Predictive Performance of Equations to Estimate Creatinine Clearance in Hospitalized Elderly Patients

To ascertain the clinical accuracy of equations that estimate creatinine clearance to predict the correct drug doses in hospitalized elderly patients

Single 24-hour creatinine clearance measurement compared with estimated creatinine clearances derived from eight equations using total and modified ideal body weight

Nonintensive care medical and surgical units at a county hospital

15 patients with urethral catheters were enrolled in each of three age groups: 65–75, 76–85, and ≥86 years

Drug–dose predictions, bias, precision, and absolute errors

The bias for all equations was −4.0−42.0 mL/min (–0.07–0.70 mL/s) and the precision was 10.8−47.4 mL/min (0.18–0.88 mL/s). The Jelliffe 1973, Hull et al., and Mawer et al. equations were the least biased and the Jelliffe 1973 was the most precise, followed by the Mawer et al., Hull et al., and Cockcroft-Gault equations. The percent of patients with absolute percent errors >20 percent were 38 percent for Jelliffe 1973, 36 percent for Mawer et al., 40 percent for Hull et al., and >50 percent for the other equations. The percent of patients receiving correct drug doses was 67 percent for Jelliffe 1973, 58 percent for Gates, 51 percent for Mawer et al. and Hull et al., and <50 percent for the other equations. Within various age, renal function, serum creatinine, and albumin subgroups, the Jelliffe 1973 estimates were least biased and most precise, followed by the Cockcroft-Gault estimates. Generally, estimates using modified lean body weight performed better than did those using total body weight.

The Jelliffe 1973 equation with modified lean body weight was the best equation, followed by the Cockcroft-Gault equation. Even with the best equation, 33 percent of the patients would have received an incorrect drug dose. Therefore, some elderly patients may still require a measured creatinine clearance.