StratMed: Relevance Stratification between Biomedical Entities for Sparsity on Medication Recommendation

TL;DR

StratMed introduces a relevance stratification approach with a dual-property network to improve medication recommendation accuracy and safety, especially for sparse data, demonstrating significant performance gains on the MIMIC-III dataset.

Contribution

The paper proposes a novel stratification strategy and dual-property network to better handle long-tailed data distribution and balance safety and accuracy in medication recommendation.

Findings

Reduces safety risk by 15.08%

Improves accuracy by 0.36%

Reduces training time by 81.66%

Abstract

With the growing imbalance between limited medical resources and escalating demands, AI-based clinical tasks have become paramount. As a sub-domain, medication recommendation aims to amalgamate longitudinal patient history with medical knowledge, assisting physicians in prescribing safer and more accurate medication combinations. Existing works ignore the inherent long-tailed distribution of medical data, have uneven learning strengths for hot and sparse data, and fail to balance safety and accuracy. To address the above limitations, we propose StratMed, which introduces a stratification strategy that overcomes the long-tailed problem and achieves fuller learning of sparse data. It also utilizes a dual-property network to address the issue of mutual constraints on the safety and accuracy of medication combinations, synergistically enhancing these two properties. Specifically, we construct a pre-training method using deep learning networks to obtain medication and disease representations. After that, we design a pyramid-like stratification method based on relevance to strengthen the expressiveness of sparse data. Based on this relevance, we design two graph structures to express medication safety and precision at the same level to obtain patient representations. Finally, the patient's historical clinical information is fitted to generate medication combinations for the current health condition. We employed the MIMIC-III dataset to evaluate our model against state-of-the-art methods in three aspects comprehensively. Compared to the sub-optimal baseline model, our model reduces safety risk by 15.08\%, improves accuracy by 0.36\%, and reduces training time consumption by 81.66\%.