# The External Validation of SACrA (Sex, Albumin, Creatinine, and APACHE II) Score for Predicting Nonemergent Renal Replacement Therapy Initiation: A Retrospective Study Based on the Medical Information Mart for Intensive Care - Version IV Database

## TL;DR

This study validates a new score (SACrA) for predicting when patients in the ICU will need nonemergent kidney treatment, using a large database of ICU patients.

## Contribution

The SACrA score is externally validated for predicting nonemergent renal replacement therapy initiation in ICU patients using the MIMIC-IV database.

## Key findings

- The SACrA score showed good discrimination with an AUC of 0.81 for predicting nonemergent RRT initiation.
- Calibration plots indicated the SACrA score was well-calibrated with a slope of 1.002 and intercept of 0.007.
- Decision curve analysis suggested the clinical utility of the SACrA score in guiding nonemergent RRT initiation.

## Abstract

Background and objective

The SACrA [sex, albumin, creatinine, and Acute Physiology and Chronic Health Evaluation II (APACHE II)] score was developed to objectively predict nonemergent renal replacement therapy (RRT) initiation. This study aimed to validate the SACrA score externally by using the Medical Information Mart for Intensive Care - Version IV (MIMIC-IV) database.

Methods

We conducted a retrospective cohort study using the MIMIC-IV database. Patients admitted to the ICU with a total hospital stay of ≥7 days were included, whereas those with chronic kidney disease grade 5, end-stage renal disease, post-kidney transplant status, or urgent RRT indications were excluded from this study. The primary outcomes were the discrimination [area under the curve (AUC)] and calibration of the SACrA score to predict nonemergent RRT initiation, defined as blood urea nitrogen (BUN) ≥112 mg/dL or oliguria lasting >72 hours.

Results

Among the 16,360 ICU patients who met the inclusion criteria, 14,226 were analyzed after applying the exclusion criteria. Of them, 658 (4.6%) met the criteria for nonemergent RRT initiation, and 351 patients (2.5%) actually received RRT. The SACrA score showed good discrimination, with an AUC of 0.81 in the primary dataset and similar performance across multiple imputed datasets (AUCs ranging from 0.81 to 0.82). Calibration plots indicated that the SACrA score was well-calibrated, with a slope of 1.002 and an intercept of 0.007. Decision curve analysis indicated the potential clinical utility of this score in guiding nonemergent RRT initiation.

Conclusions

The SACrA score is a reliable tool for predicting nonemergent RRT initiation in critically ill patients. Although it accurately identifies high-risk patients, further randomized controlled trials (RCTs) are required to determine whether targeted interventions based on SACrA scores can improve patient outcomes.