Developing a Predictive Model for Adverse Birth Outcomes Based on Nursing Assessment Indicators in Primary and Secondary Health Facilities in Anambra State

Background: Early nursing identification of adverse birth outcome risk during antenatal care has the potential to avert preventable perinatal deaths, yet no validated nurse-based predictive model exists for Nigerian primary and secondary care contexts. Aim: This study developed a predictive model for adverse birth outcomes using nursing assessment indicators at primary and secondary health facilities in Anambra State, South East Nigeria. Methods: A prospective cohort design followed 500 pregnant women from first antenatal contact to delivery. Nursing assessment data including vital signs trends, fundal height measurements, proteinuria, and foetal movement patterns were systematically collected. Machine learning algorithms including logistic regression, random forest, and gradient boosting were compared to identify the optimal predictive model. Results: The random forest model achieved the best discrimination with area under the ROC curve of 0.84. Key predictors included proteinuria trajectory, fundal height discordance, blood pressure trends, and gravidity. The model flagged high-risk women at a mean of 6.4 weeks before adverse events. Conclusion: A nurse-administered predictive model based on routine clinical indicators can substantially improve early identification of at-risk pregnancies in Nigerian primary and secondary care settings. The model has direct implementation implications for nurse antenatal risk assessment protocols. Keywords: adverse birth outcomes, predictive model, nursing assessment, antenatal care, Anambra State.