A Novel Alternating Joint Longitudinal Model for Post-ICU Hemoglobin Prediction

• DOI: 10.48550/arxiv.2302.09110
• Publication Date: 2023-01-01
• Authors: Gabriel Demuth; Curtis B. Storlie; Matthew A. Warner; Daryl J. Kor; Phillip J. Shulte; Andrew C. Hanson
• Keywords: Proxy (statistics); Longitudinal Study

Anemia is common in patients post-ICU discharge. However, which will develop or recover from anemia remains unclear. Prediction of this population complicated by hospital readmissions, can have substantial impacts on hemoglobin levels due to surgery, blood transfusions, being a proxy for severe illness. We therefore introduce novel Bayesian joint longitudinal model over time, includes specific parametric effects admission and These themselves depend patient's at time hospitalization; given function that complete history admissions discharges up until time. because the an discharge do not themselves, well defined. validate our retrospective cohort 6,876 Rochester Epidemiology Project using cross-validation, find it accurately estimates predicts anemic status readmission 30 days post-discharge with AUCs .82 .72, respectively.