapshot provides data from areas in low and middle-income regions of the world, that has rarely been reported before. The GSN was designed to track patients before AKI confirmation, thus providing an opportunity to ascertain early AKI etiology. Our results shows that pediatric AKI-associated mortality is disproportionately higher in low resource areas when compared to high-resource areas. These results also confirm our previous findings [16,47] that better investment in education and public health system infrastructure, including human resources, have the potential to significantly improve outcomes in children with AKI. We deeply hope that the evidence provided by this report will generate actionable items, and inform the design of current "0by25" pilot projects that aim to determine pathways to reduce the currently dramatic pediatric AKI morbidity and mortality around the world.