Show simple item record

dc.contributor.authorMutegyeki, Walter
dc.date.accessioned2024-12-10T08:29:03Z
dc.date.available2024-12-10T08:29:03Z
dc.date.issued2024
dc.identifier.urihttp://hdl.handle.net/10570/13936
dc.descriptionA dissertation submitted to the School of Public Health in partial fulfillment of the requirements for the award of a Master of Health Informatics Degree of Makerere Universityen_US
dc.description.abstractBackground: Postpartum haemorrhage (PPH) remains one of the leading preventable causes of maternal mortality in Uganda, particularly affecting young mothers. Early identification and intervention are crucial to reduce its impact. However, current screening methods often rely on expensive laboratory tests and require skilled healthcare workers, who may not always be readily available. Objective: This study developed a predictive model for PPH risk classification among pregnant women in the Kampala district using supervised machine learning. Methods: PPH risk factors and the presence dataset of pregnant mothers above the age of 18 who attended antenatal services and delivered at Naguru National Referral Hospital in Uganda were extracted from electronic patient medical records. Four supervised machine learning algorithms, including decision tree, random forest, support vector machine, and k-nearest neighbours, were applied to analyse the data. Each algorithm's performance was evaluated through hyperparameter optimization and 10-fold cross-validation, followed by a comparative analysis based on accuracy, sensitivity, specificity, and area under the ROC curve. The best-performing algorithm was then utilized to build a web application. Results: The random forest algorithm outperformed support vector machines, k-nearest neighbours, and decision trees, achieving an accuracy of 88.7%, a precision of 83.8%, a recall of 94.8%, an F1 score of 88.9%, a sensitivity of 94.8%, a specificity of 88.7%, and an Area Under the Receiver Operating Characteristic Curve (ROC-AUC) of 94.3%. A web application was developed to allow health workers to select relevant risk factors and predict the likelihood of PPH using the prediction model. This method accurately predicted the presence or absence of PPH in pregnant women without the need for laboratory tests, kits, or devices, and provided real-time results. Conclusion: The study demonstrated the potential of a machine learning application can be deployed as a web-based tool for predicting PPH among pregnant mothers in the Uganda population. It is recommended that health workers adopt this tool, as it has the potential to significantly support efforts to improve health outcomes and save lives in low-income settings.en_US
dc.language.isoenen_US
dc.publisherMakerere Universityen_US
dc.subjectPost-partum haemorrhageen_US
dc.subjectPregnant womenen_US
dc.subjectKampala Districten_US
dc.subjectMachine learningen_US
dc.subjectPPHen_US
dc.subjectMaternal mortalityen_US
dc.subjectUgandaen_US
dc.titleDeveloping a predictive model to improve post-partum haemorrhage risk classification among pregnant women in Kampala District using machine learningen_US
dc.typeThesisen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record