El uso la inteligencia artificial en farmacovigilancia en pacientes con enfermedades crónicas: Una revisión bibliográfica

Abstract

The actions involved in pharmacovigilance of patients with chronic diseases have made it possible to guarantee the safety of drugs and the applications of artificial intelligence developing these actions. The objective of this work was to evaluate the efficacy, safety, and feasibility of AI used in pharmacovigilance in chronic diseases, through a bibliographic review. The methodology followed PRISMA guidelines: search in PubMed, Scopus, Web of Science, ScienceDirect and Redalyc, titles and abstracts were selected, applying inclusion and exclusion criteria, finding 10 articles for review. The results show a diverse use of techniques: natural language processing (NLP) and TF-IDF to extract ADRs from unstructured texts, disproportionality algorithms (PRR/ROR) in large report databases, Machine Learning models (Random Forest, neural networks) for risk stratification, rule mining (A priori) for polypharmacy patterns, OCR to improve adherence and advanced techniques (NAS, Bayesian) in oncology; Advantages include early signal detection, improved risk stratification, therapy optimization for polypharmacy users, and time savings; however, it faces disadvantages such as data bias, underreporting in spontaneous reports, false alerts, and privacy and data security challenges. On the other hand, the geographic comparison revealed that the United States and the United Kingdom lead in ML and Big Data, Europe uses mixed methods, China and Korea apply digital platforms and OCR, and India explores advanced optimizations. In conclusion, AI holds promise for improving pharmacovigilance in chronic diseases, but its implementation requires interoperability, data governance, anonymization, and ongoing clinical validation.