Author: Lachenbruch Peter
Publisher: Taylor & Francis Ltd
ISSN: 0361-0926
Source: Communications in Statistics: Theory and Methods, Vol.38, Iss.18, 2009-01, pp. : 3268-3281
Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.
Abstract
This article discusses predicting damage in patients with juvenile myositis after treatment with various medications. These data were taken from medical records and not randomized. Investigators advocate using propensity scores for analysis of such non randomized studies in order to reduce the effect of selection of treatment. Thus far, the studies have typically been comparing drug administration versus no drug after including a propensity score to compensate for potential bias in selecting patients for use of the agent. In this study, we use propensity scoring for multiple treatments given singly or in combination. We study two methods. We use a multiple logistic regression model with continuous propensity scores and a model that develops strata based on the dichotomous treatment assignment (received drug or not). We find the multiple logistic regression models predict damage better than the dichotomous model. In many cases, the propensity score also accounts for the effect of the treatment.
Related content
Design of Panel Studies for Disease Progression with Multiple Stages
By Hwang W-T.
Lifetime Data Analysis, Vol. 9, Iss. 3, 2003-09 ,pp. :
Statistics in Biosciences, Vol. 5, Iss. 2, 2013-11 ,pp. :
By Vanderweele Tyler J. Hong Guanglei Jones Stephanie M. Brown Joshua L.
Journal of the American Statistical Association, Vol. 108, Iss. 502, 2013-06 ,pp. :