Development and external validation of a prediction model for the premature circuit clotting of continuous renal replacement therapy in critically ill patients

Abstract

Objective

This study aimed to develop and validate a prediction model for premature circuit clotting of continuous renal replacement therapy (CRRT) in critically ill patients.

Design

A retrospective cohort study was conducted on ICU patients undergoing CRRT. The Medical Information Mart for Intensive Care-III Clinical Database CareVue subset and Medical Information Mart for Intensive Care-IV were utilized for model development, while the eICU Collaborative Research Database was employed for external validation. Predictive factors were selected through Least Absolute Shrinkage and Selection Operator Regression and univariate logistic regression. A prediction model was then developed using binary logistic regression. Internal and external validations assessed the model's discrimination, calibration, and clinical net benefit.

Results

This study encompassed 2531 patients overall, with a premature circuit clotting rate of 31.88 %. The prediction model comprises five variables: body temperature, anticoagulation, mean arterial pressure, maximum transmembrane pressure change within two hours, and vasopressor. The model demonstrated robust predictive performance, achieving an area under the receiver operating characteristic curve of 0.897 (95 % CI: 0.879–0.915) in the training set and 0.877 (95 % CI: 0.852–0.902) in the external validation set. Internal validation yielded a Brier score of 0.087, while external validation showed a Brier score of 0.120. Calibration curves indicated good model calibration for both validations. The decision curve analysis indicates that the model yields a clinical net benefit across a wide range of decision thresholds.

Conclusion

The model demonstrates robust discrimination, calibration, and clinical net benefit, with readily available variables indicating substantial potential for valuable clinical applications.

Implications for clinical practice

Healthcare providers in the ICU can leverage the model to evaluate the risk of premature circuit clotting in critically ill patients undergoing continuous renal replacement therapy, facilitating timely intervention to mitigate its incidence.