Artificial Intelligence for the Future of Pharmacovigilance in Pakistan: Barriers, Opportunities, and Comparison with Other Low and Middle-Income Countries
DOI:
https://doi.org/10.61980/3049-7639/2026/4/40Keywords:
Pharmacovigilance (PV), Artificial Intelligence (AI), Adverse Drug Reactions (ADRs), Low- and Middle-Income Countries (LMICs) , Electronic Health Records (EHRs),, Natural Language Processing (NLP), Drug SafetyAbstract
Artificial intelligence (AI) is progressing day by day, and with it, the number of unreported adverse drug reactions (ADRs) linked to inadequate pharmacovigilance (PV) in Pakistan is increasing. PV is defined as the science of identifying, evaluating, and preventing side effects or other drug-related issues in order to make sure that the advantages of medications balance the risks over the course of their whole lifecycle. It has been estimated that more than 75% of healthcare professionals display poor practices when it comes to reporting ADRs. However, the majority of them, in principle, have a positive attitude towards reporting. In comparison, self-medication completely bypasses the formal system and is carried out by over 61.2% of the population in certain samples. Ongoing problems such as underreporting, insufficient professional awareness, a fragmented infrastructure, and linguistic diversity continue to limit the country's ability to identify medication-related harm on a large scale. Internationally, AI has been suggested as a means of overcoming this gap by enabling real-time detection of ADRs, using predictive analytics, and automatically extracting adverse events from unstructured clinical text. This narrative review brings together the available evidence on Pakistan's current PV system, compares its performance with that of other low- and middle-income countries (LMICs) using the World Health Organization (WHO) Indicator-based Pharmacovigilance Assessment Tool (IPAT), and sets out a route for the implementation of AI-assisted PV. We drew on evidence from AI-based pharmacovigilance studies in low-resource settings, research using electronic health records, and a controlled trial of a mobile ADR-reporting application in Uganda to identify strategies relevant to Pakistan. Based on this evidence, the immediate priority should be to build sound policies and a qualified workforce rather than focusing on advanced modelling. Once these foundations are in place, lightweight, text-based AI interventions can be introduced in settings where drug exposure already takes place. Hence, this review also proposes a three-tier implementation roadmap based on Pakistan’s local context and evidence from other LMICs. The roadmap aims to align the growing use of these technologies with the regulatory and health-system constraints within which they must be implemented.