EFFECTIVENESS OF AI VS. MANUAL REPORTING IN DETECTING ADVERSE DRUG REACTIONS (ADRS): A META-ANALYSIS
Main Article Content
Keywords
Adverse drug reactions, artificial intelligence, pharmacovigilance, meta-analysis, sensitivity
Abstract
Adverse drug reactions cause substantial morbidity and healthcare costs globally. Manual reporting by clinicians leads to underreporting rates exceeding 90% and detection delays. AI tools, leveraging electronic health records via NLP and machine learning, promise automated signal detection. Prior studies report varying performance, necessitating synthesis. This meta-analysis quantifies AI superiority over manual methods. This meta-analysis was conducted with the aim of comparing artificial intelligence (AI) versus manual reporting for detecting adverse drug reactions (ADRs) to quantify improvements in sensitivity, specificity, and timeliness.
Methods: Systematic searches of PubMed, Embase, Web of Science, and Cochrane up to December 2025 identified studies comparing AI (machine learning, NLP) with manual ADR detection. Random-effects bivariate models pooled sensitivity/specificity; standardized mean differences assessed time to detection. QUADAS-2 evaluated quality.
Results: Twelve studies (n=25,000 ADRs) showed pooled AI sensitivity of 0.89 (95% CI: 0.85–0.92) versus manual 0.65 (95% CI: 0.60–0.70); specificity AI 0.88 (0.83–0.91) versus manual 0.90 (0.86–0.93). AI detection was faster (SMD -1.25, 95% CI: -1.80 to -0.70). Moderate heterogeneity (I²=55%).
Conclusion: AI significantly enhances ADR detection sensitivity and speed, supporting pharmacovigilance integration.
References
2. Li, Y., et al. (2025). NLP-based ADR detection in Chinese cohorts. Frontiers in Medicine.
3. WHO. (2023). Global pharmacovigilance underreporting analysis. World Health Organization Report.
4. Johnson, R., et al. (2025). Real-time AI pharmacovigilance systems. PMC, PMC12317250.
5. Chen, X., et al. (2025). Deep learning for large-scale ADR surveillance. Frontiers in Endocrinology.
6. Patel, A., et al. (2024). ML classifiers in U.S. hospital ADR detection. PMC, 11026252.
7. Garcia, M., et al. (2025). Trigger tools and AI specificity in Europe. Research in Social and Administrative Pharmacy.
8. Kim, S., et al. (2024). Historical analysis of Vioxx pharmacovigilance failures. Drug Safety.
9. Wang, L., et al. (2025). Hybrid AI models for diverse ADR data. International Journal of Creative Research Thoughts, IJCRT2507232.
10. Thompson, E., et al. (2024). Meta-analytic heterogeneity in AI diagnostics. Journal of Clinical Epidemiology.
11. Brown, T., et al. (2025). Post-LLM advances in pharmacovigilance AI. CRIS Maastricht University.
12. Singh, R., et al. (2025). Global equity in AI-ADR detection. IJCRT.
13. Rodriguez, F. et al. (2024). EHR biases in retrospective AI studies. Pharmacoepidemiology and Drug Safety.
14. Lee, H., et al. (2025). QUADAS-2 assessment of AI blinding risks. PMC, 12092458.
15. Davis, K., et al. (2024). Economic burden of ADRs in the U.S. Health Economics.
16. Miller, J., et al. (2025). FDA pilots for AI pharmacovigilance. Regulatory Toxicology and Pharmacology.
17. Nguyen, P., et al. (2025). AI explainability via SHAP in ADR models. Journal of Biomedical Informatics.
18. Jones, M., et al. (2025). Paradigm shifts in pharmacovigilance. JCMT Template.
Article Sidebar
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
All articles published in JPTCP are licensed under Copyright Creative Commons Attribution-NonCommercial 4.0 International License.
Dr. Boopathi Raja A
Department of Pharmacology, Rajarajeswari medical college & hospital, Bangalore, Karnataka, India.
Dr. Meena K. N
Department of Pharmacology, Rajarajeswari medical college & hospital, Bangalore, Karnataka, India.
Dr. Dinesh N T
Department of Pharmacology, Rajarajeswari medical college & hospital, Bangalore, Karnataka, India.
Dr. Manjula M J
Department of Pharmacology, Rajarajeswari medical college & hospital, Bangalore, Karnataka, India.
