Artificial neural network, a computer-based technology that uses nonlinear statistics to recognize the relationship between input variables and an output variable, has been previously applied to outcome prediction in adult kidney recipients. In this study, we evaluated the effectiveness of a neural network model to predict a delayed decrease of serum creatinine in pediatric kidney recipients. The neural network was constructed with a training set of pediatric kidney recipients (n = 107) by using 20 input variables and assuming for the output variable, the time after 3 days to reach a serum creatinine level 50% below that before kidney transplantation. In the final model, the following input variables showing higher predictive values were retained: serum creatinine on day 1 post transplant, urine volume in the first 24 hours, diagnostic category, pretransplant dialysis mode, patient sex, donor sex, body weight on day 1 posttransplant, and patient age. The model was validated in a second set of patients (n = 41) by blinding the network for the output variable. The overall accuracies of the neural network for the training set, the validation set, and the whole patient cohort were 89.1%, 76.92%, and 87.14%, respectively. A comparative logistic regression analysis revealed only serum creatinine on day 1 posttransplant to be an independent predictor for the output variable (overall accuracy: 79.05%). The neural network showed sensitivity and specificity for the whole patient cohort to be 0.875 and 0.87, respectively, whereas using logistic regression sensitivity and specificity yields 0.37 and 0.94, respectively. This study proposes a neural network model that seemed to predict a delayed decrease in serum creatinine among pediatric kidney recipients. The availability of the source code may allow development of stand-alone neural networks to validate our model in prospective studies.

Application of an artificial neural network model to predict delayed decrease of serum creatinine in pediatric patients after kidney transplantation.

SANTORI, GREGORIO;VALENTE, UMBERTO
2007-01-01

Abstract

Artificial neural network, a computer-based technology that uses nonlinear statistics to recognize the relationship between input variables and an output variable, has been previously applied to outcome prediction in adult kidney recipients. In this study, we evaluated the effectiveness of a neural network model to predict a delayed decrease of serum creatinine in pediatric kidney recipients. The neural network was constructed with a training set of pediatric kidney recipients (n = 107) by using 20 input variables and assuming for the output variable, the time after 3 days to reach a serum creatinine level 50% below that before kidney transplantation. In the final model, the following input variables showing higher predictive values were retained: serum creatinine on day 1 post transplant, urine volume in the first 24 hours, diagnostic category, pretransplant dialysis mode, patient sex, donor sex, body weight on day 1 posttransplant, and patient age. The model was validated in a second set of patients (n = 41) by blinding the network for the output variable. The overall accuracies of the neural network for the training set, the validation set, and the whole patient cohort were 89.1%, 76.92%, and 87.14%, respectively. A comparative logistic regression analysis revealed only serum creatinine on day 1 posttransplant to be an independent predictor for the output variable (overall accuracy: 79.05%). The neural network showed sensitivity and specificity for the whole patient cohort to be 0.875 and 0.87, respectively, whereas using logistic regression sensitivity and specificity yields 0.37 and 0.94, respectively. This study proposes a neural network model that seemed to predict a delayed decrease in serum creatinine among pediatric kidney recipients. The availability of the source code may allow development of stand-alone neural networks to validate our model in prospective studies.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/262205
Citazioni
  • ???jsp.display-item.citation.pmc??? 3
  • Scopus 10
  • ???jsp.display-item.citation.isi??? 6
social impact