PreSev Study: Bringing Machine Learning to Preeclampsia Forecasting
Preeclampsia with severe features remains a dangerous threat during pregnancy—especially in low-resource settings like Nigeria, where early detection can save lives. While risk scoring tools exist, most are tailored to high-income countries and may not fit African populations. The PreSev Study, newly outlined in BMC Pregnancy and Childbirth, is set to change that with a locally optimized, machine-learning model designed for Lagos.
Why Focus on Severe Preeclampsia?
Preeclampsia—marked by high blood pressure and organ dysfunction—accounts for a significant share of maternal and fetal deaths in Nigeria. Late recognition can mean missed chances to initiate aspirin, increase monitoring, or deliver at the right time. Yet, most current risk models use data from non-African populations, potentially missing local patterns and predictors. The PreSev Study sets out to fill this gap with evidence tailored to the realities of Lagos maternity care.
How the Study Will Work
Recruitment: Over 900 pregnant women at moderate-to-high risk for preeclampsia (based on age, BMI, parity, medical history, and more) will be enrolled from four major hospitals in Lagos between 14–24 weeks gestation.
Data Collection: Detailed clinical and epidemiologic data will be gathered on each participant, covering key risk factors relevant to the Nigerian context.
Model Development: Five machine learning algorithms—random forest, XGBoost, multilayer perceptron, LightGBM, and glmnet—will be trained and compared using robust cross-validation.
Validation: The best model will be tested on a separate group to ensure its accuracy and reliability, with the goal of reaching an area under the curve (AUC) above 0.7.
Deployment: The winning algorithm will be translated into an online risk calculator for real-world clinical use.
What Makes This Study Different?
Context matters: By using local data and focusing on clinical (not laboratory) predictors, the model is designed to be practical, affordable, and scalable for frontline healthcare workers in Nigeria.
Transparency: The study uses advanced tools (like SHAP values) to explain which risk factors drive predictions, building trust among clinicians.
Impact: If successful, the tool could help target care, prioritize referrals, and personalize prevention for women most at risk—potentially reducing maternal and perinatal deaths.
Conclusion
The PreSev Study is paving the way for data-driven, locally relevant preeclampsia risk assessment in Nigeria. By leveraging machine learning and real-world data, this model could become a vital frontline tool to protect pregnant women and their babies.
Key Takeaways:
Severe preeclampsia is a leading cause of maternal and perinatal death in Nigeria.
Existing risk tools often fail to capture local realities; the PreSev Study aims to fix this.
Over 900 at-risk women will participate, with machine learning models built and validated on Lagos data.
The end goal is an easy-to-use, online preeclampsia risk calculator for Nigerian clinicians.
Context-specific prediction can guide earlier intervention and ultimately save lives.
Citation:
Babasola O, Kehinde O, Ayokunle O, Adeboje-Jimoh F, Adenekan M, Ojo O, Akinsanya G, Ojo T, Uthman O. Development of antenatal risk prediction model for preeclampsia with severe features in Lagos, Nigeria (PreSev study): protocol of a prospective cohort study. BMC Pregnancy and Childbirth. 2026; [Epub ahead of print]. doi:10.1186/s12884-026-09537-9
