Using artificial intelligence to predict the equilibrated postdialysis blood urea concentration

FERNÁNDEZ, ELMER ANDRÉS; VALTUILLE, RODOLFO; PERAZZO, CARLOS ALBERTO; WILLSHAW, PETER

BLOOD PURIFICATION.

KARGER

Lugar: Basel; Año: 2001

0253-5068

Total dialysis dose (Kt/V) is considered to be a majordeterminant of morbidity and mortality in hemodialyzedpatients. The continuous growth of the blood urea concentrationover the 30- to 60-min period following dialysis,a phenomenon known as urea rebound, is a criticalfactor in determining the true dose of hemodialysis. Themisestimation of the equilibrated (true) postdialysisblood urea or equilibrated Kt/V results in an inadequatehemodialysis prescription, with predictably poor clinicaloutcomes for the patients. The estimation of the equilibratedpostdialysis blood urea (eqU) is therefore crucialin order to estimate the equilibrated (true) Kt/V. In thiswork we propose a supervised neural network to predictthe eqU at 60 min after the end of hemodialysis. The useof this model is new in this field and is shown to be betterthan the currently accepted methods (Smye for eqU andDaugirdas for eqKt/V). With this approach we achieve amean difference error of 0.22 B 7.71 mg/ml (mean %error: 1.88 B 13.46) on the eqU prediction and a meandifference error for eqKt/V of ?0.01 B 0.15 (mean % error:?0.95 B 14.73). The equilibrated Kt/V estimated with the eqU calculated using the Smye formula is not appropriatebecause it showed a great dispersion. The Daugirdasdouble-pool Kt/V estimation formula appeared to beaccurate and in agreement with the results of the HEMOstudy