GRAIN: Molecules Are Not the Right Granularity -- Active-Ingredient Modeling for Safe Medication Recommendation

2026-07-30

AI Summary

This work addresses the challenge of balancing recommendation accuracy and drug–drug interaction (DDI) safety in polypharmacy scenarios, where existing medication recommendation systems—typically modeling at the drug or molecular substructure level—struggle to achieve both objectives simultaneously. The study proposes a novel paradigm centered on active pharmaceutical ingredients, leveraging RxNorm for ingredient normalization. It integrates temporal modeling of patient clinical trajectories via selective state space models with multi-source knowledge graphs, including drug-level and ingredient-level DDI networks as well as co-prescription graphs. A proportional controller is introduced to dynamically balance accuracy and safety during recommendation. Experiments on MIMIC-IV demonstrate a Jaccard similarity of 0.4983, a 50% reduction in drug-level DDI rate to 0.0948, and the first formal definition and empirical validation of an ingredient-level DDI rate.

Abstract

Medication recommendation from electronic health records must balance predictive accuracy against the risk of adverse drug-drug interactions (DDIs) under polypharmacy. Existing safety-aware recommenders operate at one of two granularities: the drug code, which treats each medication as an indivisible token, or the molecular substructure, which is finer than pharmacological interaction knowledge is actually organized. We argue that the active ingredient is the missing granularity, and introduce GRAIN, a medication recommendation framework built around it. GRAIN encodes longitudinal patient trajectories (diagnoses, procedures, past medications) with a selective state space backbone that handles long, irregular visit sequences in linear time. On top of it we introduce a joint objective unifying three knowledge sources aligned to a common medication vocabulary: a drug-level DDI graph, an ingredient-level DDI graph obtained by normalizing medication codes to active ingredients via RxNorm, and an EHR-derived co-prescription graph. A proportional controller adapts the accuracy-safety trade-off to the observed validation DDI rate rather than fixing it a priori. Under strictly matched settings -- identical preprocessing, cohort, vocabulary, split, and evaluation code -- GRAIN improves over a re-implemented MambaHealth baseline on MIMIC-IV across all standard multi-label metrics (Jaccard 0.4488 to 0.4983, PRAUC 0.6911 to 0.7485, F1 0.5989 to 0.6453) while reducing the drug-level DDI rate from 0.1875 to 0.0948. We further define an ingredient-level DDI rate, a safety measure invisible to drug-code-level evaluation. The results indicate that ingredient-level normalization recovers predictive signal erased by code-level aggregation, and that it is complementary to, rather than in competition with, accurate sequence modeling.

Problem

• medication recommendation
• drug-drug interactions
• active ingredient
• granularity
• polypharmacy

Innovation

• active ingredient
• medication recommendation
• drug-drug interaction
• state space model
• safety-aware AI