SDS Seminar Series – Giovanni Toto, University of Texas at Austin

The Fall 2026 SDS Seminar Series continues on October 2nd from 2:00 p.m. to 3:00 p.m. with Giovanni Toto (Postdoctoral fellow, Department of Statistics & Data Science, UT Austin). This event is in-person in the Avaya Room (POB 2.302). 

Title: Condition matrix completion for polypharmacy medication information

Abstract: We propose a hierarchical Bayesian approach for matrix completion in presence of non-random missingness. Our method is motivated by a polypharmacy application in which patients' medical conditions are not directly observed. Insurance (ICD) codes and information on prescribed medications can be used to infer the unobserved conditions. In this application, the entries in the partially unknown matrix specify the presence of conditions, with patients as rows and conditions as columns. The binary entries are independent Bernoulli realizations. The success probabilities are modeled using a factor model, which provides a parsimonious representation of patient- and condition-specific characteristics. ICD codes are treated as noisy observations of the true conditions, while medication information provides additional evidence based on the set of conditions for which each medication is typically prescribed.