562 – Joint Modeling
A Joint Latent Class Analysis for Adjusting Survival Bias with Application to Trauma Transfusion Study
Jing Ning
MD Anderson Cancer Center
Mohammad Hossein Rahbar
University of Texas Health Science Center at Houston
Sangbum Choi
University of Texas at Houston
Chuan Hong
Jin Piao
University of Texas at Houston
Deborah J. del Junco
The University of Texas Health Science Center at Houston
Erin E. Fox
The University of Texas Health Science Center at Houston
Elaheh Rahbar
The University of Texas Health Science Center at Houston
John B. Holcomb
The University of Texas Health Science Center at Houston
There is no clear classification rule to identify trauma patients who have severely hemorrhaged and may need substantial blood transfusions. A surrogate measure of severe hemorrhage, massive transfusion, has traditionally been defined as the transfusion of at least 10 units of red blood cells (RBCs) within 24 hours of emergency department admission. This definition suffers from misclassification due to a survival bias that arises because such patients may die before 24 hours. Accordingly, we propose a latent class model that adjusts for the survival bias by incorporating baseline information at emergency department admission, observed number of RBC units transfused, and survival time. The statistical challenges include induced dependent censoring for the amount of RBCs transfused. We propose a pseudo-likelihood function by using the inverse weighting principle and develop an expectation-maximization algorithm for the estimation. We evaluate the performance of the proposed method in classifying patients with severe hemorrhage and compare it to the existing definition of massive transfusion through simulations and an application to the Prospective Observational Multi-center Major Trauma Transfusion study.