A novel age-norm-capped baseline serum creatinine approach for acute kidney injury diagnosis in critically Ill children: a retrospective cohort study.

Background: A reliable baseline serum creatinine (SCr) for pediatric acute kidney injury (AKI) diagnosis is often missing, and simply using admission SCr may underestimate AKI. We aimed to evaluate a novel approach for defining baseline SCr by comparing its implications for AKI classification and prognosis with those of admission SCr. Methods: This retrospective study utilized data from the Paediatric Intensive Care (PIC) database. We compared two distinct baseline SCr definitions: 1) the "Admission SCr" and 2) a novel "Norms-Capped SCr", defined as the admission SCr if it was less than or equal to the age-specific normal maximum (NormsMax), or the NormsMax value if the admission SCr exceeded it. Both baseline definitions were then used to classify AKI according to three different diagnostic criteria. We subsequently compared the resulting AKI incidence, stage distribution, and prognostic performance for clinical outcomes across these different baseline-criterion combinations. Results: A total of 3,021 critically ill children were included. Norms-Capped SCr identified more AKI than admission SCr across KDIGO (41.7% vs. 27.4%), pROCK (38.7% vs. 27.5%), and pRIFLE (55.1% vs. 32.6%). It also showed more consistent associations with in-hospital mortality, although pRIFLE Stage 3 was not significant. Apparent AUCs ranged from 0.797 to 0.817 for Norms-Capped SCr models and from 0.783 to 0.785 for admission SCr models; the corresponding optimism-corrected AUCs were 0.784-0.804 and 0.767-0.770, respectively. Calibration was comparable, while prediction error was slightly lower with Norms-Capped SCr. Conclusion: The Norms-Capped SCr approach showed modestly better overall performance than admission SCr for defining baseline SCr in critically ill children without known pre-admission values, particularly when combined with the pROCK criteria.