Accommodating Maternal Age in CRIB Scale: Quantifying the Effect on the Classification

Abstract

Receiver operating characteristic (ROC) curves are a well-accepted measure of accuracy of diagnostic tests using in continuous or ordinal markers. Based on the notion of using a threshold to classify subjects as positive (diseased) or negative (no diseased), a ROC curve is a plot of the true positive fraction (TPF) versus the false positive fraction (FPF) for all possible cut points. Thus, it describes the whole range of possible operating characteristic for the test and hence its inherent capacity for distinguish between two status. The clinical severity scale CRIB - Clinical Risk Index for Babies, emerged in 1993 to predict the mortality of newborn at less than 32 weeks of gestation and very low birth weight (< 1500gr) [4]. In previous work of Braga [3] this index was reported as showing a good performance in assessing risk of death for babies with very low birth weight (less than 1500 g weight). However, in some situations, the performance of the diagnostic test, the ROC curve itself and the Area Under the Curve(AUC) can be strongly influenced by the presence of covariates, whether continuous or categorical [5], [32], [33]. The World Health Organization and the Ministry of Health, defined as "late pregnancy" that thus occurs in women over 35 years. In this work, using the conditional ROC curve, we analyze the effect of one covariate, maternal age, on the ROC curve that representing the diagnostic test performance. We chose two age status, less than 35 years old and equal or greater than 35 years old, to verify the effects on the discriminating power of CRIB scale, in the process classification using R and STATA software.