Abstract
BACKGROUND: Hepatitis B Virus (HBV) remains a significant global public health challenge, particularly among pregnant women due to the risk of vertical transmission. Despite increasing reports of HBV in Nigeria, data specific to Idah, Kogi State, are limited. This study aimed to estimate the prevalence and identify predictors of HBV infection among pregnant women in this setting. METHODS: A crosssectional survey was conducted among 245 pregnant women attending their first antenatal clinic visit at Specialist Hospital, Idah, Kogi State, Nigeria (March-June 2024). HBV infection was defined as HBsAg positivity confirmed by rapid test and ELISA neutralization. Independent variables included age, marital status, educational level, exposure to sharp objects, alcohol use, tribal marks, and multiple sexual partners. Adjusted logistic regression models were fitted using penalized likelihood methods in R (v4.4.2). Model fit was assessed with likelihood ratio and Wald tests, Hosmer-Lemeshow goodnessoffit, calibration slope/intercept, and AUC. Predictive performance was evaluated using confusion matrix metrics with Wilson score confidence intervals. Missing data were addressed using single imputation under a Missing at Random (MAR) assumption. RESULTS: HBV prevalence was 5.7% (95% CI: 3.2-9.4%). Significant predictors included marital status (OR = 23.66, 95% CI: 1.04-654.93), exposure to sharp objects (OR = 23.88, 95% CI: 1.52-3519.22), tribal marks (OR = 9.29, 95% CI: 1.84-99.56), and multiple sexual partners (OR = 66.79, 95% CI: 4.80-1665.01). The model demonstrated excellent discrimination (AUC = 0.9635) and good calibration (Hosmer-Lemeshow χ(2) = 4.29, df = 8, p = 0.83). At the Youden threshold (0.30), sensitivity was 0.9610 (95% CI: 0.93-0.98), specificity 0.6429 (95% CI: 0.39-0.85), PPV 0.9780 (95% CI: 0.95-0.99), NPV 0.5000 (95% CI: 0.27-0.73), F1 score 0.969, and balanced accuracy 0.8019. CONCLUSION: HBV infection among pregnant women in Idah was significantly associated with marital status, cultural practices (tribal marks), exposure to sharp objects, and multiple sexual partners. The predictive model demonstrated strong discrimination and calibration, supporting its potential utility for HBV risk stratification in antenatal care. However, modest specificity and NPV highlight the need for cautious interpretation and external validation in larger, multisite cohorts.