RT Journal Article
SR Electronic
T1 A hybrid mortality prediction model of patients with cavitary pulmonary carcinomas
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 341156
DO 10.1101/341156
A1 Maliazurina Saad
YR 2018
UL http://biorxiv.org/content/early/2018/06/11/341156.abstract
AB To date, developing a reliable mortality prediction model remains challenging. Although clinical predictors like age, gender and laboratory results are of considerable predictive value, the accuracy often ranges only between 60-80%. In this study, we proposed prediction models built on the basis of clinical covariates with adjustment for additional variables that was radiographically induced. The proposed method exhibited a high degree of prediction accuracy of between 83-92%, as well as overall improvement of between 6-20% in all other metrics, such as ROC Area, False Positive Rates, Recall and Root Mean Square Error. We provide a proof of concept that there is an added value for incorporating the additional variables while predicting 24-month mortality in pulmonary carcinomas patients with cavitary lesions. It is hoped that the findings will be clinically useful to the medical community.