Local Estimation of Patient Prognosis

• Activity Number: 262
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Alison Kosel* and Patrick Heagerty
• Companies: and University of Washington
• Keywords: local prediction ; non-parametric

Abstract

We develop an algorithm and associated inference for creating local, nonparametric predictions of patient prognosis. The intended goal of the methods is to provide an estimate of the full outcome distribution for a given subject by providing summary data for a specific axis-parallel neighborhood with a fixed subset size. We develop inference for the local predictions and implement the methods using a dynamic computational interface. We illustrate the methods with a large electronic health records based back pain cohort, and comment on extensions of the methods to comparative estimation and longitudinal data.