2015/08/29 by R. Yates Coley, Aaron J. Fisher, Coley, R. Yates +9
Mathematics · Medicine · #Advanced Causal Inference Techniques #Applications (stat.AP) #FOS: Computer and information sciences #Methodology (stat.ME) #Prostate Cancer Diagnosis and Treatment #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1508.07511
openalex publication_date 2015/08/29 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In this article, we present a Bayesian hierarchical model for predicting a\nlatent health state from longitudinal clinical measurements. Model development\nis motivated by the need to integrate multiple sources of data to improve\nclinical decisions about whether to remove or irradiate a patient's prostate\ncancer. Existing modeling approaches are extended to accommodate measurement\nerror in cancer state determinations based on biopsied tissue, clinical\nmeasurements possibly not missing at random, and informative partial\nobservation of the true state. The proposed model enables estimation of whether\nan individual's underlying prostate cancer is aggressive, requiring surgery\nand/or radiation, or indolent, permitting continued surveillance. These\nindividualized predictions can then be communicated to clinicians and patients\nto inform decision-making. We demonstrate the model with data from a cohort of\nlow risk prostate cancer patients at Johns Hopkins University and assess\npredictive accuracy among a subset for whom true cancer state is observed.\nSimulation studies confirm model performance and explore the impact of\nadjusting for informative missingness on true state predictions. R code and\nsimulated data available at\nhttps://github.com/rycoley/prediction-prostate-surveillance.\n