a b s t r a c t

Background: Identifying Acute kidney injury early can inform medical decisions key to mitigation of injury. An AKI risk stratification tool, the renal angina index (RAI), has proven better than creatinine changes alone at predicting AKI in critically ill children.

Objective: To derive and test performance of an “acute” RAI (aRAI) in the Emergency Department (ED) for predic- tion of inpatient AKI and to evaluate the added yield of urinary AKI biomarkers.

Methods: Study of pediatric ED patients with sepsis admitted and followed for 72 h. The primary outcome was inpatient AKI defined by a creatinine N1.5x baseline, 24-72 h after admission. Patients were denoted renal angina positive (RA+) for an aRAI score above a population derived cut-off. Test characteristics evaluated predictive per- formance of the aRAI compared to changes in creatinine and incorporation of 4 urinary biomarkers in the context of renal angina were assessed.

Results: 118 eligible subjects were enrolled. Mean age was 7.8 +- 6.4 years, 16% required intensive care admission. In the ED, 27% had a +RAI (22% had a N50% creatinine increase). The aRAI had an AUC of 0.92 (0.86-0.98) for pre- diction of inpatient AKI. For AKI prediction, RA+ demonstrated a sensitivity of 94% (69-99) and a negative pre- dictive value of 99% (92-100) (versus sensitivity 59% (33-82) and NPV 93% (89-96) for creatinine >=2x baseline). Biomarker analysis revealed a higher AUC for aRAI alone than any individual biomarker.

Conclusions: This pilot study finds the aRAI to be a sensitive ED-based tool for ruling out the development of in- hospital AKI.

(C) 2020

1. Introduction

Acute Kidney Injury is a common complication in hospitalized children with recent studies showing the incidence to be nearly 30% in the pediatric intensive care unit and 5% in non-critically ill, hospitalized children [1,2]. Children with AKI have an increased risk of death, inde- pendent of underlying disease pathology [2]. Those children who sur- vive have an increased risk of chronic kidney disease, known to be

associated with cardiovascular morbidity, anemia, and growth failure, and hypertension [3,4]. Many interventions, such as medications, proce- dures, and fluid-overload, in the hospital-setting are known to be neph- rotoxic and exacerbate AKI [5-7]. Recognizing that the Emergency Department (ED) is often the patient’s first exposure to hospital care, it is an opportune setting to identify children at risk of AKI and to begin limiting nephrotoxic interventions while offering renal support as necessary.

Abbreviations: AKI, Acute Kidney Injury; aRAI, acute Renal Angina Index; ED, Emergency Department; IL-18, Interleukin 18; KDIGO, Kidney Disease Improving Global Outcomes; KIM- 1, kidney injury molecule-1; L-FABP, liver fatty acid binding protein; NGAL, neutrophil gelatinase-associated lipocalin; RA, Renal Angina; RAI, Renal Angina Index; SCr, serum creatinine.

* Corresponding author at: Division of Pediatric Emergency Medicine, Monroe Carell Jr Children’s Hospital at Vanderbilt, 2200 Children’s Way, VCH B-319, Nashville, TN 37232-9001, United States of America.

E-mail addresses: [email protected] (H.R. Hanson), [email protected] (M.A. Carlisle), [email protected] (R.S. Bensman), [email protected] (T. Byczkowski), [email protected] (H. Depinet), [email protected] (R. Knox), [email protected] (S.L. Goldstein), [email protected] (R.K. Basu).

¹ Division of Pediatric Emergency Medicine, Monroe Carell Jr. Children’s Hospital at Vanderbilt, Vanderbilt University Medical Center, 2200 Children’s Way, VCH-B-319, Nashville, TN 37232.

² Division of Pediatric Critical Care Medicine, Children’s Hospital Colorado, 13,123 East 16th Ave, Aurura, CO 80045.

³ Division of Pediatric Critical Care Medicine, Children’s Healthcare of Atlanta, Emory School of Medicine, 1405 Clifton Rd NE, Atlanta, GA 30322.

https://doi.org/10.1016/j.ajem.2020.01.046

0735-6757/(C) 2020

Currently the standard measure of AKI in the ED is the elevation of serum creatinine (SCr). There is a large reliance on laboratory “normals” to help identify SCr values outside of an acceptable range and, therefore, indicate a diagnosis of AKI. This method can result in an under diagnosis of AKI as SCr, dependent on muscle mass, varies greatly between chil- dren of the same age making these laboratory “normals” unreliable [8,9]. This laboratory test is performed often without adjudication of AKI risk. Recent literature has proposed a single definition for AKI, termed the KDIGO (Kidney Disease Improving Global Outcomes) criteria, that utilizes changes in SCr to diagnose AKI [10]. Any change in SCr that is 1.5 times above the patient’s baseline SCr (defined as the lowest SCr in the last 6 months) is consistent with a diagnosis of AKI. In a recent study, only 8% of children were found to have had a SCr mea- sured in the 6 months prior to their ED visit, therefore, leaving the pro- vider without a baseline SCr for comparison [11]. This highlights the difficulty in comprehensively identifying AKI in children in the ED.

Proper triage of patients by AKI risk may ultimately facilitate a

targeted approach to initial management in the ED. Similar to stroke and acute coronary syndrome, AKI is associated with known risk factors. A combination of risk with signs of injury has led to algorithms for rapid management of both stroke and acute coronary syndrome. These syn- dromes inspired a parallel model of context-driven adjudication of early injury signs that was used to create the renal angina prodrome. Tested first in critically ill children, the concept of renal angina (RA) demonstrates a higher predictive sensitivity and discrimination for se- vere AKI than risk factors or signs of injury alone [12]. The score for assessing RA, the renal angina index (RAI), is a simple multiplication of patient risk by creatinine change (Fig. 1) [13,14]. In initial validation testing, biomarker assessment in patients positive for RA (RAI >= 8) fur- ther optimized prediction of severe AKI. This construct is similar to how cardiac enzyme testing is utilized in patients presenting with chest pain. In acute coronary syndrome, troponin I efficiently and accu- rately aids in diagnosis in the appropriate patient context. When used in a heterogeneous population without further discrimination for acute coronary syndrome it becomes much less useful [15,16].

We hypothesized that a modification of the original RAI to include “acute” components, objectively available in the ED (Fig. 1), would facil- itate the prediction of in-hospital AKI among children with possible sep- sis. Recognizing the limitations of SCr alone, this tool would allow application of a risk factor stratification to help further discriminate change in SCr and aid in AKI prediction. The aims of this study were

(1) to compare the performance of the “acute” RAI (aRAI) with SCr alone to predict inpatient AKI and (2) to evaluated the ability of urinary

AKI biomarkers, used with the aRAI, to enhance prediction of inpatient AKI.

1. Methods
   1. Study design and setting

This is a prospective, observational cohort study of children aged N28 days to b25 years performed at a single, high-volume, tertiary pedi- atric hospital. Enrollment began 8/1/2015 and concluded 5/9/2016. The hospital’s institutional review board approved this study with a waiver of informed consent.

• 1. Selection of participants

Patients were included in this study if they presented to the hospital’s ED, had a concern for sepsis, were admitted to the hospital, and had both a SCr measured and urine specimen obtained as part of routine ED care. Subjects with “concern for sepsis” were identified by a positive electronic health record automatic sepsis alert (screening tool based on vital signs, perfusion variables and high risk underlying condition; or hypotension); this screening tool had been implemented several years prior to this study and was developed as part of a national quality improvement collaborative [17]. As this is a screening tool, sub- jects were also required to have received fluid resuscitation (>=10 ml/kg isotonic intravenous fluid) as part of their ED coarse as this would fur- ther indicate a higher concern for possible shock. Excluded subjects were those who had previously been enrolled in this study, had a his- tory of chronic kidney disease stage IV or V, were anephric, or were ac- tively receiving dialysis. A report from the automated tool was generated each morning identifying all patients seen in the ED in the prior 24 h with a positive screen. These potential subjects were further reviewed for inclusion criteria and then, if deemed appropriate, were followed prospectively through their hospital course.

• 1. Methods and measurements

For each eligible subject, a manual chart review of the electronic medical record was performed by three of the authors (HRH, MAC, RSB) using a detailed manual of operations. Abstracted data was cap- tured using REDCap (Research Electronic Data Capture) tool hosted at Cincinnati Children’s Hospital Medical Center [18]. To increase validity and reliability, 5% of the charts were separately reviewed and any

Fig. 1. Description of the acute renal angina index contrasted to the original renal angina index.

discrepancies in abstracted data prompted a discussion such that all dis- agreement was resolved by consensus review. Variables including de- mographics, past medical history, administered intravenous fluids, procedures, length of stay, disposition, and all measured SCr values were abstracted for each eligible subject. Subjects were followed for the first 72 h after hospital admission or until discharge, whichever oc- curred first. SCr was ordered at the discretion of the patient’s admitting physician, was not specifically obtained for research purposes, and, therefore, was not necessarily present each day. The last SCr measured in each 24-h block after admission, was recorded. For example, if the pa- tient was admitted at midnight on a Sunday then the SCr measurement closest to midnight on Monday was recorded.

A Urine sample (catheter, clean-catch, or bag) was obtained for bio- marker testing from left-over urine that was collected as part of the sub- jects’ ED course. Urine was centrifuged and stored in cryovials at minus 80 ?C, according to standard laboratory procedure, until the time of bio- marker analysis.

• 1. Renal angina

The modifications made to the original RAI, in forming the aRAI, are depicted in Fig. 1. These modifications were necessary in order to estab- lish an index that could be applied early in the ED course. The derivation of the original variables in the RAI is explained in the manuscript by Basu, et al [12]. The risk strata variables in the aRAI were intended to parallel the variables in the original RAI and incorporate variables al- ready present in the automated sepsis alert.

RA was calculated for each enrolled subject while in the ED. Subjects were classified as RA(+) and RA(-) based on the score obtained using the RAI. For the risk component, the number correlating with the highest risk category that the subject fulfilled was used (Fig. 1). For ex- ample, if the subject had both a history of solid organ transplant and was intubated in the ED they received a score of 5. The score from the risk component was multiplied by the score from the injury component to produce the aRAI score. The aRAI ranges from 1 to 40 and, based on pre- vious derivation of the original RAI, an aRAI score >=8 was used to define RA(+) [12].

• 1. AKI biomarkers

Urinary Neutrophil Gelatinase-Associated Lipocalin , kidney injury molecule-1 (KIM-1), interleukin 18 (IL-18), and liver fatty acid binding protein (L-FABP) were each analyzed in the Center for Acute Care Nephrology Biomarker Core Laboratory by a laboratory technician with no knowledge of study outcomes. Urinary NGAL was assayed using a human-specific commercially available enzyme-linked immunosor- bent assay (ELISA, AntibodyShop, Grusbakken, Denmark). IL-18 and L- FABP were measured using commercially available ELISA kits (Medical & Biological Laboratories Co., Nagoya, Japan, and CMIC Col., Tokyo, Japan, respectively) per manufacturer’s instructions. Urine KIM-1 was measured by ELISA using commercially available reagents (R&D Sys- tems, Inc., Minneapolis, Minnesota). The a priori cut-off values for each biomarker were determined by sensitivity analysis in the initial RAI data sets [12].

• 1. Outcomes

The primary outcome of this study was AKI anytime between 24 and 72 h after admission from the ED as defined by KDIGO Stage I-III by el- evated SCr (any SCr N1.5 times baseline) [10]. This outcome measure was termed AKI_Inpt. The comparison group was all patients in the hospi- tal at 24-72 h after admission who were not identified as having AKI (ei- ther they had a SCr that was unchanged or they did not have a repeat SCr obtained). Baseline creatinine, necessary for KDIGO staging, was de- fined as the lowest measured SCr in the previous 6 months. For all sub- jects without a baseline SCr from prior laboratory assessment, an

accepted alternative approach was used to impute a baseline SCr [19]. The imputation was performed using the Schwartz formula which takes into account height and assumes a normal creatinine clearance of 120 ml/min/1.73 m² [19,20].

• 1. Sample size

We assumed a 10% incidence of AKI in the reported pediatric popu- lation with shock and an area under the curve (AUC) for the RAI of 0.7-0.8 based on previous work [12,21]. To detect a difference in AUC by 0.10, and a two-sided test at 0.05 alpha, 116 subjects provided 80% power. For a planned 116 total patients, we expected to have 12 patients with AKI and 104 without.

• 1. Analysis

All statistical analysis was made using the software packages SAS version 9.3 (SAS Institute, Inc., Cary, NC) and StataSE version 14 (StataCorp, College Station, TX). Descriptive statistics were used to iden- tify and describe the population of children with RA in the ED. Differ- ences between those who were RA(+) and RA(-) were assessed using chi-square or fisher’s exact test and t-tests for categorical and con- tinuous variables, respectively.

The aRAI was evaluated as a diagnostic test and compared to un- stratified changes in SCr measured in the ED. Sensitivity, specificity, pos- itive predictive value (PPV), negative predictive value (NPV), and receiver-operating characteristic (ROC) analysis were calculated to evaluate the Predictive performance of an aRAI >=8 (RA definition). AUC was calculated for the model. A sensitivity analysis was performed excluding all subjects who did not have a SCr measured during inpatient admission. Based on the original derivation and validation studies of the RAI in the critical care setting, we hypothesized an AUC N 0.80 and a NPV N 90% [12]. Simple and Multivariable logistic regression models were used to predict AKI_Inpt using aRAI and ED obtained, patient- related variables. Test characteristics of each biomarker were also assessed, biomarker concentrations were used as continuous variables, and the increase in AUC was calculated for each model using DeLong’s method [22]. Classification analysis by sequential iteration was per- formed to separate the entire cohort into terminal node cohorts with in- dividual probability and risks for the primary outcome (derivation of classification and regression tree analysis). A p b 0.05 was considered significant for all analysis.

1. Results
   1. Characteristics of study subjects

There were 128 subjects initially identified for potential study en- rollment based on information in the automated sepsis alert trigger. After chart review was initiated 10 subjects were found to be ineligible (Fig. 2). Therefore, 118 individual subjects were included in data analy- sis. Table 1 describes the patient demographics of all subjects in the en- rolled cohort, as well as, a comparison of those with and without RA in the ED. Of all included subjects, 52% were female, 63% were white, and the mean age was 7.8 +- 6.4 years. 16% were admitted to the intensive care unit. 69% of subjects were still admitted after 48 h and 47% were still admitted after 72 h. There were no deaths during the first 72 h in any patient in this study and no patients required renal replacement therapy.

• 1. Elevated SCr in the ED

There were 26/118 (22%) of subjects in the ED with a N1.5x increase in creatinine above baseline. Of these, 50% were stage 1 (1.5x increase in SCr from baseline), 38% were stage 2 (2x increase in SCr from base- line), and 12% were stage 3 (3x increase in SCr from baseline). 41

Fig. 2. Study flowchart.

subjects (34%) had no SCr measured after the initial measurement in the ED, of these none had an elevated SCr in the ED. The rate of AKI_Inpt was 17/81 (21%). Of the 81 subjects who remained in the hospital after 24 h, 25 subjects had no repeat SCr after their initial ED SCr. These subjects were assumed to have no AKI for primary analysis.

Table 1

Demographic and clinical information.

Characteristic, n (%)	Total cohort	aRA(+) n = 32	aRA(-) n = 86	p-value
	n = 118
Age, years (mean +- SD)	7.8 +- 6.4	7.6 +- 6.0	7.8 +- 6.6	0.873
body surface area (mean +- SD)	0.97 +- 0.6	0.93	0.98	0.685
		+- 0.4	+- 0.6
Female	61 (52)	21 (66)	40 (47)	0.065
Race				0.048
White	74 (63)	16 (50)	58 (68)
Black	19 (16)	6 (19)	13 (15)
Other	21 (18)	9 (28)	12 (14)
Unknown	4 (3)	1 (3)	3 (3)

Risk strata b0.001

Moderate	75 (64)	5 (16)	70 (82)
High	24 (20)	12 (37)	12 (14)
Very High	19 (16)	15 (47)	4 (4)
Baseline SCr present	72 (61)	30 (94)	42 (49)	b0.001
Transplant history	8 (7)	6 (19)	2 (2)	0.002
History of oncologic disease	24 (20)	13 (41)	11 (13)	0.001
History of AKI (n = 114)	6 (5)	4 (14)	2 (2)	0.017
History of CKD (n = 114)	6 (5)	4 (14)	2 (2)	0.017
ED intubation	2 (2)	1 (3)	1 (1)	0.463
N 40 mL/kg fluid in ED	17 (14)	14 (44)	3 (4)	b0.001
ICU admission	19 (16)	11 (34)	8 (9)	0.001
ICU length of stay				0.952
<=24 h	6 (32)	4 (36)	2 (25)
N72 h	2 (11)	4 (36)	3 (38)
OR during 1st 72 h	6 (5)	1 (3)	5 (6)	0.555
ED admission Dx
Shock/serious infection	109 (92)	31 (97)	78 (91)	0.261
Medical cardiac	3 (3)	2 (6)	1 (1)	0.119
Respiratory illness	16 (14)	3 (9)	13 (15)	0.418
Post-surgical/minor	1 (1)	0	1 (1)	0.540
trauma/ortho
CNS dysfunction	11 (9)	0	11 (13)	0.034
Pain/sedation management	1 (1)	0	1 (1)	0.540
AKIInpt (n = 81)	17 (21)	16 (50)	1 (1)	b0.001
Discharged home in b72 h	62 (53)	12 (38)	50 (58)	0.050

1. aRAI

aRA(+) occurred in 32/118 (27%) patients while in the ED. Com- pared with patients who were aRA(-), patients with aRA(+) were more likely to have received N40 ml/kg of isotonic fluid in the ED, been admitted to the intensive care unit, and to have required a longer than 72-h hospital admission (Table 1). Additionally, these subjects were more likely to have had a history of AKI or chronic kidney disease. There were 46/118 (39%) subjects in the cohort who required their baseline SCr to be imputed because they did not have a previous SCr measured in the 6 months prior to their ED visit. Only 2/46 of these sub-

jects were RA(+) in the ED and none had AKI_Inpt.

1. aRAI versus context free creatinine increases

The aRAI had a NPV of 0.99 (0.92-1.00) and AUC of 0.92 (0.86-0.98)

for the prediction of AKI_Inpt (Table 2). Sensitivity was 94% for the aRAI as compared to 59% for an elevation in SCr noted to be at least two times greater than baseline while in the ED (consistent with the consensus definition for severe AKI). Additionally, Table 3 shows that aRA fulfill- ment is the only variable, of the variables tested, that is independently associated with AKI_Inpt.

1. aRAI and urine biomarkers

AUC estimates from both the aRAI and individual urine biomarkers obtained in the ED revealed a higher AUC for aRAI alone than any indi- vidual biomarker, followed by NGAL (Table 4). Additionally, the classifi- cation analysis identified a terminal node of RA(+)/NGAL(+) with a probability of inpatient AKI of 60% higher than any of the other nodes.

1. Discussion

Interpreting elevations in SCr without a Clinical context in children who present to the ED is challenging. Not only can it be difficult to make a diagnosis of AKI, but it is also difficult to know how to tailor medical interventions in light of this elevation. A tool that accounts for clinical risk and accurately screens for inpatient AKI in the ED would better inform early changes in medical inventions in order to help mit- igate further damage. The aRAI is a sensitive and practical stratification model, and the first of its kind, that improves the ability to identify chil- dren in the ED at the highest risk for AKI during hospital admission.

Table 2

aRAI compared to elevations in creatinine in the ED for prediction of inpatient AKI

	n	Sensitivity	Specificity	PPV	NPV	+LR
aRAI (+)	32	94	84	50	99	5.9
	(27%)	(71-100)	(76-91)	(39-61)	(92-100)	(3.7-9.5)
aRAI (-)	86	6	16	1	50	0.1
	(73%)	(0-29)	(9-25)	(0-7)	(39-61)	(0.0-0.5)
SCr <= 1x	35	0	65	0	79	0
	(30%)	(0-2)	(55-75)		(77-81)
SCr N 1 to b1.5	57	12	44	4	75	0.2
	(48%)	(1-36)	(35-55)	(1-12)	(69-80)	(0.1-0.8)
SCr >= 1.5 to b2.0	13	29	92	38	89	3.7
	(11%)	(10-56)	(85-97)	(19-63)	(85-91)	(1.4-10.0)
SCr >= 1.5	26	88	89	58	98	8
	(22%)	(64-99)	(81-94)	(43-71)	(92-99)	(4.5-14.5)
SCr >= 2.0	13	59	97	77	93	19.8
	(11%)	(33-82)	(92-100)	(51-92)	(89-96)	(6.1-64.7)

All serum creatinine (SCr) values are represented as comparisons to baseline SCr; data are presented as percentages (95% confidence interval).

Table 3

Predictors of AKI during admission using multivariable logistic regression

Variable Odds Ratio (95% CI) p-value

Acute renal angina	1.177 (1.077, 1.286)	0.0003
Age	1.038 (0.921, 1.171)	0.5385
Prior history of AKI	0.820 (0.600, 1.127)	0.0614
ICU admission	3.207 (0.375, 27.422)	0.2872

In this study, we compared the utility of the aRAI to the current ED standard for AKI diagnosis, elevations in SCr. We found that using the aRAI improves precision for the prediction of inpatient AKI compared to using elevated SCr alone. In this study, 83 patients had a SCr N1x baseline and only 17 had inpatient AKI, therefore 80% of the time, when using SCr alone, there is an inaccurate prediction of inpatient AKI. Conversely, only 32 patients in the ED were RA(+) and 16 had in- patient AKI which reduced the error rate to 50%, or the correct predic- tion goes from 20% to 50%, a 150% increase in precision. Additionally, the aRAI was found to have a NPV of 99%, similar to the NPV of 92-99% seen in the original RAI, making this scoring system extremely effective at ruling out the potential for AKI during admission [12]. In an ED setting where resource utilization is of the utmost importance, this test, with high sensitivity and increased precision, would allow for more effective triaging of patients, both in delegation of effort and in- vestment of resources. Given the recent emphasis on reducing medical waste, this proves significant.

The original RAI used on the day of intensive care admission had an

AUC of 0.74-0.81 for predicting AKI on day 3 of admission and outperformed the current KDIGO criteria [12]. Since the original deriva- tion and validation multiple studies have demonstrated similar results including a multicenter, multinational study, where the RAI used in chil- dren on admission to the intensive care unit predicted AKI on day 3 of hospitalization better than SCr with a relative risk of 1.61 (p b 0.0001) [23,24]. The RAI has been validated both in developed and in Developing countries and has shown similar predictive results [25,26]. Additionally, resource-limited settings have highlighted its ease of use [25]. This

index has been modified for both a post-cardiac surgery and an adult population and has shown utility as a potential screening test [27,28].

In our study, children with possible sepsis who were aRA(+) were more likely to receive high volumes of crystalloid for resuscitation in the ED and be admitted to an intensive care unit. This is likely a reflec- tion of the illness severity in children with AKI. In a recent study of chil- dren admitted from the ED after having a SCr measured there was a 2% mortality rate in children with AKI, compared to 0.4% in the non-AKI population [11]. These children were also more likely to receive isotonic fluids and be admitted to an intensive care unit [11]. In our study, the aRAI was the only independent variable tested that was associated with progression to AKI. This is likely two-fold, the aRAI considers many factors, similar to an illness severity score, and therefore provides an opportunity for a more inclusive prediction. Secondly, the study was not specifically powered for this analysis and more subjects may have changed this result. It is hypothesized that the negative association be- tween history of AKI and current AKI is likely secondary to this factor.

This study used an automated sepsis alert trigger that had been em-

bedded in the electronic medical record to identify a high-risk popula- tion, those with the potential for sepsis, for earlier ED care. This alert is used to both heighten physician awareness of a potentially ill patient and provide an opportunity for medical intervention at the earliest pos- sible time point. The electronic medical record has been used, similarly, to predict nephrotoxin-mediated AKI before it occurs on Inpatient units, and has been the topic of a large consensus group within Nephrology [29-31]. In developed countries one of the leading causes of AKI is sepsis [32]. Incorporating the aRAI in the electronic medical record has the po- tential to aid in earlier recognition of children at risk for AKI. Addition- ally, the strong association of sepsis with AKI provides a unique opportunity for the merging of a dual-use tool for prediction of sepsis and AKI in the ED setting.

The RA paradigm provides an opportunity to delineate a high-risk population on which kidney injury biomarkers may prove more useful. In prior study, the incorporation of AKI biomarkers with the RAI im- proved discrimination for AKI [33]. When evaluating each of the 4 bio- markers, we found that no one biomarker had an AUC that was higher than the aRAI alone, however NGAL had the best performance. Our

Table 4

aRAI and Biomarkers Alone for Prediction of Inpatient AKI.

Test cut-off value	aRAI 8.0	NGAL 21.5	KIM-1 939.9	IL-18 109.4	L-FABP 8.7
Sensitivity	94 (71-100)	77 (55-93)	59 (33-82)	59 (33-82)	59 (33-82)
Specificity	84 (76-91)	61 (48-73)	61 (48-73)	53 (40-66)	58 (45-70)
PPV	50 (39-61)	34 (26-44)	29 (20-40)	25 (17-35)	27 (18-38)
NPV	99 (92-100)	91 (80-96)	85 (75-91)	83 (72-90)	84 (74-91)
AUC-ROC	0.92	0.71	0.65	0.49	0.57
	(0.86-0.98)	(0.57-0.84)	(0.49-0.80)	(0.34-0.64)	(0.42-0.73)

Data are presented as the percentage (95% confidence interval).

classification tree did demonstrate that the additional of NGAL(+) pa- tients to RA(+) ED patients proved to have the highest rate of inpatient AKI. This construct may be used as foundation for further research that provides an individualized approach to the care of children at risk for AKI. For instance, an automated alert in the electronic medical record fires indicating a patient is at risk for sepsis, the patient is found to be RA(+), NGAL is tested and is negative and now we have a patient who may more safety receive Contrast agents. Conversely, the patient is found to be RA(+) and NGAL(+) and we choose ultrasound as imag- ing modality over CT, when possible.

This study has several limitations. First, it is an observational study and as such causation cannot be assessed. The retrospective nature of chart review poses risks of Inaccurate data, errors in data extraction, and incomplete data. The authors attempted to minimize this by reviewing 5% of all charts for accuracy. This study was a pilot study per- formed at a single, tertiary care center and therefore the results may not be generalizable in all healthcare populations. Additionally, the study population was children with possible sepsis and the results may not be generalizable to a more heterogeneous population. The aRAI was de- veloped with modification from the original RAI and was intended as a first step towards developing a standardized risk stratification model however more work will be required to adjudicate and refine this model. Imputation of SCr using the Schwartz formula may not be an ac- curate measure of each individuals’ baseline SCr and thus may have skewed the results. Additionally, not all subjects had a SCr performed on each day of admission so it is possible that there was a higher rate of AKI than this study identified due to incomplete information.

This study has direct advantages for informing the use of nephro-

toxic medications, for example radiocontrast agents for imaging studies, and volume used for fluid resuscitation. The use of the aRAI provides an opportunity to identify a high-risk population of children and poten- tially at the earliest time point in medical care. This study begins an im- portant discussion on applications to further mitigate injury. This is the first step in a more individualized approach to pediatric AKI.

In conclusion, in this pilot study, the aRAI was shown to be a sensi-

tive test that can be used in the ED and that outperforms using a change in SCr to predict AKI 24 h after admission to the hospital. This test has promise for potential application into the electronic medical record as an automated trigger tool. In future study, this tool should be evaluated in a broadened, heterogenous population in the ED and the prediction performance of the tool with biomarker use additionally explored.

Funding

This work was supported by the Arnold W. Strauss Fellow Award at Cincinnati Children’s Hospital Medical Center.

Prior presentation of work

American Society of Nephrology Annual Conference, November 2016.

CRediT authorship contribution statement

Holly R. Hanson:Conceptualization, Methodology, Investigation, Formal analysis, Funding acquisition, Visualization, Writing - original draft, Project administration.Michael A. Carlisle:Investigation, Writing

- review & editing.Rachel S. Bensman:Investigation, Writing - review & editing.Terri Byczkowski:Formal analysis, Data curation, Writing - re- view & editing.Holly Depinet:Software, Resources, Writing - review & editing.Tara C. Terrell:Project administration, Investigation, Writing - review & editing.Hilary Pitner:Investigation, Writing - review & editing.Ryan Knox:Investigation, Writing - review & editing.Stuart L. Goldstein:Supervision, Writing - review & editing.Rajit K. Basu:Super- vision, Funding acquisition, Conceptualization, Methodology, Writing - original draft.

Acknowledgements

The authors of this study would like to thank the Center for Acute Care Nephrology Biomarker Core Laboratory and statistician Huaiyu Zang, both at Cincinnati Children’s Hospital Medicine Center, for assis- tance in obtaining biomarker data.

References

1. McGregor TL, Jones DP, Wang L, et al. Acute kidney Injury incidence in noncritically ill hospitalized children, adolescents, and young adults: a retrospective observa- tional study. Am J Kidney Dis 2016;67(3):384-90.
2. Kaddourah A, Basu RK, Bagshaw SM, Goldstein SL, Investigators A. Epidemiology of acute kidney injury in critically ill children and young adults. N Engl J Med 2017; 376(1):11-20.
3. Askenazi DJ, Feig DI, Graham NM, Hui-Stickle S, Goldstein SL. 3-5 year longitudinal follow-up of pediatric patients after acute renal failure. Kidney Int 2006;69(1): 184-9.
4. Sigurjonsdottir VK, Chaturvedi S, Mammen C, Sutherland SM. Pediatric acute kidney injury and the subsequent risk for chronic kidney disease: is there cause for alarm? Pediatr Nephrol 2018;33(11):2047-55.
5. Goldstein SL, Kirkendall E, Nguyen H, et al. Electronic health record identification of nephrotoxin exposure and associated acute kidney injury. Pediatrics 2013;132(3): e756-67.
6. Karcaaltincaba M, Oguz B, Haliloglu M. Current status of contrast-induced nephrop- athy and nephrogenic systemic fibrosis in children. Pediatr Radiol 2009;39(Suppl. 3):382-4.
7. Goldstein SL. fluid management in acute kidney injury. J Intensive Care Med 2014; 29(4):183-9.
8. Boer DP, de Rijke YB, Hop WC, Cransberg K, Dorresteijn EM. Reference values for serum creatinine in children younger than 1 year of age. Pediatr Nephrol 2010;25 (10):2107-13.
9. Ceriotti F, Boyd JC, Klein G, et al. Reference intervals for serum creatinine concentra- tions: assessment of available data for global application. Clin Chem 2008;54(3): 559-66.
10. Khwaja A. KDIGO clinical practice guidelines for acute kidney injury. Nephron Clin Pract 2012;120(4):179-84.
11. Hanson HR, Babcock L, Byczkowski T, Goldstein SL. Describing pediatric acute kidney injury in children admitted from the emergency department. Pediatr Nephrol 2018; 33(7):1243-9.
12. 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. Kid- ney Int 2014;85(3):659-67.
13. Goldstein SL, Chawla LS. Renal angina. Clinical journal of the American Society of Ne-

phrology: CJASN 2010;5(5):943-9.

1. Basu RK, Chawla LS, Wheeler DS, Goldstein SL. Renal angina: an emerging paradigm to identify children at risk for acute kidney injury. Pediatr Nephrol 2012;27(7): 1067-78.
2. Agewall S, Giannitsis E, Jernberg T, Katus H. Troponin elevation in coronary vs. non- coronary disease. Eur Heart J 2011;32(4):404-11.
3. Stein R, Gupta B, Agarwal S, et al. prognostic implications of normal (b0.10 ng/ml) and borderline (0.10 to 1.49 ng/ml) troponin elevation levels in critically ill patients without acute coronary syndrome. Am J Cardiol 2008;102(5):509-12.
4. Paul R, Melendez E, Wathen B, et al. A quality improvement collaborative for pediat- ric sepsis: lessons learned. Pediatr Qual Saf 2018;3(1):e051.
5. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)-a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42(2): 377-81.
6. Zappitelli M, Parikh CR, Akcan-Arikan A, Washburn KK, Moffett BS, Goldstein SL. As- certainment and epidemiology of acute kidney injury varies with definition inter- pretation. Clinical journal of the American Society of Nephrology: CJASN 2008;3 (4):948-54.
7. Schwartz GJ, Brion LP, Spitzer A. The use of plasma creatinine concentration for es- timating glomerular filtration rate in infants, children, and adolescents. Pediatr Clin North Am 1987;34(3):571-90.
8. Sutherland SM, Byrnes JJ, Kothari M, et al. AKI in hospitalized children: comparing the pRIFLE, AKIN, and KDIGO definitions. Clinical journal of the American Society of Nephrology: CJASN 2015;10(4):554-61.
9. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Bio- metrics 1988;44(3):837-45.
10. Basu RK, Kaddourah A, Goldstein SL, Investigators AS. Assessment of a renal angina index for prediction of severe acute kidney injury in critically ill children: a multicentre, multinational, prospective observational study. Lancet Child Adolesc Health 2018;2(2):112-20.
11. Sethi SK, Raghunathan V, Shah S, et al. fluid overload and renal angina index at ad- mission are associated with worse outcomes in critically ill children. Front Pediatr 2018;6:118.
12. Kaur R, Dhooria GS, Pooni PA, et al. Utilization of the renal angina index in PICU of a developing country for prediction of subsequent severe acute kidney injury. Pediatr Nephrol 2018;33(11):2185-91.
13. Menon S, Goldstein SL, Mottes T, et al. Urinary biomarker incorporation into the renal angina index early in intensive care unit admission optimizes acute kidney

injury prediction in critically ill children: a prospective cohort study. Nephrology, di- alysis, transplantation: official publication of the European Dialysis and Transplant Association-European Renal Association 2016;31(4):586-94.

1. Matsuura R, Srisawat N, Claure-Del Granado R, et al. Use of the renal angina index in determining acute kidney injury. Kidney Int Rep 2018;3(3):677-83.
2. Fierro MA, Ehieli EI, Cooter M, et al. Renal angina is a sensitive, but nonspecific iden- tifier of postcardiac surgery acute kidney injury. J Cardiothorac Vasc Anesth 2019;33 (2):357-64.
3. Park S, Baek SH, Ahn S, et al. Impact of electronic Acute kidney injury alerts with automated nephrologist consultation on detection and severity of AKI: a qual- ity improvement study. Am J Kidney Dis 2018;71(1):9-19.
4. Sutherland SM, Chawla LS, Kane-Gill SL, et al. Utilizing electronic health records to predict acute kidney injury risk and outcomes: workgroup statements from the 15 (th) ADQI consensus conference. Can J Kidney Health Dis 2016;3:11.
5. Kirkendall ES, Spires WL, Mottes TA, Schaffzin JK, Barclay C, Goldstein SL. Develop- ment and performance of electronic acute kidney injury triggers to identify pediatric patients at risk for nephrotoxic medication-associated harm. Appl Clin Inform 2014; 5(2):313-33.
6. Sutherland SM, Kwiatkowski DM. Acute kidney injury in children. Adv Chronic Kid- ney Dis 2017;24(6):380-7.
7. Basu RK, Wang Y, Wong HR, Chawla LS, Wheeler DS, Goldstein SL. Incorporation of biomarkers with the renal angina index for prediction of severe AKI in critically ill children. Clinical journal of the American Society of Nephrology: CJASN 2014;9(4): 654-62.