# International Delphi consensus on acute kidney injury: Foundations for AI-driven digital twin development in critical care nephrology

## TL;DR

Experts reached consensus on key factors for acute kidney injury in critical care, creating a foundation for AI-based digital twin models.

## Contribution

A structured causal framework for AKI in critical care was developed through international expert consensus.

## Key findings

- Hemodynamic instability and nephrotoxicity are leading contributors to AKI.
- Experts agreed on integrating biomarkers beyond serum creatinine and urine output.
- A Directed Acyclic Graph was created to represent the causal framework for AKI.

## Abstract

Acute kidney injury (AKI) in critically ill patients is clinically complex and heterogeneous, limiting the development of structured simulation models. Digital twin approaches require clearly defined causal relationships grounded in expert consensus. We aimed to establish an international Delphi consensus to inform the development of an interpretable, AI-driven digital twin framework for AKI in critical care.

We conducted a modified Delphi study involving up to three survey rounds. Experts in nephrology and critical care were invited to evaluate statements addressing AKI etiology, biomarker integration, biopsy indications, contrast use, and ICU management. Consensus was predefined as ≥75% agreement.

Fifty-six experts were invited, and 49 completed the first round. Consensus was achieved after two rounds. Hemodynamic instability and nephrotoxicity were identified as leading contributors to AKI. Experts supported integrating biomarkers beyond serum creatinine and urine output, although agreement varied for specific assays. Daily biomarker assessment reached consensus, whereas contrast use in advanced AKI stages did not. Desmopressin use prior to kidney biopsy in patients with markedly elevated blood urea nitrogen achieved consensus. A structured Directed Acyclic Graph was developed to represent the expert-derived causal framework.

This Delphi study established a clinically grounded framework for AKI in critical care while highlighting areas of practice variability. The resulting causal structure provides a transparent foundation for future AI-driven digital twin development and prospective validation.