Artificial Intelligence in Pharmaceutical Clinical Decision-Making: A Scoping Review Protocol

This project is a scoping review entitled "Artificial Intelligence in Pharmaceutical Clinical Decision-Making." It maps the scientific evidence on artificial intelligence (AI) tools that pharmacists use to support clinical decision-making in pharmaceutical services. AI is increasingly used in healthcare for diagnosis, prognosis and clinical decision support. In pharmacy, it has potential in prescription analysis, drug interaction screening, dose adjustment and adverse event monitoring. However, its use in pharmacy practice is still incipient and mostly oriented toward operational tasks such as inventory management and process optimization. Evidence on the direct clinical use of AI by pharmacists, and on its effects on patient safety and clinical outcomes, remains scarce and dispersed. A scoping review suits this emerging, heterogeneous field, where there is not yet enough homogeneous evidence for a synthesis of effectiveness. The review is part of the project SisAPEC Farma, aims to develop AI tools to support pharmaceutical clinical decision-making. The review will therefore provide the scientific grounding for the system's pharmaceutical module. Objective The primary objective is to map the scientific evidence on the use of AI-based tools employed by pharmacists to support clinical decision-making in pharmaceutical services, in Brazil or in other countries. The research question follows the Population–Concept–Context (PCC) framework: Population: pharmacists involved in clinical services or direct patient care. Concept: the use, development, evaluation or implementation of AI-based tools or systems to support clinical decision-making. Context: clinical services provided by pharmacists in different practice settings (primary care, community pharmacies, hospitals, outpatient services and clinics). Methods The review follows the Joanna Briggs Institute (JBI) methodology, which builds on the Arksey and O'Malley framework. The protocol was written using the guidance and template of Lely et al. (2023), and the PRISMA extension for Scoping Reviews (PRISMA-ScR) will guide reporting. Eligibility: The review will include studies involving pharmacists working in clinical services or direct patient care. These studies must describe, develop, evaluate or implement AI tools (machine learning, deep learning, natural language processing, generative AI, expert systems and AI-based clinical decision support systems) applied to clinical pharmacy practice. Studies with peripheral pharmacist involvement will be excluded, as will studies focused exclusively on non-clinical activities (administrative management, logistics, inventory, dispensing automation) and studies on industrial or drug-discovery applications with no clinical pharmacy use. Only records with available full text will be included. No language or date limits will be applied. Search strategy: The strategy was developed for MEDLINE/PubMed and adapted to Embase, the Virtual Health Library (BVS), Scopus, Web of Science, CINAHL and the Cochrane Library. Grey literature (Google Scholar, ProQuest and OpenGrey), hand-searching of reference lists and expert consultation will complement it. Controlled vocabularies (MeSH, DeCS, Emtree) and free-text terms will be combined. Because Population and Context cover the same conceptual universe (pharmacists and the clinical pharmacy services in which they work), they are combined with OR, and this block is then combined with the AI concept block using AND. Selection and charting: Records will be deduplicated in EndNote 20 and screened in Rayyan by two independent, blinded reviewers in two stages (title/abstract, then full text). Disagreements will be resolved by consensus or by a third reviewer, and reasons for exclusion at full-text stage will be recorded. Data will be charted independently by two reviewers using a piloted Excel form adapted from JBI SUMARI. It will cover study identification, context and population, AI tool characteristics, results, barriers and facilitators, and gaps and opportunities. Synthesis: Analysis will be descriptive, using absolute and relative frequencies. Where pertinent, basic qualitative content analysis will combine an inductive and a deductive step. Results will be presented in tables, figures, narrative summaries and a PRISMA-ScR flow diagram. Any AI-based automation used to support the review will operate only under supervision, with all decisions validated by human reviewers. Expected Results The review is expected to produce a comprehensive map of the nature and extent of the evidence on AI tools used by pharmacists for clinical decision-making, including the types of tools, their functionalities and the clinical processes they support; a characterization of the available evidence on usability, acceptability, feasibility, performance, implementation and clinical impact, showing where evidence is concentrated and where it is lacking; an identification of the practice settings and countries in which these technologies have been most studied. A systematized overview of the technical, ethical, regulatory and organizational barriers and facilitators to implementing AI in clinical pharmacy; a clear identification of research gaps and unmet needs, pointing to opportunities for developing or improving AI solutions aligned with real clinical pharmacy needs and; by clarifying what is known and what is still missing, the review is expected to advance the field. It will support the safe, evidence-based development and implementation of AI in clinical pharmacy, including the pharmaceutical module of SisAPEC Farma. It will also inform future research, practice and policy.