Background: Plasma creatinine (Cr) is a marker of kidney function and typically measured once daily. We hypothesized that Cr measured by point-of-care technology early after ICU admission would be a good predictor of acute kidney injury (AKI) the next day in critically ill patients.
Methods: We conducted a retrospective database audit in a single tertiary ICU database. We included patients with normal first admission Cr (CrF ) and identified a Cr value (CrP ) obtained within 6-12 hours from ICU admission. We used their difference converted into percentage (delta-Cr-%) to predict subsequent AKI (based on Cr and/or need for renal replacement therapy) the next day. We assessed predictive value by calculating area under the receiver characteristic curve (AUC), logistic regression models for AKI with and without delta-Cr-%, and the category-free net reclassifying index (cfNRI).
Results: We studied 780 patients. Overall, 70 (9.0%) fulfilled the Cr AKI definition by CrP measurement. On day 2, 148 patients (19.0%) were diagnosed with AKI. AUC (95% CI) for delta-Cr-% to predict AKI on day 2 was 0.82 (95% CI 0.78-0.86), and 0.74 (95% CI 0.69-0.80) when patients with AKI based on the CrP were excluded. Using a cut-off of 17% increment, the positive likelihood ratio (95% CI) for delta-Cr-% to predict AKI was 3.5 (2.9-4.2). The cfNRI was 90.0 (74.9-106.1).
Conclusions: Among patients admitted with normal Cr, early changes in Cr help predict AKI the following day.
Keywords: acute kidney injury; creatinine; critically ill; point-of-care; prediction.
© 2020 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd.