Artificial Intelligence in Pharmacy: Predicting Adverse Drug Reactions

resumen

Moamen Abdelfadil Ismail*, Shahad Mari Alshahrani, Abdullah Ali Ibrahim Alshahrani, Ahmed Yahya, Hassan Alshehri, Khalid Saeed Abdullah Alshahrani, Ahmed Bakheet Attiah Al-Malki, Abeer Salamah Alsharif, Miral Majed Alsherbi, Jihad Saleh Alrehaili, Mohammed Saeed Aftan, Razan Ahmed Salman Alsharif, Abdullah Mahmoud Bedaiwi, Farah Abdullah Awad Alahmadi, Sumayyah Masoud Dakheel Alrhili, Naif Abdulrahman Y. Alfaifi

Background: Adverse drug reactions (ADRs) are a major cause of morbidity, hospitalizations, and healthcare costs. Traditional pharmacovigilance methods are often limited by underreporting and delays. Artificial intelligence (AI), particularly machine learning (ML) and natural language processing (NLP), offers faster, more accurate ADR detection by integrating diverse data sources such as electronic health records and clinical notes.

Methods: A systematic review was conducted following PRISMA guidelines, searching PubMed, Scopus, IEEE Xplore, Web of Science, and Google Scholar for English-language studies published from January 2010 to May 2025. Eligible studies applied AI/ML methods to ADR prediction in pharmacy settings. Two reviewers independently screened and extracted data, with risk of bias assessed using PROBAST. A narrative synthesis was used due to methodological heterogeneity, categorizing studies by AI technique and application area.

Results: Nineteen studies met inclusion criteria. Deep learning and random forest models achieved ADR detection accuracies up to 89.4% and c-indices above 0.91. AI-based dosing tools improved safety for drugs like vancomycin and warfarin. Drug interaction predictors (e.g., XGBoost) exceeded 94% accuracy. Unsupervised models flagged rare prescription errors with >95% precision. AI systems reduced dispensing errors by >75% and improved documentation. Medication therapy management supported by AI lowered care costs by 19.3% and reduced hospital visits.

Conclusions: AI consistently outperforms traditional methods in ADR prediction, dosage optimization, and error prevention. Its integration into pharmacy practice could enhance patient safety, personalize therapy, and reduce healthcare costs. Standardized validation and transparent, ethical implementation are essential for clinical adoption.