Table 3.
Performance of baseline factors in predicting serum potassium testing in the 7 days after mineralocorticoid receptor antagonist (MRA) initiation.
Model Covariates | c-statistic | Pseudo R2 |
---|---|---|
Age, gender | 0.50 | 0.0007 |
Age, gender, pre-initiation serum potassium and creatinine* | 0.56 | 0.0192 |
Age, gender, pre-initiation serum potassium and creatinine, MI, CABG, PCI, V-tach, dyslipidemia, starting dose of MRA, prescription for potassium chloride, prescription for thiazide diuretic, heart rate, systolic blood pressure† | 0.67 | 0.1129 |
Age, gender, K levels, CR levels, MI, CABG, PCI, V-tach, starting dose of MRA, drug indicators for potassium chloride and thiazide diuretic, dyslipidemia, heart rate, systolic BP, site, year of MRA start | 0.70 | 0.1620 |
Pre-initiation serum potassium and creatinine were categorized: potassium missing, ≤4.50 mmol, and > 4.50 mmol; creatinine missing, ≤1.50 mg/dL, and >1.50 mg/dL.
Median used for missing heart rate and systolic blood pressure. Addition of LVEF as continuous covariate for subjects with a quantitative measure did not significantly influence c-statistic or R.