106
reads

Yi-Ting Huang

2012-05-04
13:00:00 - 14:40:00

103 , Mathematics Research Center Building (ori. New Math. Bldg.)

Orthostatic hypotension (OH) is one kind of hypotension. Major characteristics of OH include dizziness and falls. A stroke patient with OH may fall when he or she performs the physical recovery. This may result in severe burden of medical costs. It is of important to verify whether a stroke patient has OH. To identify OH clinically, sequences of measurements such as systolic and diastolic pressures are needed and observed repeatedly. Owing to biological variations and measurement errors, the observed measures cannot be directly used. Furthermore, although a logistic regression can be used to predict a binary outcome, the parameter estimation may be problematic if covariates are highly correlated. Hwang et al. (2011) extended the two-stage model proposed by Tsiatis, DeGruttola, and Wulfsohn (1995) and the joint model proposed by Tsiatis and Wulfsohn (1997) for estimating the survival based on one repeated measure to predict a binary outcome based on one repeated measure. However, more than one repeated measure might influence the OH. Adapting the estimation method by Lin, McCulloch and Mayne (2002), this paper extends the joint model by Hwang et al. (2011) to predict a binary outcome using multiple repeated measures. Furthermore, the receive operating characteristic (ROC) curve and the area under the ROC curve (AUC) are adapted to estimate the predictive power of the proposed model. Monte Carlo simulations are performed to evaluate the feasibility of the proposed method. The result shows the estimates of variables of interest are accurate and the AUC can be used to estimate the predictive measure. In addition, the proposed model can be used to model a real data that has a predictive power equaling to 0.7.