# Persistent Severe Acute Kidney Injury Among Critically Ill Patients: Outcomes and Predictive Markers—A Single‐Center Retrospective Cohort Study

## TL;DR

This study identifies a severe form of kidney injury in ICU patients that is linked to high mortality and poor recovery, and finds that certain blood markers may help predict which patients are at higher risk.

## Contribution

The study introduces a new trajectory-based classification of acute kidney injury and identifies a novel predictive marker (NLR/PLT) for persistent severe AKI.

## Key findings

- PS-AKI was associated with 54% in-hospital mortality and only 17% renal recovery.
- The NLR/PLT ratio was independently associated with PS-AKI and had an AUC of 0.86.
- Inflammatory markers like NLR and CRP were higher in PS-AKI patients compared to others.

## Abstract

Persistent severe acute kidney injury (PS‐AKI)—recently standardized as Kidney Disease: Improving Global Outcomes (KDIGO) Stage 3 persisting ≥ 72 h, or renal replacement therapy/death after Stage 3 diagnosis—has emerged as a trajectory‐based phenotype complementing conventional KDIGO staging. Evidence in contemporary intensive care unit (ICU) cohorts remains limited.

We retrospectively studied adults admitted to a tertiary ICU (January 2024–June 2025). Acute kidney injury (AKI) was staged per KDIGO 2012, with trajectories classified as Stage 1 AKI, transient AKI (Stage 2‐3 resolving within 48 h), persistent mild–moderate AKI, or PS‐AKI. The primary outcome was in‐hospital mortality; secondary outcomes included renal recovery. Predictors of PS‐AKI were explored using logistic regression and gradient boosting with SHAP attribution.

Among 139 ICU patients with AKI screened, 106 met criteria. Most AKI was community‐acquired (97/106, 91.5%). PS‐AKI accounted for 23% (24/106) and carried the worst outcomes, with in‐hospital mortality 54% and renal recovery 17%. Within Stage 2‐3, PS‐AKI was associated with substantially worse outcomes than non‐PS trajectories (mortality 54% vs 10.8%; adjusted HR for death 2.23, 95% CI 0.69–7.21; adjusted OR for renal recovery 0.07, 95% CI 0.01–0.24). A 72‐h landmark analysis showed similar but nonsignificant trends. Inflammatory profiles distinguished PS‐AKI, with higher neutrophil‐to‐lymphocyte ratio (NLR), C‐reactive protein (CRP), and lower platelets. The composite NLR‐to‐platelet ratio (NLR/PLT) was independently associated with PS‐AKI (adjusted OR 2.51 per doubling, 95% CI 1.52–4.12; AUC 0.86), while the systemic immune‐inflammation index (SII) showed no significant association.

In this predominant community‐acquired ICU cohort, PS‐AKI was common and strongly associated with poor in‐hospital outcomes. The co‐occurrence of inflammation and thrombocytopenia, summarized by NLR/PLT, may represent a simple exploratory signal for early‐risk appraisal. These findings support further research into trajectory‐based AKI phenotypes and the potential utility of inflammation–hematologic markers in predicting persistence.