Premium
Joint model of multiple longitudinal measures and a binary outcome: An application to predict orthostatic hypertension for subacute stroke patients
Author(s) -
Hwang YiTing,
Wang ChunChao,
Wang Chiuan He,
Tseng YiKuan,
Chang YeuJhy
Publication year - 2015
Publication title -
biometrical journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.108
H-Index - 63
eISSN - 1521-4036
pISSN - 0323-3847
DOI - 10.1002/bimj.201400044
Subject(s) - orthostatic vital signs , logistic regression , stroke (engine) , medicine , receiver operating characteristic , blood pressure , statistics , covariate , tilt (camera) , rehabilitation , physical therapy , physical medicine and rehabilitation , mathematics , engineering , mechanical engineering , geometry
Stroke patients with orthostatic hypertensive responses that are one of the blood pressure regulation problems can easily fall down while doing rehabilitation, which may result in prolonged hospitalization and delayed treatment and recovery. This may result in increasing the medical cost and burden. In turn, developing a diagnostic test for the orthostatic hypertension (OH) is clinically important for patients who are suffering from stroke. Clinically, a patient needs to have a tilt testing that requires measuring the change of blood pressures and heart rate at all angles to determine whether a stroke patient has OH. It takes lots of time and effort to perform the test. Assuming there exist measurement errors when obtaining the blood pressures and heart rate at all angles, this paper proposes using multiple mixed‐effect models to obtain the true trajectories of these measurements, which take into account the measurement error and the possible correlation among multiple measurements, and a logistic regression uses these true trajectories at a given time and other fixed‐effect covariates as predictors to predict the status of OH. The joint likelihood function is derived to estimate parameters and the area under the receiver operating characteristics curve is used to estimate the predictive power of the model. Monte Carlo simulations are performed to evaluate the feasibility of the proposed methods. Also, the proposed model is implemented in the real data and provides an acceptable predictive power.