Lassa fever (LF), an acute viral hemorrhagic illness prevalent in West Africa, significantly impacts public health due to its varied clinical manifestations and high mortality rate. In severe cases of LF, the disease can progress to more critical conditions such as hemorrhaging, respiratory distress, and organ failure which is notorious for its high mortality rate, especially in cases of delayed or misdiagnosed treatment. Prevention and control efforts involve a multi-faceted approach. Hence, this study develops a predictive framework for assessing the severity of Lassa fever using a Risk Matrix approach and Machine Learning (ML) techniques for categorizing symptoms into various risk levels to prioritize healthcare responses. Traditional severity assessments were subjective, but ML provided an objective alternative. The study developed an ML based severity classification using clinical parameters from 239 confirmed LF patient records. Features included; age, blood pressure, sore throat, fever, cough, abdominal pain, vomiting, headaches, diarrhea, nose bleeding, myalgia and depression. ML models including Artificial Neural Network (ANN), Decision Tree (DT) and Random Forest (RF) were tested, with the derived Risk matrix across class levels of low, moderate and high risk, optimizing performance through cross-validation. RF achieved the highest accuracy at 93.7%, DT reached 92% and ANN followed with 83%. Therefore, RF was selected for the development and deployment of a user interface in R for predicting Lassa fever severity. The proposed framework achieved high accuracy and demonstrated potential for clinical integration to assist decision-making in resource-limited settings.

ANN Decision Tree Lassa fever Machine learning Random Forest Risk Matrix Severity prediction