Predictive data mining to support clinical decisions: An overview of heart disease prediction systems

Healthcare organizations are faced with challenges to provide cost-effective and high quality patient care. Both administrators and clinicians need to analyze a wealth of data available in the databases of healthcare information systems in order to discover knowledge and to make informed decisions. This is critical in particular to enhance the effectiveness of disease treatment and preventions. It becomes of more important in case of heart disease (HD) that is regarded as the primary reason behind death in adults. Data mining serves as an analysis tool to discover hidden relationships and patterns in HD medical data. This paper reviews five models constructed of single and combined data mining techniques to support clinical decisions in (HD) diagnosis and prediction. The five systems provide automatic pattern recognition and attempts to uncover relationships among different parameters and symptoms of HD. Each system exhibits set of strengths and limitations in terms of the type of data it handles, accuracy, ease of interpretation, reliability and generalization ability. Poor generalization ability is still a major open issue for data mining in healthcare mainly because of the lack of input data and cost of re-processing.