Data mining methods for pharmacovigilance in Africa

This work reviewed and applied methods for analysis of adverse drug reactions and drug utilization. This will help to increase knowledge about the burden of drug safety in African region. The burden is a result of multiple factors like medication errors, irrational drug use, poor drug quality, adverse reactions and counterfeits. Moreover, adequate decision making processes needs easy-to-use analytical methodologies for evidence-based decisions despite the scarcity of data. The first and second parts of the thesis respectively reviewed and applied the data mining algorithms for detection and evaluation of safety signals. This involved looking at the commonly used data mining algorithms applied to spontaneous adverse reports and applying them to late phase clinical trials safety pediatric database. The third section utilized the standard data mining analysis method of classification and regression tree (CART) to determine the predictors of prescription practices in rural based health facilities of Africa. The use of data mining and biostatistical methods for handling drug safety and utilization data collected using diverse approaches like the traditional passive reporting systems, electronic medical records, clinical trials and post marketing surveys was demonstrated.