Multimodal Pharmacy Intelligence for Linking Prescriptions, Symptoms, Laboratory Trends, Images, and

Medication decisions are rarely supported by prescriptions alone. Pharmacists may need to relate medication orders to reported symptoms, longitudinal laboratory changes, clinical images, spoken information, documentation context, and prior therapeutic events. These information forms differ in structure, reliability, timing, provenance, and clinical meaning, creating a problem that cannot be resolved through simple data concatenation or a single predictive model. This article proposes a Multimodal Pharmacy Intelligence Architecture for organizing heterogeneous patient information while preserving uncertainty and professional decision authority. The architecture contains source and identity controls, modality-specific ingestion and representation, provenance and data-quality assessment, temporal normalization, cross-modal alignment, task-specific fusion, uncertainty management, and a pharmacist-facing evidence bundle. A proposed Pharmacy Alignment Graph represents relationships among observations without treating temporal proximity as causality. Missing, stale, conflicting, or low-quality modalities remain visible rather than being silently imputed or suppressed. A Pharmacist Decision Boundary separates computational transformation from medication verification, interpretation, communication, escalation, and action. Evaluation is organized across representation integrity, alignment accuracy, missing-modality robustness, task performance, human–AI interaction, workflow consequences, medication safety, equity, and local transportability. The architecture is an original non-empirical synthesis rather than a validated clinical system. Its usefulness is conditional on task-specific evidence, semantic interoperability, prospective human-factors assessment, subgroup evaluation, incident monitoring, and governed change control. Multimodal agreement does not establish clinical truth, and model output does not independently authorize medication decisions. The proposed contribution offers a structured basis for developing and evaluating traceable multimodal decision support in pharmacy practice.