The adjustment methods in AUC calculation for age-related discriminatory measure

• Main Authors: Yi-Chen Chen, 陳怡蓁
• Other Authors: Chong-Yau Fu
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/39141900056832068737

碩士 === 國立陽明大學 === 公共衛生研究所 === 101 === In medical study, a continuous biomarker used for discriminating diseased from non-diseased is commonly evaluated from receiver operating characteristic curve (ROC). The area under ROC (AUC) is an important index representing the average accuracy. In fetal study, a biomarker may be collected by ultrasound or Doppler and thus provides to discriminate between diseased and no-diseased cases. It is noted that the collected biomarkers tend to vary with gestational age. Ignoring this factor, the yield ROC or AUC will be biased Two disease outcomes, Down syndrome and Intrauterine growth restriction (IUGR) are studied from nasal bone and umbilical artery pulsability. According to this application, this study investigates the methods for adjusting the effect of gestational age on ROC curve and AUC calculation. The regarding methods are categorized as (i) based on placement values and (ii) based on logistic regression analysis. A simulation study had been conducted for further verifying the adjusted methods in AUC calculation.