Artificial intelligence-powered pharmacovigilance: A review of machine and deep learning in clinical text-based adverse drug event detection for benchmark datasets

Li, Yiming; Tao, Wei; Li, Zehan; Sun, Zenan; Li, Fang; Fenton, Susan; Xu, Hua; Tao, Cui

doi:10.1016/j.jbi.2024.104621

Cited by 13 publications

(5 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It is evident that the model exhibits a predominant challenge in dealing with boundary mismatch, false positives and false negatives, which can be attributed to several factors. The quality and representativeness of the training data play a significant role; inconsistent or limited annotations can lead to mismatches and incorrect identifications [ 60 , 61 ]. The inherent complexity of distinguishing similar and potentially overlapping entities adds to the challenge.…”

Section: Discussionmentioning

confidence: 99%

AE-GPT: Using Large Language Models to extract adverse events from surveillance reports-A use case with influenza vaccine adverse events

Li,

et al. 2024

PLoS ONE

Self Cite

View full text Add to dashboard Cite

Though Vaccines are instrumental in global health, mitigating infectious diseases and pandemic outbreaks, they can occasionally lead to adverse events (AEs). Recently, Large Language Models (LLMs) have shown promise in effectively identifying and cataloging AEs within clinical reports. Utilizing data from the Vaccine Adverse Event Reporting System (VAERS) from 1990 to 2016, this study particularly focuses on AEs to evaluate LLMs’ capability for AE extraction. A variety of prevalent LLMs, including GPT-2, GPT-3 variants, GPT-4, and Llama2, were evaluated using Influenza vaccine as a use case. The fine-tuned GPT 3.5 model (AE-GPT) stood out with a 0.704 averaged micro F1 score for strict match and 0.816 for relaxed match. The encouraging performance of the AE-GPT underscores LLMs’ potential in processing medical data, indicating a significant stride towards advanced AE detection, thus presumably generalizable to other AE extraction tasks.

show abstract

Section: Discussionmentioning

confidence: 99%

AE-GPT: Using Large Language Models to extract adverse events from surveillance reports-A use case with influenza vaccine adverse events

Li,

et al. 2024

PLoS ONE

Self Cite

View full text Add to dashboard Cite

show abstract

“…In addition, a recent and thorough review focusing on AI-driven PV has examined the capabilities of machine learning (ML) and deep learning (DL) techniques in detecting ADRs [ 44 ]. This analysis highlights the significant potential that AI holds in the field of PV, detailing both the strengths and challenges of the existing approaches and proposing directions for future research aimed at improving the extraction of ADRs from varied data sources [ 44 ]. The implications of this work for drug safety monitoring and healthcare outcomes are profound, signaling a move towards leveraging sophisticated computational techniques to advance the field of PV.…”

Section: Discussionmentioning

confidence: 99%

Pharmacovigilance Practices by Healthcare Providers in Oncology: A Cross-Sectional Study

Alkofide,

Almalag,

Alromaih

et al. 2024

Pharmaceuticals

View full text Add to dashboard Cite

Investigating pharmacovigilance (PV) practices among oncology healthcare providers (HCPs) is crucial for patient safety in oncology settings. This study aimed to assess the awareness, attitudes, and practices towards PV and identify barriers to effective adverse drug reaction (ADR) reporting for HCPs working in oncology-related settings. Employing a cross-sectional survey design, we collected data from 65 HCPs, focusing on their experiences with ADR reporting, education on ADR management, and familiarity with PV protocols. The results showed that about half of the responders were pharmacists. Around 58.9% of the respondents reported ADRs internally, and 76.9% had received some form of ADR-related education. However, only 38.5% were aware of formal ADR review procedures. Methotrexate and paclitaxel emerged as the drugs most frequently associated with ADRs. The complexity of cancer treatments was among the common reasons for the low reporting of ADRs by the study participants. The findings highlight the need for enhanced PV education and standardized reporting mechanisms to improve oncology care. We conclude that reinforcing PV training and streamlining ADR-reporting processes are critical to optimizing patient outcomes and safety in oncology, advocating for targeted educational interventions and the development of unified PV guidelines.

show abstract

“…Additionally, spatial analysis examines the similarity of AEFI across different regions, providing insights into spatial variations, vaccine brands, and populations [40]. The majority of AEFI are preventable [41] [43]. The method accounted for underreporting and zero-inflation in passive surveillance systems and identified 14 AEFI with significantly heterogeneous reporting rates over the years, including two events showing an increasing trend [43].…”

Section: Introductionmentioning

confidence: 99%

“…Additionally, spatial analysis examines the similarity of AEFIs across different regions, providing insights into spatial variations, vaccine brands, and populations [ 40 ]. The majority of AEFIs are preventable [ 41 ]; thus, analyzing these AEFIs reported enables public health researchers and officials to understand their spatial patterns, potential causal factors, and overall impact, supporting evidence-based decision-making and targeted interventions. The Vaccine Adverse Event Reporting System (VAERS), a comprehensive database that collects reports of AEFIs across different states and periods in the United States, proves instrumental in conducting temporal and spatial monitoring of COVID-19 vaccine–related AEFIs [ 42 ].…”

Section: Introductionmentioning

confidence: 99%

Adverse Events of COVID-19 Vaccines in the United States: Temporal and Spatial Analysis

Li,

Dang

et al. 2024

JMIR Public Health Surveill

Self Cite

View full text Add to dashboard Cite

Background: COVID-19, a highly contagious respiratory disease, has rapidly emerged as a global pandemic. By 3 March 2024, 103.4 million confirmed cases of COVID-19 had been reported in the United States. Vaccination against COVID-19 is currently considered one of the most effective strategies to control its spread. However, due to variations in populations and the ongoing development of vaccines, it is expected that a certain number of individuals may experience adverse events following immunization (AEFI) for COVID-19 vaccines.Objective: In this study, we exclusively examined AEFI associated with COVID-19 vaccination reported by vaccine recipients. Our objective was to analyze the temporal and spatial characteristics of these reported AEFI to gain insights into their patterns and distribution. Methods:We analyzed the administration data from the Centers for Disease Control and Prevention (CDC) (N = 663,822,575) and the reports from the surveillance system-Vaccine Adverse Event Reporting System (VAERS) (N = 900,522) between 2020 and 2022. To gain a broader understanding of reported post-vaccination AEFI, we categorized them into System Organ Classes (SOCs) according to the Medical Dictionary for Regulatory Activities (MedDRA). Additionally, temporal analysis was conducted to examine the trends of AEFI in all VAERS reports, those related to Pfizer and Moderna as well as the Top 10 AEFI trends in serious reports. We also compared the similarity of symptoms across various regions within the United States.Results: Our findings revealed that the most frequently reported symptoms following COVID-19 vaccination were headache (N = 141,186), pyrexia (N = 122,120), and fatigue (N = 121,910). The most common symptom combination was chills and pyrexia (N = 56,954). Initially, General disorders and administration site conditions (SOC 22) were the most prevalent class reported. Moderna exhibited a higher reporting rate of AEFI compared to Pfizer. Over time, we observed a decreasing reporting rate for COVID-19-related AEFI. And the overall rates of AEFI between the Pfizer and Moderna vaccines were comparable. In terms of spatial analysis, the middle and north regions displayed a higher reporting rate of AEFI for COVID-19 vaccines, while the southeast and south central regions showed notable similarity in reported COVID-19 symptoms. Conclusions:This study presents a potentially beneficial method for AEFI surveillance, especially for severe cases, which can intensify research efforts in areas where abnormal or severe presentations occur. The findings suggest that increasing vaccination coverage reduces the reported AEFI, thereby enhancing vaccine confidence. Our study highlights the importance of robust health policies for managing AEFI in future vaccination campaigns. Specifically, health authorities should consider implementing targeted monitoring programs to detect and manage AEFI early, especially for symptoms that were severe and frequently reported in our analysis. These programs can help improve public confidence in COVID-...

show abstract

Artificial intelligence-powered pharmacovigilance: A review of machine and deep learning in clinical text-based adverse drug event detection for benchmark datasets

Cited by 13 publications

References 29 publications

AE-GPT: Using Large Language Models to extract adverse events from surveillance reports-A use case with influenza vaccine adverse events

AE-GPT: Using Large Language Models to extract adverse events from surveillance reports-A use case with influenza vaccine adverse events

Pharmacovigilance Practices by Healthcare Providers in Oncology: A Cross-Sectional Study

Adverse Events of COVID-19 Vaccines in the United States: Temporal and Spatial Analysis

Contact Info

Product

Resources

About