Screening test data analysis for liver disease prediction model using growth curve

Young Sun Kim, So Young Sohn, Chang No Yoon

Research output: Contribution to journalArticlepeer-review

25 Citations (Scopus)


This study was done based on screening test data accumulated from 1994 to 2001 for studying of risk factor related with liver disease and prediction model of liver disease. In the existing study related with liver, the main current is studying on liver cancer, not on liver disease, previous step into liver cancer. As a result of estimating prediction model through the risk factors of liver disease and the growth curve on the basis of data, it is shown that most of the risk factors about liver disease are also those about known well as liver cancer. In addition, to investigate liver disease prevalence from the viewpoint of the future, this study presumed risk factor through the various growth curve analysis and examined logistic regression, decision tree and neural network from those estimators. In the case of neural network using growth curve estimator of Xi(5)iiT+εiT, accuracy of liver disease was 72.55% and sensitivity was 78.62%. On the other hand, in the case of liver disease prediction model using recent screening test data estimator, accuracy was 72. 09% and sensitivity was 71.72%. Those are lower than liver disease prediction model of growth curve analysis. In the various liver disease prediction models assumed by growth curve and many distinction models, when growth curve estimator was used, sensitivity value was improved.

Original languageEnglish
Pages (from-to)482-488
Number of pages7
JournalBiomedicine and Pharmacotherapy
Issue number10
Publication statusPublished - 2003 Dec

Bibliographical note

Funding Information:
This work was supported partly by National Research Laboratory program of the Ministry of Science and Technology, Korea.

All Science Journal Classification (ASJC) codes

  • Pharmacology


Dive into the research topics of 'Screening test data analysis for liver disease prediction model using growth curve'. Together they form a unique fingerprint.

Cite this