Risk of obstructive acute kidney injury: derivation and internal validation of a risk stratification tree

Acute kidney injury (AKI) is a frequent condition among patients presenting to the emergency department (ED). Obstructive AKI constitutes a urological emergency requiring rapid diagnosis and intervention. The objective of this study was to derive and internally validate a predictive model of obstructive AKI in ED.

We conducted a retrospective derivation and internal validation cohort study. Adult patients presenting to the EDs of Toulouse University Hospital with AKI of any Kidney Disease: Improving Global Outcomes (KDIGO) stage between 1 July and 31 December 2019 were eligible. Included patients were randomly assigned in a 2:1 ratio to a derivation cohort (DC) and an internal validation cohort (VC). The primary outcome was obstructive AKI, defined as hydronephrosis identified on imaging and requiring either urological intervention or Foley catheter insertion. A risk stratification decision tree was developed (Kidney Injury Tree For Identification of obSTructive Origin (KIT-FISTO) model).

The prevalence of obstructive AKI was 9% in the DC (64/727) and 7% in the VC (27/364). Patients presenting with lumbar, flank or hypogastric pain were classified as 'high risk' in each cohort, with a corresponding obstructive AKI risk of 55% (95% CI 45% to 64%) and 54% (95% CI 39% to 69%), respectively. Patients without pain but with a history of urinary tract surgery, abdominal cancer, a solitary functional kidney or prostatic hyperplasia were classified as 'moderate risk', with obstructive AKI risks of 4% (95% CI 1% to 10%) and 2% (95% CI 0% to 12%), respectively. All remaining patients (>70%) were classified as 'low risk', with an observed obstructive AKI risk of 0% (95% CI 0% to 1%). 'Low risk' classification had a sensitivity of 98% (95% CI 92% to 100%) and 96% (95% CI 81% to 100%), respectively.

The KIT-FISTO was derived and internally validated to predict obstructive AKI in ED, but requires external and prospective validation before implementation.