Acute kidney injury (AKI) occurring during critical illness is not a trivial complication. Accumulating data show poor outcomes after an episode of AKI, including greater risk of mortality, major morbidity, and health services use (1). Understanding the epidemiology of AKI and its burden has been the focus of several pediatric studies; however, findings across these studies have been variable, mainly due to differences in study design, AKI definitions, and cohort case-mix assessed. Few pediatrics studies have applied the recent Kidney Disease: Improving Global Outcomes (KDIGO) AKI consensus definition, although most have omitted the urine output criteria (2–5). Recently, the Epidemiology of Acute Kidney Injury in Critically Ill Children and Young Adults (AWARE) study, a large prospective multinational study, addressed many of these concerns and validated the utility of the KDIGO definition in diagnosing AKI and predicting its association with outcomes in critically ill children (6). In AWARE, AKI occurred in 27% and severe AKI in 12% of children, respectively. Mortality was highest among those with severe AKI and was further associated with greater use of mechanical ventilation, renal replacement therapy (RRT), and longer ICU stay.
The population-level incidence of AKI has yet to be described. Determination of the AKI incidence in critically ill children is important to establish its overall burden and optimally characterize those at greater risk. This is vital to enable devising preventative and therapeutic strategies. Our objectives were to describe the population-based incidence, risk factors, and outcomes associated with AKI in a large well-defined population.
MATERIALS AND METHODS
The study was approved by the University of Alberta Health Research Ethics Board (File Number PRO00063318). The need for written informed consent was waived.
This study was a retrospective population-based cohort study utilizing prospectively collected data on critically ill children in the province of Alberta, Canada. Alberta Health Services provides all hospital care to the residents of Alberta (2015 population 4,177,527, of which 906,245 were children between 1 month and 17 yr old). Critically ill children in Alberta are managed in three university-affiliated multidisciplinary closed PICUs (two medical-surgical PICU and one cardiac PICU). The study population consisted of all children (1 mo to 17 yr) residing in Alberta admitted to any of the three PICUs between January 1, 2015, and December 31, 2015. Patients were excluded for any of the following: 1) presence of chronic kidney disease (defined as glomerular filtration rate < 60 mL/min per 1.73 m2 for more than 3 mo), 2) history of kidney transplant, 3) admission for less than 24 hours, and 4) transfer into PICU from out of province. If a patient had more than one PICU admission during the study period, only the first admission was included.
The data source was Alberta “eCritical” which is composed of a bedside clinical information system (MetaVision; iMDsoft, Germany) and a data repository/clinical analytics system (TRACER) (7). MetaVision provides full electronic interdisciplinary clinical documentation and collation of demographic, diagnostic (i.e., comorbidity, diagnostic classification, surgical status), laboratory, and device data (i.e., monitor, ventilator, RRT). “eCritical” TRACER provides a comprehensive, multimodal and integrated data repository of patient-specific ICU data enabling creation of reports and data extracts for administrative, quality, and research purposes (8). The size and demographics of the pediatric population in Alberta during the study period was obtained from Statistics Canada, a government agency commissioned with producing population-related statistics (9).
Standard demographic and clinical data were retrieved. Demographic information included age, weight, sex, and date of admission. Severity of illness was assessed using Pediatric Index of Mortality (PIM)–3 at admission (10). Seven main admission diagnostic criteria were assigned based on the primary admission diagnosis: shock, respiratory failure, CNS dysfunction, cardiac disease (including both medical and surgical cardiac patients), postsurgery (other than cardiac surgery), and trauma. Diagnoses that did not fit any of these categories were labeled “other.” Clinical data encompassed hourly fluid intake and output from all sources, receipt of noninvasive or invasive mechanical ventilation, infusions of vasoactive medications, or RRT. Urine output was recorded hourly if patient had urinary catheter, or otherwise averaged over 12 hours. All serum creatinine (SCr) values measured in the 3 months prior to and during PICU admission and index hospitalization were captured. Chronic kidney disease status was obtained based on baseline SCr and patient charts.
AKI was defined using KDIGO definition, utilizing both SCr and urine output criteria (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/B92). When the two criteria resulted in different stages, the greater severity stage was used. Baseline SCr was defined as the lowest SCr in the 3 months preceding admission. When baseline SCr was unavailable, we used the average SCr norms for age and sex used in Alberta clinical laboratories (Supplementary Table 2, Supplemental Digital Content 1, http://links.lww.com/PCC/B92). In patients with no SCr measurements during PICU admission, only the urine output criteria were applied. If a patient had only one SCr value measured during PICU admission, the criteria of absolute SCr increase of greater than or equal to 26.5 mmol/L was not applied. We defined severe AKI as KDIGO stage 2 and 3 based on prior pediatric studies showing these stages are associated with worse outcomes (4,6,11). Early AKI was defined as AKI occurring within the first 24 hours following PICU admission, whereas late AKI was defined as AKI occurring later in the PICU course after the first 24 hours. Fluid overload was defined as peak percent fluid overload more than 10% during the first 10 days of admission (12), calculated as follows:
(Cumulative daily fluid intake in liters–cumulative daily output in liters)/PICU admission weight in kilogram × 100%
The primary outcome was PICU mortality. Secondary outcomes included hospital and 1-year mortality, duration of mechanical ventilation, duration of vasoactive support, receipt and duration of RRT, and length of PICU and hospital stay.
We performed descriptive statistics, reported for the whole cohort and stratified according to severe AKI status and PICU mortality. Categorical variables were reported as proportions and compared using bivariate logistic regression. For continuous variables, data are reported as medians with interquartile ranges (IQRs) and compared using bivariate linear regression.
Population-based incidence and mortality rates were calculated using the pediatric population of Alberta in 2015 as the denominator. Incidence rate ratio (IRR) was calculated for age and sex categories as follow: (number of AKI cases in category/number of category population in Alberta)/(number of other pediatric AKI cases not in category/number of Alberta children population not in category).
Multivariate logistic regression was used to identify factors independently associated with severe AKI. We included in the multivariate model variables carrying associations of p value of less than 0.10 with severe AKI in the bivariate analysis, in addition to the prespecified variables (weight, diagnosis, PIM 3 score, and site). To examine the association of severe AKI with outcomes, we used multivariate logistic and linear regression, as appropriate. We adjusted for the following prespecified clinically important variables: weight, diagnosis, PIM 3 score, and site. Results are presented as odds ratios (ORs) with 95% CIs for logistic regression and B regression coefficient with 95% CIs for linear regression. Finally, we performed survival analysis through 1 year using Cox proportional hazards regression adjusting for the same four variables (weight, diagnosis, PIM 3 score, and site) and reported hazard ratios with 95% CIs. Two-sided analyses were performed with p value of less than or equal to 0.05 considered to be statistically significant. Analysis was performed using Stata Version 15.1 (Stata Corp, College Station, TX).
During the 12-month study period, we analyzed data for 1,017 children who met the eligibility criteria (Fig. 1). The rate of PICU admissions was 112 (95% CI, 105–119) per 100,000 children-year. The median (IQR) age was 24 months (6.7–96.0 mo). Fifty-six percent were males. Details of baseline demographics and clinical characteristics are presented in Table 1.
Acute Kidney Injury Incidence and Risk Factors
AKI occurred in 308 patients (30.3%; 95% CI, 28.1–33.8%) and severe AKI (stage 2 or 3) occurred in 124 patients (12.2%; 95% CI, 10.3–14.4%). The population incidence rates for critical illness-associated AKI and severe AKI were 34 (95% CI, 30–38) and 14 (95% CI, 11–16) per 100,000 children-year, respectively. Annual incidence rate for severe AKI was higher in males (IRR, 1.55; 95%, 1.08–2.33) compared with females and infants younger than 1 year old compared with older children (IRR, 14.77; 95% CI, 10.36–21.07) (Supplementary Table 3, Supplemental Digital Content 1, http://links.lww.com/PCC/B92).
Cardiac diagnosis was the most common diagnostic category in AKI patients, representing 41% of AKI patients and 43% of severe AKI patients. Although cardiac diagnosis was associated with severe AKI on bivariate analysis, this association did not persist on multivariate analysis. Factors associated with the development of severe AKI on multivariate analysis included PIM 3 score, the receipt of vasoactive support, fluid overload, and PICU site (Table 1). Fluid overload occurred more frequently in those with severe AKI (50.8% vs 30.2%; p < 0.001). Median (IQR) peak fluid overload % during the first 10 days was higher in those with severe AKI (9.4% [3.6–17.9%] vs 5.7% [2.6–10.2%]; p < 0.001).
Timing and Characteristics of AKI
The majority of AKI cases (57.8%) were diagnosed during the first 24 hours of admission. Early AKI (in the first 24 hr) had higher proportion of severe AKI cases compared with AKI occurring later in the PICU course (50.6% vs 26.2%; p < 0.001). Early AKI occurred more frequently in cardiac patients, whereas late AKI occurred more commonly in respiratory and postsurgical patients. The median (IQR) incidence day in the late AKI group was day 2 (2–4). Only nine AKI cases (2.9%) were diagnosed after the first week of critical illness (Table 2).
The average time to AKI resolution (based on SCr decreasing to < 1.5 times baseline and urine output > 0.5 mL/kg/hr for more than 12 hr) was 3.02 days (95% CI, 1.62–4.41 d). Both early and severe AKI were associated with longer duration of AKI (p < 0.001) compared with late and stage 1 AKI. The diagnosis of the majority of AKI cases was made using the SCr criteria (145 cases, 47.1%), whereas 97 cases (31.5%) were diagnosed based on urine output criteria, and 66 cases (21.4%) met both urine and creatinine criteria (Fig. 2). AKI based on both criteria was more severe and of longer duration compared with AKI diagnosed based on one criterion (p < 0.001). In early AKI, the majority of cases (55%) were diagnosed based on the SCr criteria, whereas in late AKI, most cases (52%) were diagnosed based on urine output criteria. Detailed data of AKI characteristics and timing are provided in Table 3 and Supplementary Tables 4–8 (Supplemental Digital Content 1, http://links.lww.com/PCC/B92).
Urine output data were available for all included patients. On the other hand, 18.7% of the cohort had no measured SCr during PICU admission. In those, AKI was diagnosed less frequently compared with the children with at least one creatinine measurement (8.4% vs 35.3%; p < 0.001). PIM 3 scores (median [IQR]) were lower in patients with no measured SCr (–6.15 [–6.51 to –4.93] vs –4.51 [–5.60 to –3.89]; p < 0.001).
Baseline SCr was unavailable and was estimated for 49.4% of children. AKI occurred less commonly in those without baseline SCr compared with those with available baseline SCr (21% vs 40%; p < 0.001). When omitting all patients with no baseline SCr, the incidence of AKI increased to 39.6%.
Twenty-one AKI patients (2.1%) received RRT with 17 patients treated with continuous RRT and four with peritoneal dialysis.
AKI Association With Mortality
Thirty-two patients (3.1%) died during PICU admission, and of those 21 (65%) had AKI. The annual AKI-associated PICU mortality rate was 2.3 (95% CI, 1.4–3.5) per 100,000 children. AKI-associated PICU mortality rate was higher in infants younger than 1 year old compared with older children (IRR, 13.87; 95% CI, 5.89–32.67). There was no significant difference in mortality risk between males and females (Supplementary Table 9, Supplemental Digital Content 1, http://links.lww.com/PCC/B92).
When comparing survivors and nonsurvivors, those who died had a higher PIM 3 score, were more likely to have greater than 10% fluid overload, and receive treatment with mechanical ventilation, vasoactive support, and RRT (Table 4).
Patients with AKI had higher mortality rates compared with patients with no AKI (6.8% vs 1.1%; OR, 4.64; 95% CI, 2.21–9.75). There was gradient increase in mortality rates according to increasing AKI severity, however, only stage 2 and stage 3 AKI were associated with PICU mortality. Severe AKI conferred an incremental risk of PICU mortality after adjustment for weight, diagnosis, PIM 3 score, and PICU site (adjusted OR, 11.93; 95% CI, 4.68–30.42).
Mortality rates did not differ between those with estimated baseline SCr and measured baseline SCr (2.7% vs 3.4%; p = 0.52). However, patients with no measured SCr during admission had significantly lower mortality rates compared with those with at least one measured SCr (0.5% vs 3.8%; p = 0.02).
Thirty-eight patients (3.7%) died before hospital discharge and 56 died (5.5%) within 1 year of PICU admission. Postdischarge mortality (n = 18) was higher in patients with severe AKI (OR, 2.84; 95% CI, 1.04–7.81). In adjusted analysis, severe AKI was associated with greater hospital (OR, 9.20; 95% CI, 4.00–21.19) and 1-year mortality (OR, 5.50; 95% CI, 2.76–10.96). Cox proportional hazards analysis, adjusting for the same variables, showed that stage 2 and stage 3 AKI were significantly associated with 1-year mortality (Fig. 3).
AKI Association With Other Outcomes
In adjusted analyses, patients with severe AKI received longer duration of mechanical ventilation (3.31 d; 95% CI, 1.06–5.55) and vasoactive therapy (2.46 d; 95% CI, 1.44–3.48). In addition, patients with severe AKI stayed longer in PICU (3.75 d; 95% CI, 1.28–6.23) and hospital (12.09 d; 95% CI, 3.11–21.06) (Table 5). There was a statistically significant worsening of secondary outcomes according to worsening AKI status (Supplementary Table 10, Supplemental Digital Content 1, http://links.lww.com/PCC/B92).
Multivariate models adjusting for fluid overload (Supplementary Table 11, Supplemental Digital Content 1, http://links.lww.com/PCC/B92) and sensitivity analysis excluding cardiac patients (Supplementary Table 12, Supplemental Digital Content 1, http://links.lww.com/PCC/B92) showed similar significant association between severe AKI and primary and secondary outcomes.
In this large multicenter cohort study, we described the population-level epidemiology and outcomes for critically ill children with AKI. We found AKI was common, occurring in an estimated 30% of the PICU patients, with annual incidence rate of 34 per 100,000 children. Severe AKI was independently associated with greater risk of short and longer-term mortality and other important secondary outcomes, including longer receipt of mechanical ventilation and vasoactive support, and prolonged PICU and hospital stay.
To our knowledge, no pediatric study has described the population-level incidence of AKI among critically ill children. The incidence rate of AKI in our study population was higher in comparison to other acute illnesses associated with PICU admission such as respiratory failure (32 per 100,000 person-year), cardiac diseases (23 per 100,000 person-years), shock (eight per 100,000 person-years), and trauma (seven per 100,000 person-years). Several hospital-based pediatric studies, applying different AKI definitions, have described the incidence rates of AKI between 16% and 51% (2,4,6,13–16). The AWARE study prospectively evaluated the epidemiology of AKI in a large pediatric population from 32 PICUs from around the world using the KDIGO definition (6). Remarkably, the incidence and associated outcomes of AKI in our study were similar to those described in AWARE. In both studies, there was stepwise increase in mortality associated with worsening AKI severity. This consistency in findings between the two studies extends confidence in the estimates reported in AWARE and suggests that KDIGO criteria is robust for discriminating clinically relevant outcomes in pediatric populations. In addition to greater risk for mortality, we found AKI (including stage 1) to be associated with greater health services use in a stepwise fashion with increasing AKI severity. These results have significant implications on healthcare delivery, especially in areas with limited resources, and further highlights the need for effective strategies to identify, prevent, and mitigate the impact of AKI.
The long-term implications of AKI have seldom been described in children surviving critical illness, with most pediatric AKI studies only report mortality at PICU or hospital discharge. In our study, severe AKI patients were at increased risk of death beyond discharge from hospital, adding to the overall burden potentially attributable to AKI. A recent retrospective study including over 2,000 PICU patients found AKI to be associated with 5-year mortality (17). This finding is consistent with data from numerous adult studies showing AKI to have negative impact on long-term survival (18–20).
Approximately one-third of AKI patients were diagnosed based on urine output criteria alone, underscoring the importance of closely monitoring urine output during PICU stay. Similar to findings in AWARE, in our study, approximately half of AKI cases diagnosed after the first 24 hours of PICU admission would have been missed if urine output data were not evaluated (6).
We identified three clinical factors associated with the development of severe AKI including severity of illness, the receipt for vasoactive support, and fluid overload. Of these, fluid overload may potentially be modifiable and a target for interventional strategies. Our study findings align with robust evidence describing the negative association between fluid accumulation and patient outcomes, including AKI. In a recent systematic review of 44 pediatric studies, fluid overload was associated with increased mortality (OR, 4.34; 95% CI, 3.01–6.26) and AKI (OR, 2.36; 95% CI, 1.27–4.38) (12). The findings from our study suggest that future research should now evaluate potential interventions to prevent avoidable fluid accumulation in critically ill children.
Our study is strengthened by the use of prespecified protocol and the utilization of prospectively electronically collected data with no missing data points. It is one of the largest pediatrics studies to investigate AKI using both creatinine and urine output criteria. The data were obtained from three different PICUs with variable practices and processes of care increasing the generalizability of the findings. The population-based nature and inclusion of all patients in a geographically defined area governed by a single health system minimizes the risk of selection bias while providing robust estimates of disease burden. Furthermore, unlike prior work, we were able to evaluate patients for AKI during the entire PICU stay. We also described long-term outcome data, which have been infrequently reported in pediatric AKI literature. We used a novel method to estimate baseline creatinine utilizing population-based average creatinine.
The study has limitations that warrant discussion. The study is observational in design, and as such, we cannot definitively confirm the causal relationship between AKI and clinical outcomes. The population-based incidence of all pediatric AKI cases occurring in hospital cannot be determined, considering we only included patients admitted to PICU. Despite adjusting for severity of illness and other important potential confounders, the retrospective design limited our ability to account for important interventions such as exposure to nephrotoxins or the use of diuretics. This could partly explain the observed variable AKI incidence between centers. The baseline SCr was not known in about half of the patients. Similar figures have been described in other pediatric studies and can be explained by the fact that majority of children have no comorbidities that necessitate evaluating SCr in the community (6,14). Nevertheless, this could be a source of classification bias. When we excluded those patients, the incidence of AKI was higher, but the association with mortality was unchanged. Our data sources did not have access to pre-PICU urine output and creatinine data. As such, we cannot explore what proportion of early AKI patients had AKI during hospitalization prior to PICU admission. This too is a potential source of misclassification.
AKI is common in critically ill children and portends worse outcomes, including greater mortality and healthcare resource utilization. AKI therefore represents an important burden for healthcare. The findings of this study highlight the need for improved detection of AKI and for efforts to improve care for these patients.
1. Hoste EAJ, Kellum JA, Selby NM, et al. Global epidemiology
of acute kidney injury
. Nat Rev Nephrol 2018; 14:607–625
2. Selewski DT, Cornell TT, Heung M, et al. Validation of the KDIGO acute kidney injury
criteria in a pediatric
critical care population. Intensive Care Med 2014; 40:1481–1488
3. Lex DJ, Tóth R, Cserép Z, et al. A comparison of the systems for the identification of postoperative acute kidney injury
cardiac patients. Ann Thorac Surg 2014; 97:202–210
4. Sutherland SM, Byrnes JJ, Kothari M, et al. AKI in hospitalized children: Comparing the pRIFLE, AKIN, and KDIGO definitions. Clin J Am Soc Nephrol 2015; 10:554–561
5. Sanchez-Pinto LN, Goldstein SL, Schneider JB, et al. Association between progression and improvement of acute kidney injury
and mortality in critically Ill children. Pediatr Crit Care Med 2015; 16:703–710
6. Kaddourah A, Basu RK, Bagshaw SM, et al.; AWARE Investigators: Epidemiology
of acute kidney injury
in critically Ill children and young adults. N Engl J Med 2017; 376:11–20
7. Alberta Health Services: Alberta’s Strategic Clinical Networks-eCritical Available at: https://www.albertahealthservices.ca/assets/about/scn/ahs-scn-sb-cc-ecritical.pdf
. Accessed September 20, 2018
8. Brundin-Mather R, Soo A, Zuege DJ, et al. Secondary EMR data for quality improvement and research: A comparison of manual and electronic data collection from an integrated critical care electronic medical record system. J Crit Care 2018; 47:295–301
9. Statistics Canada: Table 051-0001 - Estimates of Population, by Age Group and Sex for July 1, Canada, Provinces and Territories, Annual (Persons Unless Otherwise Noted), CANSIM (Database). Available at: https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1710000501
. Accessed March 11, 2018
10. Straney L, Clements A, Parslow RC, et al.; ANZICS Paediatric Study Group and the Paediatric Intensive Care Audit Network: Paediatric index of mortality 3: An updated model for predicting mortality in pediatric
intensive care*. Pediatr Crit Care Med 2013; 14:673–681
11. Hoste EA, Bagshaw SM, Bellomo R, et al. Epidemiology
of acute kidney injury
in critically ill patients: The multinational AKI-EPI study. Intensive Care Med 2015; 41:1411–1423
12. Alobaidi R, Morgan C, Basu RK, et al. Association between fluid balance and outcomes
in critically ill children: A systematic review and meta-analysis. JAMA Pediatr 2018; 172:257–268
13. Schneider J, Khemani R, Grushkin C, et al. Serum creatinine as stratified in the RIFLE score for acute kidney injury
is associated with mortality and length of stay for children in the pediatric
intensive care unit. Crit Care Med 2010; 38:933–939
14. Alkandari O, Eddington KA, Hyder A, et al. Acute kidney injury
is an independent risk factor for pediatric
intensive care unit mortality, longer length of stay and prolonged mechanical ventilation in critically ill children: A two-center retrospective cohort study. Crit Care 2011; 15:R146
15. Mehta P, Sinha A, Sami A, et al. Incidence of acute kidney injury
in hospitalized children. Indian Pediatr 2012; 49:537–542
16. Soler YA, Nieves-Plaza M, Prieto M, et al. Pediatric
risk, injury, failure, loss, end-stage renal disease score identifies acute kidney injury
and predicts mortality in critically ill children: A prospective study. Pediatr Crit Care Med 2013; 14:e189–e195
17. Hessey E, Morissette G, Lacroix J, et al. Long-term mortality after acute kidney injury
in the pediatric
ICU. Hosp Pediatr 2018; 8:260–268
18. Lafrance JP, Miller DR. Acute kidney injury
associates with increased long-term mortality. J Am Soc Nephrol 2010; 21:345–352
19. Wald R, Quinn RR, Adhikari NK, et al.; University of Toronto Acute Kidney Injury
Research Group: Risk of chronic dialysis and death following acute kidney injury
. Am J Med 2012; 125:585–593
20. Horkan CM, Purtle SW, Mendu ML, et al. The association of acute kidney injury
in the critically ill and postdischarge outcomes
: A cohort study*. Crit Care Med 2015; 43:354–364