With the rising costs of health care, the aging American population, and the burden of diabetic morbidity, there is increasing interest in understanding the progression and natural history of type 2 diabetes and its complications. Due to the complexity of diabetes and the prolonged development of complications, a single longitudinal study of progression is expensive and quickly outdated. For this reason, diabetes researchers use computer models, with secondary data analysis, to integrate findings from recent studies and to predict the long-term benefits of early preventive intervention. However, current computer models of diabetes have limitations including 1) constraint to a single study, which requires considerable time and expense to update when new theories arise; and 2) poor estimation procedures that cannot use data from studies with differing theoretical models. Standard estimation techniques cannot use data that group nodes (e.g., do not differentiate between normo- and micro- albuminuic patients) or omit nodes (e.g., ignore angina in a model of cardiovascular disease). Recent statistical developments use these complementary data (that group or omit nodes) but are inefficient for covariates.
The specific aims of this project are to advance the methodology of chronic disease modeling by: 1) Deriving methods to use complementary data for indirect estimation with both continuous and categorical covariates, 2) Developing software for model simulation using indirect estimation, and 3)To provide internal and external validation of our indirect estimates using the Translating Research Into Action for Diabetes (TRIAD) data. We will develop statistical likelihood theory for these data, investigate large- and small- sample properties of our estimators via simulation, derive necessary conditions for estimability, and validate the model internally and externally using secondary data analysis of the TRIAD study. In response to this PA's objective """"""""to promote methodological advances in epidemiologic research using secondary data analysis"""""""", this project develops analytic tools and techniques for epidemiologic disease models using secondary data. These tools and techniques, together, enable two goals of this PA: the ability to test complex models using existing data, and rapid analysis of new data. These tools allow integration of existing data with new data as an economic alternative to expensive and time-consuming longitudinal studies to inform the design and content of new studies. The long-term goal of this project is to apply these tools and techniques to other disease areas of the NIDDK such as obesity and kidney disease. ? ? ?
| Ye, Wen; Isaman, Deanna Jm; Barhak, Jacob (2012) Use of Secondary Data to Estimate Instantaneous Model Parameters of Diabetic Heart Disease: Lemonade Method. Inf Fusion 13:137-145 |
| Barhak, Jacob; Isaman, Deanna J M; Ye, Wen et al. (2010) Chronic disease modeling and simulation software. J Biomed Inform 43:791-9 |
| Isaman, Deanna J M; Barhak, Jacob; Ye, Wen (2009) Indirect estimation of a discrete-state discrete-time model using secondary data analysis of regression data. Stat Med 28:2095-115 |
