Several investigations have demonstrated that acute

Association Between Progression and Improvement of Acute Kidney Injury and Mortality in Critically Ill Children* L. Nelson Sanchez-Pinto, MD 1 ; Stuart L. Goldstein, MD 2 ; James B. Schneider, MD 3 ; Robinder G. Khemani, MD, MsCI 1 Objective: To determine whether the progression and/or improvement of acute kidney injury in critically ill children is associated with mortality. Design: Retrospective. Setting: Multidisciplinary, tertiary care, 24-bed PICU. Patients: A total of 8,260 patients who were 1 month to 21 years old with no chronic kidney disease admitted between May 2003 and March 2012. Interventions: We analyzed patients based on their acute kidney injury stage as per the Kidney Disease Improving Global Outcomes acute kidney injury serum creatinine staging criteria on ICU admission, peak (highest acute kidney injury stage reached), and trough (lowest acute kidney injury stage after the peak) during their ICU stay. Nonrenal organ dysfunction was measured with a modified Pediatric Logistic Organ Dysfunction score. The primary outcome was 28-day mortality. p values were based on Yatescorrected chi-square test and logistic regression. *See also p. 774. 1 Department of Anesthesiology and Critical Care Medicine, Children s Hospital Los Angeles, University of Southern California Keck School of Medicine, Los Angeles, CA. 2 Center for Acute Care Nephrology, Cincinnati Children s Hospital Medical Center, Cincinnati, OH. 3 Division of Critical Care Medicine, Cohen Children s Medical Center of New York, North Shore Long Island Jewish Health System, New Hyde Park, NY. Dr. Sanchez-Pinto and Khemani conceived and designed the study and collected data and analyzed. All the authors drafted the article for important intellectual content. Dr. Goldstein consulted for Otsuka, Bard, Ikaria, and La Jolla Pharm; lectured for Baxter Gambro Renal; and has stock in Hemametrics. Dr. Goldstein and his institution consulted for Baxter Gambro Renal. His institution received grant support from Baxter Gambro Renal. Dr. Khemani lectured for the Society of Critical Care Medicine (Honorarium for lectures at SCCM courses) and has a pending patent (unrelated to this paper or any of the methods). His institution received grant support from the National Institutes of Health (K 23 Award, Not directly related to project). The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: nsanchezpinto@chla.usc.edu Copyright 2015 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies DOI: 10.1097/PCC.0000000000000461 Measurements and Main Results: Of the 8,260 patients, 529 (6.4%) had acute kidney injury on ICU admission and 974 (11.8%) had acute kidney injury during their ICU course. The 28-day mortality was 2.7% for patients with no acute kidney injury and 25.3% for patients with acute kidney injury. Patients in whom acute kidney injury developed or had worsening acute kidney injury from admission to peak and reached acute kidney injury stage 2 or 3 had higher mortality than those who remained at an acute kidney injury stage 1 (17.3 17.8% vs 32.2 37.9%; p 0.003). Patients whose acute kidney injury resolved after the peak had lower mortality than those who retained the same degree of acute kidney injury (9 13.5% vs 37.3 44%; p 0.04). Patients with acute kidney injury that resolved still had higher mortality than those who never developed acute kidney injury (11.2% vs 2.7%; p < 0.001). Multivariate regression demonstrated that the association between mortality and acute kidney injury progression was independent of severity of illness at admission and the severity of nonrenal organ dysfunction during the first week of ICU stay (p 0.01). Conclusion: Progression of acute kidney injury per the Kidney Disease Improving Global Outcomes staging criteria is independently associated with increased mortality in the PICU while its improvement is associated with a stepwise decrease in mortality. (Pediatr Crit Care Med 2015; 16:703 710) Key Words: acute kidney injury; critical care; mortality; organ dysfunction; pediatrics Several investigations have demonstrated that acute kidney injury (AKI) is associated with morbidity and mortality in pediatric intensive care settings (1 3). The implications of this relationship between AKI and mortality in the ICU are significant (4 8), yet there are few epidemiologic data to support that there is a relationship between AKI progression or improvement and outcome in children. Several adult AKI studies have shown an association between recovery of kidney function and improved outcomes (9 11). Most of these studies have defined renal recovery as independence from dialysis and have focused on long-term outcomes, but there is also evidence that early recovery of kidney function, Pediatric Critical Care Medicine www.pccmjournal.org 703

Sanchez-Pinto et al especially after cardiac bypass, has an association with reduced hospital mortality (12, 13). There are no studies looking at the relationship between improvement of kidney function and mortality after AKI in the PICU population. Understanding this relationship might help us better define the populations at risk and guide the design of future AKI research studies. The goal of our study was to explore the relationship between ICU mortality and the progression and/or improvement of AKI during ICU admission for critically ill children. We explored this relationship by analyzing the dynamic change in severity of AKI as defined by the Kidney Disease Improving Global Outcomes (KDIGO) AKI staging criteria in a heterogeneous PICU population. METHODS Study Design and Data Collection We performed a retrospective analysis of all patients admitted to a multidisciplinary, tertiary PICU between May 2003 and March 2012. This 24-bed PICU serves a mixed population of medical, surgical, trauma, and solid-organ and stem cell transplantation patients, but not postoperative cardiac patients. Data were extracted from electronic medical records (Cerner Kids, Kansas City, MO; Philips/CareVue, Waltham, MA) and a locally developed quality improvement and clinical patient database (Microsoft Access, Seattle, WA) maintained by the ICU physicians delivering care. Patients were excluded if they were less than 1 month or more than 21 years, had documented chronic kidney disease, were perioperative from a kidney transplant, or had no serum creatinine (SCr) concentrations measured. Each ICU admission was treated independently. This study was approved with a waiver of informed consent by the Institutional Review Board of Children s Hospital Los Angeles. Definition of AKI and Other Organ Dysfunction AKI was defined by the KDIGO AKI staging SCr criteria (14). SCr levels measured during the ICU stay were compared to a baseline creatinine, which was the most recent documented SCr within 6 months of the ICU admission. If a prior SCr was unavailable, the upper limit of normal for age and gender was used (1, 15, 16). Patient height was unavailable for a majority of patients, limiting our ability to use estimated glomerular filtration rate for our calculations. Initiation of renal replacement therapy (RRT) was also considered as a criterion for AKI stage 3 as per the KDIGO guidelines. When the baseline creatinine was below the low normal SCr for age and gender, the low normal for age and gender was assigned as a baseline. This was done because some patients with very low SCr levels (e.g., 0.1 mg/dl) could meet AKI criteria (0.15 or 0.2 mg/dl, in this example) with changes that are within the accepted calibration variability of the laboratory test and are still below the low normal for age and gender. A KDIGO AKI stage was assigned at three time points during the ICU stay: admission, peak (highest SCr-based AKI stage reached), and trough (lowest AKI stage reached after the peak). As per the Table 1. Acute Kidney Injury Stages as per the Serum Creatinine Kidney Diseases Improving Global Outcomes Criteria Acute Kidney Injury Stage Kidney Disease Improving Global Outcomes Criteria 1 1.5 1.9 times baseline SCr OR 0.3 mg/dl increase in SCr within 48 hr 2 2.0 2.9 times baseline SCr 3 3.0 times baseline SCr OR Increase in SCr to 4.0 mg/dl OR Initiation of renal replacement therapy OR In patients < 18 yr, decrease in estimated glomerular filtration rate < 35 ml/min per 1.73 m 2 SCr = serum creatinine. KDIGO AKI staging criteria, stage 1 is an increase of SCr 1.5 1.9 times the baseline creatinine or an increase of at least 0.3 mg/ dl in 48 hours, stage 2 is 2.0 2.9 times the baseline creatinine, and stage 3 is 3.0 times the baseline creatinine or initiation of RRT or SCr increase at least 4.0 mg/dl (14) (Table 1). To classify other organ dysfunction (OD), we created a nonrenal OD score, modified from Pediatric Logistic Organ Dysfunction (PELOD) (17 19). This was required because there would be collinearity between AKI category and PELOD since both use SCr, preventing multivariate modeling. We also needed a score that could be calculated at multiple time points during the ICU stay. PELOD uses 12 variables from six organ systems and has scores from 0 to 71. Our nonrenal OD score included the same weighted variables as PELOD excluding the renal variables (since it was the organ of interest) and the neurologic variables (since many patients are intubated and sedated, artificially changing the Glasgow Coma Scale score and pupillary response). The nonrenal OD score therefore included nine variables (heart rate, systolic blood pressure, Pao 2 /Fio 2 ratio, Paco 2, mechanical ventilation, WBC count, platelet count, serum aspartate transaminase level, and prothrombin time/international normalized ratio) with a score from 0 to 41. We computed the score daily for the first week of admission using the worst recorded value of each variable for each 24-hour period, as per the PELOD methodology (19). To deal with the nonlinearity of the mortality distribution of the modified OD score and aid clinical interpretation, patients were divided into four groups based on their maximum nonrenal OD score: no OD (score of 0), mild OD (1 10), moderate OD (11 20), and severe OD (> 20). These cutoff values were chosen based on their association with mortality. Statistical Analysis Data were analyzed using STATA version 10 (StatCorp, College Station, TX) and Statistica version 12 (Statsoft, Tulsa, OK). Each ICU episode was initially assigned to one of three groups depending on the worst stage of AKI reached during the ICU course: no AKI, stage 1, stage 2, or stage 3. Stages 2 and 3 had similar mortality, as 704 www.pccmjournal.org October 2015 Volume 16 Number 8

Feature Articles it has been shown in prior pediatric studies, so they were grouped for analysis (1, 2). Univariate analysis was performed to examine the relationship between clinical and demographic variables and AKI. Proportions for categorical data were compared using the Yates-corrected chi-square test, and medians for continuous data were compared with the Mann-Whitney U test. The primary outcome was 28-day mortality. To evaluate the association between change in AKI during ICU admission and mortality, patients were divided into 11 groups: five groups from admission to peak AKI stage, five groups from peak to trough AKI stage, and one group comprising the patients who never developed AKI during the ICU stay (Fig. 1). The mortality rates for the different groups were compared using the Yates-corrected chi-square test. Log-rank tests were used to evaluate Kaplan-Meier survival curves of the different progression/improvement groups. Secondary analysis aimed to determine the association between changes in AKI and mortality, controlling for severity of illness at admission and other OD during the first week of admission using two multivariate logistic regression models. The Pediatric Index of Mortality (PIM)-2 (20) score was used to control for severity of illness at admission, and the maximum nonrenal OD score was used to control for the severity of other OD during the first week of ICU stay. For the models, we report adjusted odds ratios (ORs) and 95% CIs. Model calibration was assessed with the Hosmer- Lemeshow goodness-of-fit test and visual inspection for influential points. Discrimination of mortality for the maximum nonrenal OD is reported using areas under the curve (AUC) of the receiver operating characteristic (ROC) plots. RESULTS Epidemiology There were 10,411 patient admissions during the study period; 2,151 were excluded, leaving 8,260 for analysis (Fig. 2). Mortality was 5.4% for the included cohort, and 3.9% for those who were excluded from the study. The prevalence of AKI using KDIGO in the included cohort was 11.8% during the ICU course, with 6.4% of patients having AKI at admission. Of those with AKI, 80.8% met AKI criteria within 72 hours of admission and 89.8% within a week of ICU admission; 62.4% of patients with AKI met stage 2 or 3 criteria during their ICU stay. The 28-day mortality was nearly nine-fold higher for those with AKI compared with those without AKI (Table 2), with a stepwise increase in mortality across AKI severity stages: 17.5% for patients with stage 1 AKI, 30.4% for patients with stage 2 AKI, and 29.7% for patients with stage 3 AKI (p < 0.001 between no AKI and stage 1, stages 1 and 2 or 3; p = 0.92 between 2 and 3). Among AKI survivors, 38% had AKI at the time of ICU discharge. For patients with AKI, the median time to admission SCr was 2.9 hours (interquartile range [IQR], 1.4 6.6 hr); to peak creatinine was 36.3 hours (IQR, 4.8 135.5 hr), and to trough creatinine was 103 hours (IQR, 32.3 162.2 hr). There was a drop in mortality over the study period (6.2% mortality before 2008 vs 4.6% after 2008; p = 0.001), but the prevalence of AKI remained stable (11.5% vs 12.1%; p = 0.37) and the proportion of children dying with AKI had an increasing trend (51.2% vs 60%; p = 0.09). Demographics Table 2 summarizes patient characteristics as a function of AKI status. Patients with AKI had significantly longer ICU lengths of stay, were less likely to be white, and were more likely to be black than patients with no AKI. The primary reason for ICU admission for patients with AKI was more likely to be respiratory distress/failure, cardiovascular compromise, metabolic derangement, or hemorrhage/coagulopathy and less likely to be postoperative care or neurologic compromise. Patients with AKI had significantly higher PIM-2 risk of mortality and nonrenal OD scores than patients without AKI. Nine high-risk diagnoses associated with pediatric AKI were identified in the literature, and their association with AKI was explored in our cohort (Table 2). Figure 1. The 28-d mortality for the different acute kidney injury (AKI) groups depending on the changes in the Kidney Disease Improving Global Outcomes AKI staging: from admission to the peak AKI (A) and from peak to trough AKI (lowest AKI stage achieved after the peak, B). Differences between groups: a,b,f p < 0.001, c p = 0.003, d p = 0.04, e p = 0.002. Of note, the mortality is not shown to scale. Use of RRT One hundred seventy-four patients with AKI (17.7%) received RRT and were therefore classified as KDIGO AKI stage 3. Thirty-eight of those patients (21.8%) reached AKI stage 3 based solely on the use of RRT of which 50% had SCrbased stage 2 AKI, 21% had SCr-based stage 1 AKI, and 29% had no AKI per SCr criteria before initiation of RRT. In total, 46.6% of RRT patients Pediatric Critical Care Medicine www.pccmjournal.org 705

Sanchez-Pinto et al nonrenal OD scores and only slightly worse PIM-2 risk of mortality (0.5% difference) when compared with patients with resolved AKI. Patients with persistent AKI were also less likely to be recovering from surgery and more likely to have an associated diagnosis of coagulopathy, pre-icu cardiac arrest, and/ or stem cell transplantation. The majority of the deaths in the persistent AKI group were associated with respiratory failure (48.3%), sepsis (44.2%), coagulopathy (35.5%), and/or pre- ICU cardiac arrest (33.1%). Survival Analysis. Kaplan-Meier curves are shown in Figure 3 for a representative subgroup of patients. The median time to death for nonsurvivors was shorter in patients who had little or no improvement of their AKI after the peak, and the separation of the survival curves occurred early in the course. Figure 2. Flow diagram of admissions during the study period with the prevalence of acute kidney injury (AKI) according to the Kidney Disease Improving Global Outcomes AKI staging criteria and the associated mortalities. Of note, patients excluded for lack of serum creatinine measurements had a median length of stay of 1.2 d and a mortality of 3.4%. CKD = chronic kidney disease. received intermittent hemodialysis or continuous hemofiltration/hemodiafiltration, 29.3% received peritoneal dialysis, and 24.1% received more than one mode of dialysis. There was an increase in the proportion of dialyzed patients receiving hemofiltration during the study period (40.7% before 2008 vs 71.2% after 2008; p < 0.001). The mortality rate for those who received RRT was 30.5%. Primary Outcome New or Progressing AKI. Patients who developed or had progression of their AKI (peak AKI > admission AKI) had higher mortality than those who had less progression or unchanging (Fig. 1A). In general, progressing from stage 1 AKI at admission to peak stage 2 or 3 AKI had the highest associated mortality rate (37.9%), followed by those with no AKI at admission that progressed to peak stage 2 or 3 (32.2%). The differences in mortality between each progression group categorized by their admission AKI stage were all statistically significant (p 0.003) (Fig. 1A). For the 511 patients with new or progressing AKI, nonsurvivors met AKI criteria sooner than survivors (median, 36.7 hr vs 61.3 hr; p = 0.003). Improvement of AKI. Patients who resolved their AKI or had improvement of their severity (peak AKI > trough AKI) had lower mortality than those who retained the same degree of AKI (Fig. 1B). Those who remained at stage 2 or 3 AKI after the peak had the highest 28-day mortality (44%), followed by those who remained at stage 1 AKI after the peak (37.3%). Of note, those patients who had any degree of AKI at their peak and improved to no AKI (507 patients) still had higher mortality than those who had no AKI during their ICU course (11.2% vs 2.7%; p < 0.001). Patients who had persistent AKI had higher mortality than those who recovered from their AKI (40.5% vs 11.2%; p < 0.001). Table 3 summarizes the patient characteristics as a function of AKI resolution or persistence. Patients with persistent AKI had similar admission and maximum Secondary Outcome Effect of Severity of Illness at Admission. All patients had a PIM-2 score calculated at admission. PIM-2 scores had excellent discrimination of ICU mortality in this cohort (AUC ROC plot, 0.80 [95% CI 0.77 0.82]). Eight hundred seventy-five patients (10.6%) had AKI during the first week of ICU admission, and these patients were divided into four groups based on their trajectory of AKI progression and improvement during the first week of ICU stay. Table 4 shows the unadjusted ORs as well as the PIM-2 and year of admission-adjusted ORs for mortality of the different progression/improvement groups. All AKI progression/improvement groups had an independent association with mortality after controlling for severity of illness at admission (p 0.001), with appropriate model calibration (Hosmer-Lemeshow test, p = 1). Effect of Other OD. All patients had daily nonrenal OD scores computed for the first week of admission, and the maximum nonrenal OD score was used to control for the effect of other OD. The nonrenal OD score had excellent discrimination of ICU mortality in this cohort (AUC ROC plot, 0.84 [95% CI, 0.83 0.86]). The OD score category-based mortality was as follows: no OD, 0.2%; mild OD, 2.1%; moderate OD, 8%; and severe OD, 39.5%. Table 3 shows the nonrenal OD score and year of admission-adjusted ORs for mortality of the different progression/improvement groups. In general, all AKI progression/improvement groups had an independent association with mortality after controlling for other OD (p 0.01) except for the group that had no AKI at admission, developed AKI, and resolved it during the first week of ICU stay (p = 0.09). Model calibration was appropriate (Hosmer-Lemeshow test, p = 0.7). DISCUSSION We have demonstrated an independent association between the development, progression, and improvement of AKI and mortality in critically ill children. The association between mortality and AKI progression/improvement holds after controlling for severity of illness and other OD. Patients with persistent AKI during their ICU course were 10 11 times more likely to die after controlling for other OD than patients with no AKI. In contrast, patients who resolved their AKI within the first week of ICU stay had a five- to seven-fold lower mortality 706 www.pccmjournal.org October 2015 Volume 16 Number 8

Feature Articles Table 2. Descriptive Statistics of Patients With and Without Acute Kidney Injury Variables No AKI AKI p Total, n (%) 7,286 (88.2) 974 (11.8) Age, yr (IQR) 7.4 (1.8, 13.6) 6.7 (1.6, 13.9) 0.56 Weight, kg (IQR) 22.5 (11, 45) 20 (10, 46.9) 0.23 Males, n (%) 3,925 (53.9) 508 (52.2) 0.33 Race, n (%) Latino 3,669 (50.4) 503 (51.6) 0.47 White 1,371 (18.8) 134 (13.8) < 0.01 Black 527 (7.2) 91 (9.3) 0.02 Asian 453 (6.2) 70 (7.2) 0.27 Other/unknown 1,266 (17.4) 176 (18.1) 0.62 Admission type, n (%) Respiratory distress/failure 1,527 (21) 330 (33.9) < 0.01 Cardiovascular compromise 607 (8.3) 251 (25.8) < 0.01 Metabolic derangement 208 (2.9) 111 (11.4) < 0.01 Neurologic compromise 1,456 (20) 93 (9.5) < 0.01 Hemorrhage/coagulopathy 186 (2.6) 37 (3.8) 0.03 Postoperative 2,868 (39.4) 108 (11.1) < 0.01 Other 436 (6) 44 (4.5) 0.08 High-risk diagnoses, n (%) Respiratory failure/acute respiratory distress syndrome 721 (9.9) 310 (31.8) < 0.01 Sepsis/systemic inflammatory response syndrome 435 (6.0) 241 (24.7) < 0.01 Coagulopathy 169 (2.3) 163 (16.7) < 0.01 Cardiac disease (not postoperative) 417 (5.7) 160 (16.4) < 0.01 Leukemia/lymphoma 228 (3.1) 131 (13.4) < 0.01 Hepatic failure 203 (2.8) 103 (10.6) < 0.01 Cardiac arrest 100 (1.4) 102 (10.5) < 0.01 Stem cell transplantation 56 (0.8) 80 (8.2) < 0.01 Dehydration 119 (1.6) 75 (7.7) < 0.01 Pediatric Index of Mortality-2 risk of mortality percent at admission (IQR) 0.8 (0.4, 0.9) 1.2 (0.8, 3.9) < 0.01 Nonrenal OD score at admission (IQR) 1 (0, 11) 11 (1, 12) < 0.01 Maximum nonrenal OD score in first week (IQR) 10 (0, 11) 11 (10, 20) < 0.01 Length of stay, d (IQR) 2.3 (1.2, 4.9) 7.1 (2.5, 16.5) < 0.01 28-d mortality, n (%) 200 (2.7) 246 (25.3) < 0.01 AKI = acute kidney injury, IQR = interquartile range, OD = organ dysfunction. p values based on Yates-corrected chi-square test for categorical variables and Mann-Whitney U test for continuous variables. compared with patients who had persistent AKI, even after controlling for other OD. Our findings are in line with several adult studies which have shown an association between renal recovery and decreased mortality (9 13). Although ours is the first pediatric study to examine the relationship between the improvement of kidney function and mortality after AKI, there is evidence that worsening or lack of improvement of multiple OD is associated with poor prognosis. Leteurtre et al (17) studied a PICU population and showed that nonsurvivors were less likely than survivors to demonstrate improvements in their daily PELOD scores. Pediatric Critical Care Medicine www.pccmjournal.org 707

Sanchez-Pinto et al Table 3. Descriptive Statistics of Patients With Resolved Acute Kidney Injury Versus With Persistent Acute Kidney Injury After Their Peak Acute Kidney Injury in the ICU Variables Resolved AKI Persistent AKI p Total, n (%) 507 (52.1) 467 (47.9) Age, yr (IQR) 6.3 (1.4, 13.3) 7.6 (1.7, 14.7) 0.10 Weight, kg (IQR) 19.2 (9.3, 45) 22 (10.5, 50) 0.15 Males, n (%) 275 (54.2) 233 (49.9) 0.20 Race, n (%) Latino 263 (51.9) 240 (51.4) 0.93 White 68 (13.4) 66 (14.1) 0.82 Black 49 (9.7) 42 (9) 0.80 Asian 41 (8.1) 29 (6.2) 0.31 Other/unknown 86 (17) 90 (19.3) 0.39 Admission type, n (%) Respiratory distress/failure 182 (35.9) 148 (31.7) 0.19 Cardiovascular compromise 121 (23.9) 130 (27.8) 0.18 Metabolic derangement 55 (10.9) 56 (12) 0.65 Neurologic compromise 45 (8.9) 48 (10.3) 0.53 Hemorrhage/coagulopathy 14 (2.8) 23 (4.9) 0.11 Postoperative 70 (13.8) 38 (8.1) < 0.01 Other 20 (3.9) 24 (5.1) 0.46 High-risk diagnoses, n (%) Respiratory failure/acute respiratory distress syndrome 164 (47.8) 146 (45.5) 0.77 Sepsis/systemic inflammatory response syndrome 114 (29) 127 (37.4) 0.10 Coagulopathy 62 (13.9) 101 (27.6) < 0.01 Cardiac disease (not postoperative) 84 (19.9) 76 (19.4) 0.97 Leukemia/lymphoma 66 (15) 65 (16.2) 0.75 Hepatic failure 50 (10.9) 53 (12.8) 0.52 Cardiac arrest 31 (6.5) 71 (17.9) < 0.01 Stem cell transplantation 31 (6.5) 49 (11.7) 0.02 Dehydration 47 (10.2) 28 (6.4) 0.07 Pediatric Index of Mortality-2 risk of mortality percent at admission (IQR) 1.1 (0.8, 3.9) 1.6 (0.8, 3.9) < 0.01 Nonrenal OD score at admission (IQR) 11 (1, 12) 11 (1, 12) 0.77 Maximum nonrenal OD score in first week (IQR) 11 (10, 20) 11 (2, 20) 0.34 Length of stay, d (IQR) 10.9 (4.6, 22.1) 3.9 (1.5, 10.2) < 0.01 28-d mortality, n (%) 57 (11.2) 189 (40.5) < 0.01 AKI = acute kidney injury, IQR = interquartile range, OD = organ dysfunction. p values based on Yates-corrected chi-square test for categorical variables and Mann-Whitney U test for continuous variables. There are several potential explanations for our findings of an association between changes in AKI and mortality, independent of other OD. The kidney plays a central role in homeostatic control mechanisms, such as fluid balance, electrolyte management, acid-base status, vascular tone, and toxin clearance (21) that can pose a threat to other organs and the overall system viability when this homeostasis is acutely deranged (22, 23). The kidney s complex physiologic functions make the 708 www.pccmjournal.org October 2015 Volume 16 Number 8

Feature Articles Figure 3. Kaplan-Meier curves for four representative progression/ improvement groups based on the Kidney Disease Improving Global Outcomes acute kidney injury (AKI) staging criteria at admission, peak (highest stage achieved), and trough (lowest stage after the peak) attained during the ICU course. relationship between other OD and kidney injury a potentially two-way street: while AKI is generally accepted as a result of other OD mainly cardiovascular and respiratory it is possible that kidney injury also exacerbates or delays the recovery of other OD by the mechanisms mentioned above. Another explanation could be an unmeasured confounding variable. Although we have attempted to control for the most likely alternative explanation (i.e., other OD), there are other possible confounding variables. The specific etiology of the kidney injury and the inherent susceptibility of the patient may have a role in the progression and/or improvement of AKI and its associated mortality. This is supported by our finding of little to no difference in the severity of illness and nonrenal OD scores between patients with persistent or resolved AKI despite the large difference in mortality between these two groups. This is also similar to the results of the adult Beginning and Ending Supportive Therapy for the Kidney study, which showed that OD and severity-of-illness scores were not good predictors of AKI outcomes (24). Renal recovery is multifactorial, and the genetic components underlying the injury and its subsequent recovery might play a key role in the differences in outcomes after AKI (25 27). Future studies will hopefully shed further light on this area. Other pediatric studies have demonstrated an association between AKI and mortality in critically ill children, including the study by Schneider et al (1), which analyzed a subset of our patient population and found a mortality of 30% in patients with AKI (2, 3). Although these studies have consistently found an association between the presence of AKI and mortality, understanding the relationship of changes in kidney function and outcomes in the PICU could help us better define the populations at risk and guide the design of future preventive and therapeutic studies. Although our study has only demonstrated an association between AKI changes and mortality, it highlights the potential importance of predicting AKI progression. Prior studies have shown the ability to predict the development or progression of AKI by the use of clinical criteria such as the Renal Angina Index (28) or a standardized furosemide stress test (29). Our results provide justification that these and other AKI prediction methods should be studied to see if identifying patients at risk and preventing the development or progression of AKI is associated with reduced mortality. Our study has several limitations. First, it is a single-center retrospective study and relies on data collected electronically primarily for clinical care, and not research, which limits the completeness and accuracy of the data. Second, the study period spans 9 years, which increases the likelihood of changes in clinical practice affecting the outcomes of patients, as reflected by the drop in mortality seen in this cohort. However, this effect appeared to be independent of time trend in our multivariate models. Third, we used SCr, which is a late marker of AKI and limits our ability to do time-sequence analysis of AKI and other OD. SCr also has the potential to be diluted, especially in fluid overloaded patients, potentially underestimating the prevalence of AKI (30, 31). Fourth, we needed to impute baseline creatinine for many patients. We elected to use the upper limit of normal for age to bias toward the null, Table 4. Effect of Severity of Illness and Nonrenal Organ Dysfunction Scores AKI Progression/ Improvement Groups (Admit to Peak to Trough) n Unadjusted OR (95% CI) p Pediatric Index of Mortality-2 Adjusted OR (95% CI) p Nonrenal Organ Dysfunction Adjusted OR (95% CI) p No AKI 7,385 1 Baseline 1 Baseline 1 Baseline No AKI to AKI to no AKI 169 3.9 (2.4, 6.6) < 0.001 2.7 (1.6, 4.7) < 0.001 1.6 (0.9, 2.7) 0.09 No AKI to AKI to AKI 177 25.8 (18.6, 35.7) < 0.001 19.4 (13.6, 27.9) < 0.001 11.6 (8, 16.9) < 0.001 AKI to AKI to no AKI 188 4.4 (2.8, 7.1) < 0.001 2.4 (1.5, 4) 0.001 1.9 (1.2, 3.2) 0.01 AKI to AKI to AKI 341 17.1 (13.1, 22.2) < 0.001 10.5 (7.9, 14) < 0.001 10.4 (7.7, 14) < 0.001 AKI = acute kidney injury, OR = odds ratio. Multivariate logistic regression for the progression/improvement groups based on whether they had acute kidney injury as per the Kidney Disease Improving Global Outcomes acute kidney injury criteria at admission, peak (highest acute kidney injury stage), and trough (lowest acute kidney injury stage after the peak) attained during the first week of ICU course. Adjusted odds ratios with 95% CIs are reported after controlling for the Pediatric Index of Mortality-2 score at admission, maximum nonrenal organ dysfunction score during the first week of ICU stay, and year of admission. Pediatric Critical Care Medicine www.pccmjournal.org 709

Sanchez-Pinto et al but this may have caused us to underestimate the prevalence of AKI (32). Fifth, we did not have height information for most patients, precluding an actual glomerular filtration rate estimation for the classification of AKI. Sixth, we did not specifically study the etiology of the kidney injury, which could be associated with outcome. Finally, our nonrenal OD score, even though it is based on a validated scoring system and had excellent discrimination of ICU mortality in our cohort, has not been independently validated in other studies. CONCLUSION The development or progression of AKI in the PICU is independently associated with mortality. In addition, patients who have improvement in their SCr after AKI have significantly lower mortality than those who do not resolve their AKI. In most cases, these effects do not appear to be explained by the severity of illness at admission or the degree of other OD during the first week of ICU stay. These findings are intriguing and deserve further study. Whether AKI is a modifiable risk factor that if prevented or improved can decrease the degree of other OD or reduce mortality remains to be seen. REFERENCES 1. 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 2. Akcan-Arikan A, Zappitelli M, Loftis LL, et al: Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 2007; 71:1028 1035 3. 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 4. Askenazi D: Evaluation and management of critically ill children with acute kidney injury. Curr Opin Pediatr 2011; 23:201 207 5. Hsu CW, Symons JM: Acute kidney injury: Can we improve prognosis? Pediatr Nephrol 2010; 25:2401 2412 6. Goldstein SL, Chawla LS: Renal angina. Clin J Am Soc Nephrol 2010; 5:943 949 7. Mehta RL, Kellum JA, Shah SV, et al; Acute Kidney Injury Network: Acute Kidney Injury Network: Report of an initiative to improve outcomes in acute kidney injury. Crit Care 2007; 11:R31 8. Goldstein SL: Acute kidney injury in children: Prevention, treatment and rehabilitation. Contrib Nephrol 2011; 174:163 172 9. Pannu N, James M, Hemmelgarn B, et al; Alberta Kidney Disease Network: Association between AKI, recovery of renal function, and long-term outcomes after hospital discharge. Clin J Am Soc Nephrol 2013; 8:194 202 10. Schiffl H: Renal recovery after severe acute renal injury. Eur J Med Res 2008; 13:552 556 11. Bagshaw SM, Laupland KB, Doig CJ, et al: Prognosis for long-term survival and renal recovery in critically ill patients with severe acute renal failure: A population-based study. Crit Care 2005; 9:R700 R709 12. Swaminathan M, Hudson CC, Phillips-Bute BG, et al: Impact of early renal recovery on survival after cardiac surgery-associated acute kidney injury. Ann Thorac Surg 2010; 89:1098 1104 13. Lassnigg A, Schmidlin D, Mouhieddine M, et al: Minimal changes of serum creatinine predict prognosis in patients after cardiothoracic surgery: A prospective cohort study. J Am Soc Nephrol 2004; 15:1597 1605 14. Kidney Disease: Improving Global Outcomes Acute Kidney Injury Work Group: KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl 2012; 2:1 138 15. Schwartz GJ, Haycock GB, Spitzer A: Plasma creatinine and urea concentration in children: Normal values for age and sex. J Pediatr 1976; 88:828 830 16. Bellomo R, Ronco C, Kellum JA, et al; Acute Dialysis Quality Initiative workgroup: Acute renal failure definition, outcome measures, animal models, fluid therapy and information technology needs: The Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004; 8:R204 R212 17. Leteurtre S, Duhamel A, Grandbastien B, et al: Daily estimation of the severity of multiple organ dysfunction syndrome in critically ill children. CMAJ 2010; 182:1181 1187 18. Garcia PC, Eulmesekian P, Branco RG, et al: External validation of the paediatric logistic organ dysfunction score. Intensive Care Med 2010; 36:116 122 19. Leteurtre S, Martinot A, Duhamel A, et al: Development of a pediatric multiple organ dysfunction score: Use of two strategies. Med Decis Making 1999; 19:399 410 20. Slater A, Shann F, Pearson G; Paediatric Index of Mortality (PIM) Study Group: PIM2: A revised version of the Paediatric Index of Mortality. Intensive Care Med 2003; 29:278 285 21. Lote CJ: Principles of Renal Physiology. Fifth Edition. New York, Springer, 2012 22. Basu RK, Devarajan P, Wong H, et al: An update and review of acute kidney injury in pediatrics. Pediatr Crit Care Med 2011; 12:339 347 23. Elapavaluru S, Kellum JA: Why do patients die of acute kidney injury? Acta Clin Belg Suppl 2007; 2:326 331 24. Uchino S, Bellomo R, Morimatsu H, et al; Beginning and Ending Supportive Therapy for the Kidney (B.E.S.T. Kidney) Investigators: External validation of severity scoring systems for acute renal failure using a multinational database. Crit Care Med 2005; 33:1961 1967 25. Fekete A, Treszl A, Tóth-Heyn P, et al: Association between heat shock protein 72 gene polymorphism and acute renal failure in premature neonates. Pediatr Res 2003; 54:452 455 26. Treszl A, Tóth-Heyn P, Kocsis I, et al: Interleukin genetic variants and the risk of renal failure in infants with infection. Pediatr Nephrol 2002; 17:713 717 27. Goldstein SL, Chawla L, Ronco C, et al: Renal recovery. Crit Care 2014; 18:301 28. Basu RK, Zappitelli M, Brunner L, et al: Derivation and validation of the renal angina index to improve the prediction of acute kidney injury in critically ill children. Kidney Int 2014; 85:659 667 29. Chawla LS, Davison DL, Brasha-Mitchell E, et al: Development and standardization of a furosemide stress test to predict the severity of acute kidney injury. Crit Care 2013; 17:R207 30. Basu RK, Andrews A, Krawczeski C, et al: Acute kidney injury based on corrected serum creatinine is associated with increased morbidity in children following the arterial switch operation. Pediatr Crit Care Med 2013; 14:e218 e224 31. Liu KD, Thompson BT, Ancukiewicz M, et al; National Institutes of Health National Heart, Lung, and Blood Institute Acute Respiratory Distress Syndrome Network: Acute kidney injury in patients with acute lung injury: Impact of fluid accumulation on classification of acute kidney injury and associated outcomes. Crit Care Med 2011; 39:2665 2671 32. Zappitelli M, Parikh CR, Akcan-Arikan A, et al: Ascertainment and epidemiology of acute kidney injury varies with definition interpretation. Clin J Am Soc Nephrol 2008; 3:948 954 710 www.pccmjournal.org October 2015 Volume 16 Number 8