Biostatistics Seminar speaker, Dr. Amita Manatunga, Emory University, will present, “A Modeling Approach for Predicting Disease Status Using Functional Data in the Absence of a Gold Standard.”
Statistical methods can play a vital role in risk predictions that can help to inform patients as well as to guide important medical decision-making. Data from clinical studies involving risk predictions provide a wealth of opportunities for statistical research in particular to the development of prediction models and their evaluations. Based on our recently funded NIH statistical methodological grant, I will introduce a specific clinical decision making problem in nuclear medicine, present its statistical challenges and discuss some potential solutions. In our study, two consecutive curves over time are observed per subject and in some cases, the second curve for some subjects are not observed. There is no gold standard for determining disease status, instead, the ratings for disease status from multiple experts are available We consider a latent class modeling approach for predicting disease status of a subject based on observed functional data and its ratings from multiple experts. I will present our preliminary work including the modeling procedure, prediction models consisting of several prediction schemes, and their evaluation via simulation studies. I will demonstrate the practicality of our method using a preliminary renal study, which motivated the grant application. The proposed modeling procedure reasonably captures the patterns of observed curves and provide sensible clinical interpretations. I will conclude with a brief discussion of future work.