Comparing pRIFLE, AKIN, KDIGO for AKI Diagnosis and Outcomes in Hospitalized Children

Introduction

A standardized definition for AKI did not exist until 2004, when the Acute Dialysis Quality Initiative (ADQI) group introduced the RIFLE criteria.¹ Since then, the RIFLE framework has undergone several modifications to better reflect pediatric physiology. The pediatric RIFLE (pRIFLE) criteria were an early adaptation, followed by the Acute Kidney Injury Network (AKIN) definition, and finally the Kidney Disease: Improving Global Outcomes (KDIGO) criteria, which were published in 2012.^2-4 The standardization of AKI definitions has led to increased recognition and better understanding of the epidemiology of AKI in hospitalized children.⁵

Despite the significant progress in the standardization of AKI definition, the absence of a clearly superior criterion that is universally accepted leads to heterogeneity in the selection of diagnostic criteria across studies. The continued use of both pRIFLE and AKIN definitions in recent research underscores the lack of consensus on the best definition.^6-10 Emerging criteria, such as Reference Change Value Optimized for AKI (pROCK), aim to improve specificity in pediatric AKI diagnosis, particularly in children with low baseline creatinine levels. Although not evaluated in our study, their findings highlight the ongoing need for pediatric-appropriate definitions.¹¹

Uniform and accurate diagnosis and staging are crucial to understanding AKI burden and improving outcomes globally. Variation in diagnostic definitions complicates data comparison across regions. The recent pediatric ADQI (pADQI) consensus emphasized the need for studies from resource-limited settings to better inform global epidemiology and outcomes of pediatric AKI.⁵

The present study aimed to compare the incidence of AKI in hospitalized children using three different criteria (pRIFLE, AKIN, and KDIGO). In addition, it was proposed to assess the impact of defining and staging AKI using three different criteria on clinical outcomes, such as mortality and the need for KRT.

Materials and Methods

This prospective observational cohort study was conducted in the department of pediatrics at a tertiary care hospital in India from April 2023 to March 2024. Written informed consent was obtained from parents or legal guardians. All hospitalized children aged 1 month to 18 years were eligible for inclusion. Children with CKD or primary glomerular disease, AKI at admission with a hospital stay <48 hours, serum bilirubin >5 mg/dL, and no Scr within 48 hours of admission were excluded. Neonates were excluded from analysis to ensure uniform application of the three criteria to all patients. Since the variable under question defines AKI alone, children with predisposing factors for CKD, like CAKUT, were also included.

Baseline data were recorded within 24 hours of admission using a standardized case record form. Children were classified as critically ill if they required PICU admission, mechanical ventilation, or vasopressor support.¹² Scr was monitored every 48 hours during hospital stay; urine output was monitored every 12 hours in patients with indwelling urinary catheters.

AKI was diagnosed using creatinine-based criteria [Supplementary Table 1].^2,3,13 For clarity, we will refer to the Risk, Injury, and Failure stages of the pRIFLE criteria as stages 1, 2, and 3, respectively. Only the Scr criteria were applied for the primary analysis. Since our cohort included both ward and ICU patients, and urine output assessment was challenging, given that not all patients had indwelling catheters, we did not use urine output to define AKI.

Baseline SCr was defined as the lowest value within the preceding 3 months and was rounded to two decimal places. The creatinine values were estimated using the modified Jaffe method. GFR was calculated using the modified Schwartz formula.¹⁴ When unavailable, baseline GFR was assumed to be 120 mL/min/1.73 m².¹⁵ In fluid-overloaded patients, SCr was adjusted using the equation: Adjusted SCr = Measured SCr × (1 + Fluid balance/Total body water).¹⁶ For AKIN and KDIGO, baseline SCr was back-calculated from an assumed eGFR of 120 mL/min/1.73 m² using the Schwartz formula: Baseline SCr = (k × height in cm)/120.

Potential AKI exposures recorded included sepsis, dehydration, shock, nephrotoxic drug exposure, post-cardiac surgery state, and contrast administration. When more than one precipitating factor was present, the episode was classified as multifactorial AKI.

Fluid overload was calculated using the formula, Fluid overload (%) = [(Cumulative fluid balance – Urine output)/Admission weight] × 100.

Duration of AKI was measured in days. Recovery from AKI was defined as improvement in urine output, no need for KRT, and a Scr level within 0.3 mg/dL and <1.5 times the baseline value.¹⁷

Primary outcomes were the incidence of AKI by each definition. Secondary outcomes were mortality (in hospital), KRT requirement, length of hospital stay, need for mechanical ventilation, and recovery of kidney function at discharge.

Results

A total of 502 children were included after screening for exclusion criteria [Figure 1]. The median age was 6.0 years (2.0, 11.0), with 64.7% (n=325) males; 102 (20.3%) were critically ill. Critical illness was more frequent among those with AKI (p<0.001 across all definitions). AKI was significantly more common in critically ill children (OR 17.8, p < 0.001) compared to non-critically ill children.

Close

Figure 1:

Patient enrollment flowchart. AKI: Acute kidney injury, pRIFLE: Pediatric-modified risk, injury, failure, loss, and end-stage, AKIN: Acute kidney injury network; KDIGO: Kidney disease improving global outcomes.

Export to PPT

Baseline estimated GFR was slightly lower in AKI cases (median ∼118 mL/min/1.73m²) compared to non-AKI (120 mL/min/1.73m²). The etiology and underlying disease have been listed in Table 1.

Neurologic (27.7%) and hematologic (23.3%) involvement was most prevalent. Prior or in-hospital exposure to nephrotoxic medications was documented in 24.3% of all cases and was higher among those with AKI (p<0.001). The common diagnosis for each organ system involved has been given in Supplementary Table 2. AKI in our cohort was predominantly multifactorial (58.1% of cases) [Table 1].

Incidence and staging of AKI

AKI was diagnosed in 12.0%, 11.0%, and 10.8% of the patients based on pRIFLE, AKIN, and KDIGO definitions, respectively. The overall incidence of AKI was 12.3% (95% CI, 9.8–15.5%). Of all AKI cases, 60.0% (pRIFLE), 65.5% (AKIN), and 64.8% (KDIGO) were stage 1. pRIFLE criteria had picked more stage 3 AKI (20%) compared to AKIN (12.7%) and KDIGO (13.0%). However, neither the overall incidence nor the stage distribution of AKI differs statistically across the definitions. AKI occurred in 42.2% of ICU patients and 4.2% of ward patients by pRIFLE, 39.2% vs. 3.8% by AKIN, and 38.2% vs. 3.8% by KDIGO (all p < 0.001) [Table 2].

Table 2: AKI classification and inter-definition agreement

Category	pRIFLE	AKIN	KDIGO
AKI cases	60 (12.0)	55 (11.0)	54 (10.8)
No AKI	442 (88.1)	447 (89.0)	448 (89.2)
Stage 1	36 (60.0)	36 (65.5)	35 (64.8)
Stage 2	12 (20.0)	12 (21.8)	12 (22.2)
Stage 3	12 (20.0)	7 (12.7)	7 (13.0)
Overall agreement κ (%) (vs. other definitions)	0.912 (95.4) (vs. AKIN), 0.901 (96.2) (vs. KDIGO)	0.990 (97.1) (vs KDIGO), 0.912 (95.4) (vs pRIFLE)	0.990 (97.1) (vs AKIN), 0.901 (95.4) (vs pRIFLE)
Stage 1 κ (%) (vs. other definitions)	1.000 (99.8) (vs. AKIN), 1.000 (99.6) (vs. KDIGO)	0.985 (99.7) (vs. KDIGO), 1.000 (99.6) (vs. pRIFLE)	0.985 (99.7) (vs. AKIN), 1.000 (99.6) (vs. pRIFLE)
Stage 2 κ (%) (vs. other definitions)	0.659 (91.5) (vs. AKIN), 0.732 (92.8) (vs. KDIGO)	0.761 (93.1) (vs. KDIGO), 0.659 (91.5) (vs. pRIFLE)	0.761 (93.1) (vs. AKIN), 0.732 (92.8) (vs. pRIFLE)
Stage 3 κ (%) (vs. other definitions)	0.659 (vs. AKIN), 0.732 (vs. KDIGO)	0.742 (vs. KDIGO), 0.659 (vs. pRIFLE)	0.742 (vs. AKIN), 0.732 (vs. pRIFLE)
ICU AKI (%)	43 (42.2)	40 (39.2)	39 (38.2)
Non-ICU AKI (%)	17 (4.2)	15 (3.8)	15 (3.8)
p (ICU vs Ward)	< 0.001	< 0.001	< 0.001

κ: Kappa statistic for agreement. Stage-wise agreement is based on cross-tabulation between definitions.

Agreement across definitions was excellent, with overall kappa values >0.90 and percentage agreement ranging from 95.4% to 97.1%. Agreement was highest for Stage 1 (κ = 0.985–1.000) and remained substantial for Stages 2–3 (κ range: 0.659–0.761). AKIN and KDIGO showed the strongest alignment (κ = 0.990, 97.1%), while pRIFLE identified 7 (11%) additional cases not captured by the others.

Overlap analysis demonstrated that 83% children with AKI were identified by all three definitions. In contrast, pRIFLE uniquely detected 11% additional Stage 1 cases. None of these additional patients required KRT, and their median hospital stay (9 (7, 11) days) and mortality (0%) were comparable to those without AKI [Figure 2].

Close

Figure 2:

Venn diagram showing overlap of AKI diagnosis by pRIFLE, AKIN, and KDIGO definitions. Each circle represents the number of patients classified as AKI by that definition. The intersections denote cases identified by more than one definition. Fifty-two patients were common to all three, while pRIFLE uniquely detected seven.

Export to PPT

Discussion

In this prospective cohort of 502 hospitalized children, it was found that the incidence of AKI ranged from 10.8% to 12.0% depending on the AKI criteria used, with excellent agreement across definitions. AKI was strongly associated with adverse outcomes. This reinforces its clinical relevance irrespective of the definition applied. pRIFLE detected a higher number of AKI stage 1. However, it was not clinically significant. The increase in mortality rates with increasing AKI stage, the rates of renal recovery, and the need for KRT with increasing AKI stage were similar across the three groups.

A recent meta-analysis from India, using KDIGO, showed similar AKI incidence across HICs and low- and lower-middle-income countries, but higher mortality in LLMICs. However, we had a lower incidence of AKI, likely due to the predominance of ward cases in our cohort.¹⁸ Our findings are consistent with those of Sutherland et al., who, in a large cohort study, reported that all three AKI definitions - pRIFLE, AKIN, and KDIGO captured AKI cases with good inter-stage discrimination. However, incidence and staging varied depending on the criterion.¹⁹ The slight overestimation of AKI by pRIFLE, particularly in detecting Stage 3 cases, may be attributed to its use of eGFR decline, which could make it more sensitive in children with low muscle mass or subclinical renal dysfunction.^19,20

In a retrospective cohort study of 483 extremely low birth weight infants, Chowdhary et al. compared the pRIFLE, AKIN, and KDIGO definitions and found that the incidence of AKI was high across all three criteria (56–60%), with advanced AKI stages consistently associated with increased mortality, prolonged NICU stay, and impaired growth outcomes.²¹ We observed nearly perfect agreement in classifying Stage 1 AKI, with slightly lower but substantial agreement in higher stages with Stages 2–3 (κ range: 0.659–0.761). This aligns with previous findings that pRIFLE and KDIGO tend to agree in early AKI detection, but diverge in staging due to their differing thresholds, as pRIFLE relies on eGFR decline. In contrast, AKIN and KDIGO are based on absolute or relative increases in creatinine.²² While our study demonstrated comparable performance across the three definitions, it is also reported that the modified KDIGO (mKDIGO) criteria outperformed standard definitions in predicting mortality among neonates undergoing cardiac surgery, highlighting the potential value of tailored criteria in specific high-risk subpopulations.²³ In a large multicenter PICU cohort, Kuai et al. mKDIGO offered better mortality prediction than standard KDIGO. However, their study focused on critically ill children and included newer definitions like mKDIGO and pROCK.²⁴

In the present study, AKI was strongly associated with increased risk of in-hospital mortality, with odds of death being highest with pRIFLE, followed by KDIGO and AKIN.

The severity of AKI further influenced outcomes, as Stages 2 and 3 were linked to the highest risk of mortality and the greatest need for KRT. Notably, in our cohort, AKI was also associated with increased need for mechanical ventilation and with nephrotoxic drug exposure.

The overlap analysis revealed that while all three definitions identified most AKI cases, seven were picked up only by pRIFLE. This suggests that pRIFLE may detect certain cases missed by Scr–only definitions, particularly in children with subtle eGFR decline and low muscle mass. However, it had no clinical relevance as there was no requirement for KRT, mortality, or an increase in hospital stay in them.

Another encouraging finding is the high rate of renal recovery (69%) across all AKI definitions and stages at discharge, suggesting a potentially reversible disease course in children when managed appropriately, as noted in previous studies.²⁵ However, the lack of long-term follow-up data in our study precludes assessment of chronic kidney disease progression or long-term morbidity, a known consequence of pediatric AKI in other studies.

Our study has several strengths, including a prospective design, the standardized application of three AKI definitions, and a comprehensive assessment of clinically meaningful outcomes, such as mortality and the need for KRT. The inclusion of children with AKI admitted to wards also makes it more generalizable compared to the ICU alone. However, it is limited by incomplete urine output data, possible misclassification from assumed baseline eGFR in many patients, and the lack of follow-up data. Additionally, small sample sizes in the Stage 2 and 3 subgroups limit the precision of those analyses.

Despite these limitations, our study provides strong evidence that the three definitions perform comparably in identifying AKI and predicting poor outcomes in hospitalized children. Although PRIFLE criteria may detect a slightly higher AKI, there is no difference in the need for KRT; however, it also has higher odds of predicting mortality.

However, given the strong alignment between AKIN and KDIGO and their global adoption, KDIGO offers a practical and widely accepted approach for clinical application and research standardization. Emerging criteria, such as mKDIGO and pROCK, require prospective validation to identify the most accurate definition in diverse pediatric settings, particularly those with resource limitations.

Few prior studies have compared AKI criteria; however, most have been retrospective with small samples or limited to post-operative cases, thereby reducing their generalizability.^19,21,23,26 We conducted this study to provide more robust and prospective evidence, particularly from a resource-limited setting [Table 4].

In conclusion, we show that the pRIFLE, AKIN, and KDIGO criteria show excellent agreement in identifying AKI and in predicting key clinical outcomes such as mortality, mechanical ventilation, and the need for KRT in hospitalized children. pRIFLE detects additional cases due to its sensitivity to estimated GFR changes, but these changes do not translate into adverse outcomes.