Large language model driven multicenter prediction and explainable risk attribution of acute kidney injury

2026-07-27 · Source: Machine learning : nature.com subject feeds · Field: Health & Wellbeing — Clinical Care & Medical Practice, Medical Devices & Health Technology, Health & Medical Research · Depth: Expert, medium

Summary

A novel large language model (LLM) framework addresses critical limitations in acute kidney injury (AKI) prediction by combining high accuracy with clinical interpretability. Developed from a multi-center retrospective cohort study of 140,637 hospital admissions across four diverse Chinese hospitals, the framework comprises two models: AKI-PM for predicting AKI within 24 hours and AKI-RAM for explainable risk attribution. AKI-PM demonstrated strong predictive performance, achieving an area under curve (AUC) of 0.95 and a positive predictive value (PPV) of 0.68 in internal validation. It maintained robust generalizability externally with AUCs of 0.92-0.96 and PPVs of 0.69-0.74. AKI-RAM provides structured, actionable risk explanations, distinguishing modifiable from non-modifiable factors and offering tailored recommendations. Clinical evaluation by six nephrologists across 200 cases yielded high Likert scores (4.18-4.88) and good inter-rater reliability (0.680-0.803).

Key takeaway

For clinical teams and AI scientists developing diagnostic tools, this LLM-driven framework offers a robust solution to overcome the high false positive rates and lack of actionable insights common in existing AKI prediction models. You should consider integrating explainable AI components like AKI-RAM to provide clinicians with structured, modifiable risk factors, thereby enhancing trust and enabling more effective, personalized patient interventions for early AKI prevention.

Key insights

An LLM framework enhances AKI prediction and provides actionable, explainable risk attribution, improving clinical utility.

Principles

• Combine prediction with explainability for clinical utility.
• Distinguish modifiable from non-modifiable risk factors.

Method

The framework uses AKI-PM for 24-hour AKI prediction and AKI-RAM for explainable risk attribution, offering tailored recommendations based on modifiable factors.

In practice

• Implement early AKI prevention strategies.
• Tailor interventions based on identified risk factors.

Topics

• Acute Kidney Injury
• Large Language Models
• Clinical Prediction
• Explainable AI
• Risk Attribution
• Healthcare AI

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• TRIAGE: Dialectical Reasoning for Explainable Risk Prediction on Irregularly Sampled Medical Time Series with LLMs — TRIAGE uses dialectical LLM reasoning to provide calibrated, continuous risk scores and interpretable rationales for medical time series.
• Multimodal deep learning model for AI-based functional prognostic risk stratification in patients undergoing radical nephrectomy — A multimodal deep learning model predicts post-surgical kidney function decline in complex renal cell carcinoma patients.
• Predicting Post-Traumatic Epilepsy from Clinical Records using Large Language Model Embeddings — LLM embeddings from clinical records can predict post-traumatic epilepsy with high accuracy.
• A safety-oriented hypothetico-deductive framework for AI-assisted differential diagnosis — Engineering diagnostic AI as a safety-oriented reasoning framework provides safer, more transparent clinical decision support.
• Large Language Models as Unified Multimodal Learners for Clinical Prediction — Converting all patient data to text for LLM fine-tuning simplifies multimodal clinical prediction while maintaining performance.
• Clinically Interpretable Sepsis Early Warning via LLM-Guided Simulation of Temporal Physiological Dynamics — LLM-guided temporal simulation improves sepsis early warning by providing interpretable physiological trajectory predictions.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.