YSPH Biostatistics Seminar: “Versatile Deep Learning Provider Profiling: A Design-Based Approach”

NOTE: BIS 525 students are required to attend in person. Others are invited to attend in person, but may also attend via Zoom.

SPEAKER: Wenbo Wu, Assistant Professor, Department of Population Health, New York University Grossman School of Medicine

TITLE: “Versatile Deep Learning Provider Profiling: A Design-Based Approach”

ABSTRACT: Provider profiling is a quality assessment process in which the performance of hospitals and clinicians is compared based on patient-centered outcomes, and outlying providers with significantly subpar services are identified. Encompassing numerous initiatives in the United States, provider profiling has evolved into a major health care undertaking with ubiquitous applications, profound implications, and high-stakes consequences. In line with such a significant profile, the literature has accumulated an enormous collection of articles dedicated to enhancing the statistical paradigm of provider profiling. Tackling wide-ranging profiling issues, these methods typically adjust for risk factors using linear predictors. While this simple approach generally leads to reasonable assessments, it can be too restrictive to characterize complex factor-outcome associations. Secondly, conventional methods, having been historically driven by the demand for controlling care expenditures, tend to amalgamate all racial/ethnic groups without accounting for their socioeconomic diversity. Thirdly, despite the paramount importance of distinguishing between cost-driven and equity-driven profiling, a methodological framework capable of addressing these different but related objectives is still lacking, due in part to the absence of a unified framework defining objective-oriented performance benchmarks. To address these issues, we consider a versatile probabilistic method based on so-called provider comparators, defined as hypothetical reference providers that correspond to specific profiling objectives. In addition, we develop a flexible deep learning approach that relaxes the linearity assumption underpinning existing profiling methods. The advantages of the proposed methods are demonstrated through simulation experiments and the profiling of kidney dialysis facilities using 2020 Medicare claims.

YSPH values inclusion and access for all participants. If you have questions about accessibility or would like to request an accommodation, please contact Charmila Fernandes at Charmila.fernandes@yale.edu. We will try to provide accommodations requested by October 26, 2023.