Accuracy of screening procedures using multilevel modeling for screening and outcome data
Date of Completion
January 2004
Keywords
Statistics|Health Sciences, Radiology
Degree
Ph.D.
Abstract
The phenomenon of observer variability has been studied in many areas of clinical medicine such as psychiatry, pathology, physical diagnosis and radiology. Observer variability is a major cause of concern for clinical procedures that require subjective decision. The efficacy of such procedures evidently depend on the subjectivity in interpretation. Previous studies have established interpretive variation in the field of mammography under test setting. In this thesis, we quantify the interpretive variability among the radiologists in the community setting. It is believed that fiscal, legal and practice environment along with radiologist's personal characteristics contribute to the variability in the field of mammography. This thesis uses hierarchical modeling techniques to understand how radiologist's characteristics affect this accuracy. We further define and discuss estimation of conditional rates. Conditional rates are defined as measures of accuracy when only a subset of the available risk factors are specified. We argue that, based upon the joint distribution of the event outcome, the screening result and the risk factor occurrence, a formal expression for such a rate can be obtained. ^
Recommended Citation
Paliwal, Prashni, "Accuracy of screening procedures using multilevel modeling for screening and outcome data" (2004). Doctoral Dissertations. AAI3144604.
https://digitalcommons.lib.uconn.edu/dissertations/AAI3144604