Abstract
Neonatal jaundice is a frequently encountered clinical condition in newborns that requires prompt recognition and management to prevent serious outcomes such as kernicterus. This study focuses on developing a statistical classification model to assess the risk of neonatal jaundice by utilizing cord blood albumin levels in conjunction with key perinatal factors such as birth weight and gestational age. Conducted as a hospital-based prospective observational study, it included 168 neonates and their mothers. Data collected comprised maternal age, birth weight, gestational age, APGAR scores and cord blood albumin levels. Discriminant function analysis (DFA) and Receiver operating characteristic (ROC) curve analysis were employed to identify optimal cut-off points and evaluate the model’s effectiveness. The ROC analysis determined that a cord blood albumin level of ≤2.8 g/dL served as a critical threshold, achieving a sensitivity of 93.5% and specificity of 89.2%. The discriminant model demonstrated a high classification accuracy of 91.7%, which was further confirmed through 5-fold cross-validation. Among the variables, cord blood albumin was identified as the most significant predictor. These results highlight the potential of integrating statistical prediction models into neonatal care for early detection of jaundice risk. The study underscores that cord blood albumin, a simple, non-invasive and cost-effective biomarker, can play a crucial role in enhancing early diagnosis and treatment strategies for neonatal jaundice when combined with advanced statistical methodologies.
Keywords: Classification, Cord Blood Albumin, Discriminant Analysis, Neonatal Jaundice, Sensitivity, Specificity.