AKI in hospitalized children: comparing the pRIFLE, AKIN, and KDIGO definitions

Abstract

Background and objectives: Although several standardized definitions for AKI have been developed, no consensus exists regarding which to use in children. This study applied the Pediatric RIFLE (pRIFLE), AKI Network (AKIN), and Kidney Disease Improving Global Outcomes (KDIGO) criteria to an anonymized cohort of hospitalizations extracted from the electronic medical record to compare AKI incidence and outcomes in intensive care unit (ICU) and non-ICU pediatric populations.

Design, setting, participants, & measurements: Observational, electronic medical record-enabled study of 14,795 hospitalizations at the Lucile Packard Children's Hospital between 2006 and 2010. AKI and AKI severity stage were defined by the pRIFLE, AKIN, and KDIGO definitions according to creatinine change criteria; urine output criteria were not used. The incidences of AKI and each AKI stage were calculated for each classification system. All-cause, in-hospital mortality and total hospital length of stay (LOS) were compared at each subsequent AKI stage by Fisher exact and Kolmogorov-Smirnov tests, respectively.

Results: AKI incidences across the cohort according to pRIFLE, AKIN, and KDIGO were 51.1%, 37.3%, and 40.3%. Mortality was higher among patients with AKI across all definitions (pRIFLE, 2.3%; AKIN, 2.7%; KDIGO, 2.5%; P<0.001 versus no AKI [0.8%-1.0%]). Within the ICU, pRIFLE, AKIN, and KDIGO demonstrated progressively higher mortality at each AKI severity stage; AKI was not associated with mortality outside the ICU by any definition. Both in and outside the ICU, AKI was associated with significantly higher LOS at each AKI severity stage across all three definitions (P<0.001). Definitions resulted in differences in diagnosis and staging of AKI; staging agreement ranged from 76.7% to 92.5%.

Conclusions: Application of the three definitions led to differences in AKI incidence and staging. AKI was associated with greater mortality and LOS in the ICU and greater LOS outside the ICU. All three definitions demonstrated excellent interstage discrimination. While each definition offers advantages, these results underscore the need to adopt a single, universal AKI definition.

Keywords: acute renal failure; children; kidney failure; mortality.

Figure 1.

Cohort creation, exclusion criteria, and subgroup definition. ICU, intensive care unit.

Figure 2.

Incidence of AKI according to three definitions. The incidence for each AKI stage is shown as a percentage. AKIN, AKI Network; KDIGO, Kidney Diseases Improving Global Outcomes; pRIFLE, pediatric RIFLE.

Figure 3.

Mortality and hospital length of stay (LOS) by AKI severity stage in ICU and non-ICU hospitalizations. (A) Mortality rates for patients with no AKI, stage 1 AKI, stage 2 AKI, and stage 3 AKI are presented as percentages with 95% confidence intervals. Mortality was compared with that of the previous stage. *P<0.05. Within the ICU, mortality was significantly higher for pRIFLE stages 1, 2, and 3; for AKIN stages 1 and 3; and for KDIGO stages 1 and 3 (P<0.05 for all). Outside of the ICU, higher AKI severity stage was not associated with higher mortality. (B) The total hospital LOS data for patients with no AKI, stage 1 AKI, stage 2 AKI, and stage 3 AKI are presented as medians (interquartile range). LOS was compared with that in the preceding stage. *P<0.05. Within the ICU, LOS was progressively higher at all three stages across all three definitions (P<0.001). Outside the ICU, LOS was progressively higher at all three stages across all three definitions except for pRIFLE stage 3 (P<0.001).

Figure 4.

Definitional overlap and outcomes for specific patient cohorts. The diagram demonstrates how many patients were diagnosed with AKI according to each of the three definitions. For example, 5406 patients were diagnosed by all three definitions, zero were diagnosed by AKIN and pRIFLE but not by KDIGO, 1720 were diagnosed by pRIFLE alone, and 427 were diagnosed by pRIFLE and KDIGO but not by AKIN. Data on mortality and LOS for patients not identified by each definition and diagnosed solely by each definition are also presented. n/a, not applicable.