Development and validation of a risk prediction model for citrate accumulation in patients undergoing continuous renal replacement therapy with regional citrate anticoagulation.

Abstract

Introduction: Regional citrate anticoagulation (RCA) is now recommended as the first choice of anticoagulation for continuous renal replacement therapy (CRRT). However, impaired citrate metabolism can lead to citrate accumulation (CA), resulting in severe metabolic acidosis and hypocalcemia, which poses a challenge for clinicians when making decision about the use of RCA.

Methods: In this retrospective cohort study performed in West China Hospital of Sichuan University, we evaluated patients who underwent RCA-based CRRT from 2021 to 2023. Participants were randomly allocated into training and validation groups at a 7:3 ratio. In the training group, significant risk factors for CA were determined by a binary logistic regression analysis and established a risk prediction model, the validation group validated and evaluated the model. A nomogram was constructed to visualize the prediction model, and calibration and receiver operating characteristic (ROC) curves were used to evaluate the prediction accuracy, and decision curve analysis (DCA) was used to evaluate the clinical effectiveness.

Results: Of the 1,259 patients with RCA-CRRT, 882 were randomly stratified into the training group and 377 into the validation group. CA was reported in 16.2% and 16.7%, respectively. We developed and validated a nomogram to predict the risk of CA, incorporating significant factors including male, age, body surface area (BSA), mean hourly citrate dosage, systolic blood pressure (SBP), lactate, total bilirubin (TBIL) and international normalized ratio (INR). The area under the ROC curve of the nomogram was 0.760 (95% CI, 0.737-0.765) and 0.752 (95% CI, 0.744-0.787) in both groups. The calibration curve further confirmed its effective discrimination and calibration abilities. DCA analysis emphasized its clinical utility when the CA probability threshold for intervention is between 11% and 76%.

Conclusion: We developed and validated a useful prediction model for CA in critically ill patients who underwent RCA-CRRT, assisting clinicians in identifying high-risk individuals.