Data mining of massive datasets in healthcare

Managing the distribution of healthcare is seen to involve data collected at every encounter of each patient with any provider, pharmacy, payer, or government agency. These data and their analysis are massive on multiple dimensions: of patient-encounter records; of variables (administrative, diagnosti,c and procedural) and their derived indicators; of the related clinical knowledge resources; of the clinical and administrative issues to be addressed; and of the diversity of the audience for the analysis. Statistical and computational strategies for massive analysis of these massive data include a principle of ``layered recalibration{''} used to maintain statistical models, and an emphasis on presentation, including report generation, and specialized software systems. These ideas are implemented in the Performance iQ products of QuadraMed Corp. developed by the author.