Development of a Machine Learning Model for Early Prediction of Sepsis in Intensive Care Unit Patients at Nigerian Teaching Hospitals

Sepsis is a leading cause of intensive care unit mortality in Nigerian teaching hospitals, and early prediction through automated analysis of routinely collected vital signs and laboratory data offers a clinically actionable tool for improving sepsis outcomes in resource-limited critical care settings. This study develops and validates a machine learning model for early sepsis prediction in ICU patients at Nigerian teaching hospitals, addressing the absence of validated prediction algorithms for the Nigerian critical care population. Retrospective clinical data from three hundred and forty ICU admissions over three years at the Lagos University Teaching Hospital were extracted from patient records by trained research assistants, covering vital signs (heart rate, respiratory rate, blood pressure, temperature, SpO2), laboratory results (white cell count, lactate, creatinine, bilirubin), and fluid balance data. Sepsis-3 criteria were applied as the outcome definition. A six-hour prediction horizon was targeted. Feature engineering produced time-series derived features including trend slopes, variability measures, and combinations representing modified early warning score components. Four machine learning algorithms, specifically logistic regression, random forest, gradient boosting, and long short-term memory neural network, were trained and evaluated by stratified five-fold cross-validation. The gradient boosting model achieved the best performance with an area under the ROC curve of 0.89, sensitivity of 83.4 percent, and specificity of 82.7 percent at the selected operating threshold. The model required only routinely available clinical data without additional laboratory testing. External validation at a second hospital in Enugu confirmed AUC of 0.86. Clinical implementation recommendations including integration with nursing workflow are discussed. Keywords: sepsis prediction, machine learning, intensive care unit, early warning, Lagos.