Assessing Regression to the Mean and Intervention Effect;Rept. for 1 Aug 01-30 Jun 03

摘要

The authors conducted research to study the implications of and provide statistical solutions to regression effects when assessing the effectiveness of an intervention. The authors developed methods to estimate the intervention effect and adjust for the regression effect in a nonrandomized study with a mismeasured binary response and a binary response outcome. The extended the proposed methods to estimate the placebo and intervention effects in placebo-controlled studies and conducted a number of simulation studies to mimic actual medical situations. The authors quantified regression effect in various probable clinical scenarios, and proposed statistical methods to remove bias. The authors developed a statistical model for estimating the treatment effect where the selection variable is continuous and the outcome is binary, but were unable to obtain a converged estimation of treatment effect. The authors proposed four possible types of validation studied to be applied to single-armed, non-controlled studies. Finally, the authors analyzed restenosis and otitis media data for illustration. The proposed research is innovative and significant because current methods for adjustment for RTM have focused on normally distributed data, and are no longer adequate to cope with the complexities of health services outcomes and clinical studies in the real world.