Dialysis Risk Prediction and Treatment Effect Estimation for AKI patients using Longitudinal Electronic Health Records

2026-04-27

AI Summary

This study addresses the prediction of progression to dialysis or end-stage kidney disease among patients with acute kidney injury (AKI) and evaluates the causal effect of pharmacological interventions on this rare outcome. Leveraging a longitudinal electronic health record cohort with a 90-day observation window and a 730-day prediction horizon, the work integrates time-series diagnostic, procedural, medication, and renal function data. It proposes a novel causal multi-head model that combines Transformer architecture with counterfactual reasoning to estimate average treatment effects at both drug and ingredient levels. Counterfactual exposure simulations are validated through robustness checks employing inverse probability of treatment weighting (IPTW), augmented IPTW (AIPW), and covariate-adjusted regression. The model achieves an AUC of 0.694 and a PR-AUC of 0.094 on the test set. Causal analyses suggest a potential protective effect of ACE inhibitors/ARBs, whereas loop diuretics may exacerbate renal deterioration.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Progression to dialysis or end-stage renal disease is a rare but clinically important outcome. Clinicians need evidence on how medication exposures influence downstream risk. We constructed a fixed-window EHR cohort (90-day observation, 730-day prediction; N=81401; dialysis/ESRD prevalence: 1.1%) and modeled sequences of diagnoses, procedures, and medications with kidney laboratory trends (creatinine, BUN, eGFR). A transformer-based causal multi-head model was trained to estimate drug- and ingredient-level average treatment effects (ATEs) using counterfactual exposure removal and insertion under a full medication history setup. On test set, predictive performance reached an AUC of 0.694 and PR-AUC of 0.094. At the selected decision threshold (0.883), the model achieved an F1 score of 0.201 with a Brier score of 0.018. Post-hoc causal analyses of lab changes (eGFR, creatinine, BUN) using IPTW, AIPW, naive, and covariate-adjusted OLS methods assessed clinical directionality. Results showed partial protective-direction support for ACE/ARB exposures and worsening-direction signals for loop diuretics.

Problem

• dialysis risk prediction
• treatment effect estimation
• acute kidney injury
• electronic health records
• causal inference

Innovation

• Transformer-based causal model
• Average Treatment Effect (ATE)
• Counterfactual exposure
• Longitudinal EHR
• Dialysis risk prediction