A Comparative Time-to-Event Analysis Across Health Systems

Abstract

While the wide spread adoption of EHR allows access to clinical data for AI modeling, constructed cohorts at institutional levels tend to be small due to inevitable data sparsity issues, which poses a challenge to build interoperable models performing well across health systems. This work addresses such interoperability challenges, focusing on the development of time-to-event models to estimate Type 2 Diabetes Mellitus (T2D) complication times.