BAYESIAN JOINT MODELING OF LONGITUDINAL TUMOR SIZE AND SURVIVAL TIME AMONG CERVICAL CANCER PATIENTS, IN THE CASE OF HAWASSA UNIVERSITY CSH

Abstract

Cervical cancer is still a public health problem in low-resource settings, for which early diagnosis and continuous treatment are inaccessible. The present research considers the simultaneous dynamics of tumor size progression and time to death among cervical cancer patients at Hawassa University Comprehensive Specialized Hospital (HUCSH) from 2019 through 2025. With a Bayesian joint model approach, longitudinal tumor volume and survival times were both modeled together to reveal significant clinical and demographic predictors. The linear mixed-effects model controlled for individual tumor growth patterns, while the survival sub-model, with an Accelerated Failure Time (AFT) assumption, estimated risk factors with implications on patient survival. Tumor stage, comorbidities, baseline weight and baseline hematocrit levels were found to have awesome implications on tumor growth as well as survival. The Bayesian approach generated robust inferences from posterior distributions and allowed better clinical insight into the evolution of the disease. The model provided confirmation of the widely accepted robust association between evolution of tumor size and increased risk of death and estimated the strongest sociodemographic and clinical predictors on the outcomes. These results provide practical implications to improve treatment plans and patient follow-up in such similar resource-poor settings.