Development and Validation of Models to Predict Major Adverse Cardiovascular Events in Chronic Kidney Disease

Abstract

Background

Accurate cardiovascular (CV) risk prediction tools may heighten awareness and monitoring, improve the use of evidence-based therapies and help inform shared decision making for patients with chronic kidney disease (CKD). The purpose of this study was to develop and externally validate a risk prediction model for incident and recurrent CV events across all stages of CKD using commonly available demographics and laboratory data.

Methods

A series of models were developed using administrative and laboratory data (n=36,317) from Manitoba, Canada, between April 1, 2006, and December 31, 2018, with external validation in health system’s data from Alberta, Canada (n=95,191), and Stockholm, Sweden (n=83,000). Adults with incident CKD stages G1-G4 were followed for the occurrence of major adverse cardiovascular events (MACE) (myocardial infraction, stroke, and CV death), and MACE including hospitalization for heart failure (MACE+). Discrimination and calibration were evaluated using the area under the receiver operating characteristic curve (AUC), Brier scores, and plots of observed vs predicted risk, and the models were compared to an existing model from the Chronic Renal Insufficiency Cohort (CRIC).

Results

In the Alberta cohort, the AUCs for predicting MACE and MACE+ were 0.77 (0.77-0.77) and 0.80 (0.79-0.80), respectively. In the Stockholm cohort, the model achieved an AUC of 0.87 (0.86-0.87) for predicting MACE and 0.88 (0.88-0.88) for MACE+. Overall performance was improved relative to CRIC.

Conclusions

A model including commonly available administrative data and laboratory results can predict the risk of MACE and MACE+ outcomes among individuals with CKD.