Blood Transfusion After Lung Transplant Using Two-Step Nonlinear Multiphase Generalized Mixed Models

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Jeevanantham Rajeswaran*+ and Eugene H. Blackstone
• Companies: Cleveland Clinic and Cleveland Clinic
• Keywords: generalized mixed models ; multi-phase longitudinal models ; non-linear generalized mixed models

Abstract

Blood transfusion has been shown to impact outcomes after surgery. But, little is known about patient factors influencing the long-term transfusion. We present a 2-step separate modeling to analyze the longitudinal transfusion data. Step 1: A multi-phase logit mixed model presented to estimate time-related prevalence of transfusion and their determinants. Step 2: A multi-phase Poisson mixed model presented to estimate the expected number of units of red blood cell transfused and their determinants. These models enable us to estimate patient-specific risk factors and population-specific prevalence and expected number of units transfused. The main features of these models are: 1. Temporal pattern can be decomposed into different time phases (eg. early, and late) using a time function; 2. Influence of covariates on transfusion may or may not change over time (time-varying coefficients).