Clinical data mining: a review.

摘要: Objective: Clinical data mining is the application of techniques using clinical data. We review literature in order to provide a general overview by identifying status-of-practice and challenges ahead. Methods: The nine steps proposed Fayyad 1996 [4] were used as main themes review. MEDLINE was primary source 84 papers retained based on our inclusion criteria. Results: has three objectives: understanding data, assist healthcare professionals, develop analysis methodology suitable for medical Classification most frequently function with predominance implementation Bayesian classifiers, neural networks, SVMs (Support Vector Machines). A myriad quantitative performance measures accuracy, sensitivity, specificity, ROC curves. latter are usually associated qualitative evaluation. Conclusion: respects its commitment extracting new previously unknown knowledge from databases. More efforts still needed obtain wider acceptance professionals generalization reproducibility extraction process: better description variables, systematic report algorithm parameters including method them, use easy-to-understand models comparisons efficiency traditional statistical analyses. more will be available miners they have methodologies infrastructures analyze increasingly complex