Statistical analysis and data mining of Medicare patients with diabetes.

Statistical analysis;Diabetes;Data mining;Medicare

Xiao Wang

University:University of Louisville

Department:Mathematics

关键词: Statistical analysis;    Diabetes;    Data mining;    Medicare;   

Others  :  https://ir.library.louisville.edu/cgi/viewcontent.cgi?article=2508&context=etd

美国|英语

来源: The Universite of Louisville's Institutional Repository

【 摘 要 】

The purpose of this dissertation is to find ways to decrease Medicare costs and to study health outcomes of diabetes patients as well as to investigate the influence of Medicare, part D since its introduction in 2006 using the CMS CCW (Chronic Condition Data Warehouse) Data and the MEPS (Medical Expenditure Panel Survey) data. In this dissertation, we introduce pattern recognition analysis into the study of medical characteristics and demographic characteristics of the inpatients who have a higher readmission risk. We also broaden the cost-effectiveness analysis by including medical resources usage when investigating the effects of Medicare, part D. In addition, we apply several statistical linear models such as the generalized linear model and data mining techniques such as the neural network model to study the costs and outcomes of both inpatients and outpatients with diabetes in Medicare. Moreover, some descriptive statistics such as kernel density estimation and survival analysis are also employed.One important conclusion from these analyses is that only diseases and procedures, rather than age are key factors to inpatients' mortality rate. Another important discovery is that at the influence of Medicare part 0, insulin is the most efficient oral anti-diabetes drug treatment and that the drug usage in 2006 is not as stable as that in 2005. We also find that the patients who are discharged to home or hospice are more likely to re-enter the hospital after discharge within 30 days. Two - way interaction effect analysis demonstrates that diabetes complications interact with each other, which makes healthcare costs and health outcomes different between a case with one complication and a case with two complications. Accordingly, we propose some useful suggestions. For instance, as for how to decrease Medicare payments for outpatients with diabetes, we suggest that the patients should often monitor their blood glucose level. We also recommend that inpatients with diabetes should pay more attention to their kidney disease, and use prevention to avoid such diseases to decrease the costs.