HIV Patient Monitoring Framework Through Knowledge Engineering

Uganda has registered more than a million deaths since the HIV virus was first offi¬cially reported in the country over 3 decades ago. The governments in partnership with different groups have implemented different programmes to address the epidemic. The support from different donors and reduction in prices of treatment resulted in the focus on antiretroviral therapy access to those affected. Presently only a quarter of the approximately 1 million infected by HIV in Uganda are undergoing antiretroviral therapy. The number of patients pause a challenge in monitoring of therapy given the overall resource needs for health care in the country. Furthermore the numbers on antiretroviral therapy are set to increase in addition to the stringent requirements in tracking and monitoring of each individual patient during therapy. This research aimed at developing a framework for adopting knowledge engineering in information systems for monitoring HIV/AIDS patients. An open source approach was adopted due to the resource constrained context of the study to ensure a cost effec¬tive and sustainable solution. The research was motivated by the inconclusive literature on open source dimensional models for data warehouses and data mining for monitor¬ing antiretroviral therapy. The first phase of the research involved a situational analysis of HIV in health care and different health care information systems in the country. An analysis of the strengths, weaknesses and opportunities of the health care system to adopt knowledge bases was done. It proposed a dimensional model for implementing a data warehouse focused on monitoring HIV patients. The second phase involved the development of a knowledge base inform of an open source data warehouse, its simulation and testing. The study involved interdisciplinary collaboration between different stakeholders in the research domain and adopted a participatory action research methodology. This involved identification of the most appropriate technologies to foster this collabora¬tion. Analysis was done of how stakeholders can take ownership of basic HIV health information system architecture as their expertise grow in managing the systems and make changes to reflect even better results out of system functionality. Data mining simulations was done on the data warehouse out of which two machine learning algorithms (regression and classification) were developed and tested using data from the data warehouse. The algorithms were used to predict patient viral load from CD4 count test figures and to classify cases of treatment failure with 83% accu¬racy. The research additionally presents an open source dimensional model for moni¬toring antiretroviral therapy and the status of information systems in health care. An architecture showing the integration of different knowledge engineering components in the study including the data warehouse, the data mining platform and user interac-tion is presented