Kidney Week

Abstract: SA-PO875

Predictors of Hyperkalemia in Patients with Advanced CKD

Session Information

• CKD: Pharmacoepidemiology
  November 09, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: CKD (Non-Dialysis)

• 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention

Authors

• Dashputre, Ankur A., University of Tennessee Health Science Center, Memphis, Tennessee, United States

Ankur A. Dashputre,

• Sumida, Keiichi, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Keiichi Sumida,

• Potukuchi, Praveen Kumar, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Praveen Kumar Potukuchi,

• Akbilgic, Oguz, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Oguz Akbilgic,

• Kar, Suryatapa, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Suryatapa Kar,

• Obi, Yoshitsugu, University of California Irvine, Irvine, California, United States

Yoshitsugu Obi,

• Thomas, Fridtjof, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Fridtjof Thomas,

• Molnar, Miklos Zsolt, University of Tennessee Health Science Center, Memphis, Tennessee, United States

Miklos Zsolt Molnar,

• Streja, Elani, Harold Simmons Center for Kidney Disease Research and Epidemiology, Orange, California, United States

Elani Streja,

• Kalantar-Zadeh, Kamyar, University of California Irvine, School of Medicine, Orange, California, United States

Kamyar Kalantar-Zadeh,

• Kovesdy, Csaba P., University of Tennessee Health Science Center, Memphis, Tennessee, United States

Csaba P. Kovesdy,

Background

Identifying predictors of hyperkalemia will help clinicians in managing patients with advanced CKD at a greater risk for future hyperkalemia events.

Methods

In 20,657 US Veterans with eGFR <30 ml/min/1.73m² and with ≥1 year of follow-up prior to dialysis initiation, we identified predictors of plasma potassium [K⁺] >6.0 mEq/l using multivariable logistic regression models with backward-selection based on Akaike’s information criterion, adjusted for demographics, comorbidities, vital signs, laboratory tests and medications. The sample was split (70%:30%) into training set (n=14,461) and test set (n=6,196). We conducted model cross validation using the leave one out cross validation method (LOOCV). We assessed model predictive discrimination using the area under the receiver-operator curve (AUC).

Results

The mean (SD) age of the patients was 67 (10) years; 98% were male, 29% were African American, with a mean (SD) baseline K⁺ of 4.6 (0.6) mEq/l, and baseline eGFR of 23.6 (4.7) ml/min/1.73m². At least one event of [K⁺] >6.0 mEq/l was experienced by 7.4% of the patients. Our final model included 16 predictor variables. The AUC (95% CI) estimates for training, test, and LOOCV were 0.765 (0.751-0.780), 0.761 (0.738-0.784), and 0.762 (0.749-0.774), respectively (Figure). Baseline K⁺ (OR [95%CI], 3.09 [2.83-3.39]), baseline Na polystyrene sulphonate use (2.51 [2.16-2.93]), and age (0.97 [0.96-0.98]) were the strongest predictors of hyperkalemia.

Conclusion

We developed and tested a prediction model with good discrimination ability to identify future hyperkalemia in patients with advanced CKD. Accurate prediction of future hyperkalemia could help implement preventive measures and may have a beneficial impact on outcomes.

Funding

• NIDDK Support