JMIR Publications

ABSTRACT

Background:

Electronic health records (EHR) store rich data involving medical entities such as diagnoses, treatment procedures, and prescribed medications, offering valuable resources for developing automated systems for hypertension medication recom-mendations. The entities in EHR show significant synergies during treatment. How-ever, existing medication recommendation methods mainly focus on homogeneous graphs, overlooking the crucial synergistic relationships among heterogeneous medical entities. Also, accurately modeling the progression of hypertension using EHR is essential for precise medication recommendations, but current approaches often lack comprehensive temporal modeling and don't fully meet clinical require-ments.

Objective:

To overcome the challenges in existing methods, introduce a novel model for hy-pertension medication recommendation that leverages the synergy and selectivity of heterogeneous medical entities.

Methods:

First, use patient EHR to construct both heterogeneous and homogeneous graphs. Then, capture the inter-entity synergies with a multi-head graph attention mecha-nism to enhance entity-level representations. Next, apply a dual-layer temporal se-lection mechanism to calculate selective coefficients between current and historical visit records and aggregate them to form refined visit-level representations. Finally, determine medication recommendation probabilities based on these comprehen-sive patient representations.

Results:

Experimental evaluations on the real-world dataset MIMIC-IV v2.2 show that the model achieves a Jaccard similarity coefficient of 55.82%, a precision-recall AUC of 80.69%, and an F1 score of 64.83%, outperforming baseline models.

Conclusions:

The findings indicate the superior efficacy of the introduced model in medication recommendation, highlighting its potential to enhance clinical decision-making in the management of hypertension.