Premium
Artificial neural network is highly predictive of outcome in paediatric acute liver failure
Author(s) -
Rajanayagam J.,
Frank E.,
Shepherd R. W.,
Lewindon P. J.
Publication year - 2013
Publication title -
pediatric transplantation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.457
H-Index - 69
eISSN - 1399-3046
pISSN - 1397-3142
DOI - 10.1111/petr.12100
Subject(s) - medicine , liver failure , artificial neural network , outcome (game theory) , liver transplantation , intensive care medicine , artificial intelligence , transplantation , mathematics , mathematical economics , computer science
Current prognostic models in PALF are unreliable, failing to account for complex, non‐linear relationships existing between multiple prognostic factors. A computational approach using ANN should provide superior modelling to PELD ‐ MELD scores. We assessed the prognostic accuracy of PELD‐MELD scores and ANN in PALF in children presenting to the QLTS, Australia. A comprehensive registry‐based data set was evaluated in 54 children (32M, 22F, median age 17 month) with PALF . PELD ‐ MELD scores calculated at (i) meeting PALF criteria and (ii) peak. ANN was evaluated using stratified 10‐fold cross‐validation. Outcomes were classified as good (transplant‐free survival) or poor (death or LT) and predictive accuracy compared using AUROC curves. Mean PELD ‐ MELD scores were significantly higher in non‐transplanted non‐survivors (i) 37 and (ii) 46 and transplant recipients (i) 32 and (ii) 43 compared to transplant‐free survivors (i) 26 and (ii) 30. Threshold PELD ‐ MELD scores ≥27 and ≥42, at meeting PALF criteria and peak, gave AUROC 0.71 and 0.86, respectively, for poor outcome. ANN showed superior prediction for poor outcome with AUROC 0.96, sensitivity 82.6%, specificity 96%, PPV 96.2% and NPV 85.7% (cut‐off 0.5). ANN is superior to PELD ‐ MELD for predicting poor outcome in PALF .