Home | DTI | 2006–07 funded proposals | Paul E. Johnson, Gedas Adomavicius, Patrick J. O'Connor, William A. Rush
Paul E. Johnson, Gedas Adomavicius, Patrick J. O'Connor, William A. Rush
Improving Chronic Disease Care Using Data Mining Technologies
The problem of medical error is pervasive and increasing. One difficulty encountered by those who work on this problem is the issue of access to clinical databases for research purposes. Major issues include data scarcity, medical confidentiality, and security compliance. In the work proposed here, we seek to address these issues through the use of statistical models of real patient data and simulation-based models of physician practice patterns. Based on results from a large federally funded project we have developed models of physician decision making in environments of chronic disease care. We propose to use these models to identify physician practice patterns that predict patient treatment errors. The specific context of this project is managing patients with type 2 diabetes mellitus, which accounts for a large proportion of current healthcare expenditures. We first develop our work in simulated physician-patient environments and then test it using databases drawn from real clinical settings. More specifically, this project proposes to construct simulated clinical databases comprised of physician-patient encounters that reflect characteristics of real physician-patient databases. Using these simulated databases we will apply data mining techniques in order to extract models and patterns that are predictive of treatment errors, thus enabling the identification and correction of physician thinking before such errors occur.
Digital Technology Center
499 Walter Library, 117 Pleasant Street SE, Minneapolis, MN 55455
P: (612) 624-9510