Table 1 Introduction to baseline models

KAME [22]

Utilizes medical ontologies to learn representations of medical codes and their hierarchies, which are then input into a neural network to predict sequential diagnoses

MMORE [15]

Generates multiple representations for each disease diagnosis via attention mechanisms, offering clinically enriched sub-classifications

SETOR [18]

Employs neural ordinary differential equations to manage irregular intervals between patient visits and captures dependencies through multi-layer transformer blocks, integrating medical ontologies to enhance data scarcity challenges