Construction and External Validation of a Predictive Nomogram for Premature Unplanned Discontinuation in Elderly Patients Undergoing Continuous Renal Replacement Therapy in the
ICU
Zhengnan Wu, Ronghua Zhao, Mingming Wang, Shuwen Ren, Titi Li, Peng Wang ABSTRACT
Introduction
To develop and validate a 2‐h early‐warning nomogram for the risk of unplanned discontinuation of continuous renal replacement therapy (CRRT) in elderly intensive care unit (ICU) patients.
Methods
This study retrospectively enrolled elderly CRRT patients from two tertiary hospitals (2022–2024) via convenience sampling for model development. Predictive performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration plots, and internal validation using 1000 bootstrap resamples, with goodness of fit assessed by the Hosmer–Lemeshow test; an external cohort (Jan–Aug 2025) was used for clinical applicability.
Results
The unplanned CRRT discontinuation rate reached 27.4% in the modeling cohort and 32.8% in an external temporal validation cohort ( n = 247 sessions). Independent predictors included hypotension, therapeutic intervention scoring system (TISS) score, transmembrane pressure (TMP) at 2 h > 86 mmHg, and diagnosis. The model demonstrated robust predictive performance, with an AUC of 0.860 (training) and 0.865 (validation), good calibration (Brier scores: 0.107 internal, 0.127 external), and clinical net benefit across thresholds per decision curve analysis.
Conclusions
The model demonstrates robust discrimination, calibration, and clinical net benefit, with readily available variables indicating substantial potential for valuable clinical applications.