An Epistemological Approach to Risk Assessment in

Pharmacovigilance and Mitigation Through Artificial Intelligence

A. D. Skali

Healthcare Innovation Consultant, CEO TheTowerBrand, ex-COO Luci Health, Head of Innovation Forum Barcelona

Branch, Future Business Centre King's Hedges Road, Cambridge, U.K.

Keywords: Pharmacovigilance, Risk Assessment, Drug Safety, Risk Management, Investment Practices, Bibliographic

Research, Benefit-Risk Evaluation, Artificial Intelligence, Epistemology.

Abstract: Pharmacovigilance, which focuses on risk assessment and management in drug safety, offers a robust

foundation for addressing inherent risks in new drug discovery. This bibliographic research article explores

innovative perspectives by drawing parallels between pharmacovigilance and investment practices, as

inspiration to establish a new, in the field of pharmacovigilance, epistemological framework for the

understanding of the main risk inducing elements in pharmacovigilance and the steps and technology we can

adopt to assess and mitigate them.

1 BACKGROUND AND

CONTEXT

Risk assessment and prevention is essential in the

field of healthcare. The judicious application of risk

assessment methodologies serves as a sentinel,

discerning latent hazards embedded within clinical

processes. This discernment facilitates the creation of

a secure healthcare milieu, diminishing the incidence

of adverse events and fortifying the sanctity of patient

well-being.

As we transition to the specific domain of

pharmacovigilance, the importance of risk

management becomes even more pronounced. In the

pharmaceutical landscape, where the stakes are

inherently high, risk assessment plays a pivotal role

in ensuring the safety and well-being of patients. The

intricate web of risks in this context includes not only

the potential side effects of medications but also

regulatory compliance and the complexities of a

globally interconnected pharmaceutical market.

Pharmacovigilance, as a subset of risk

management in healthcare, involves the systematic

monitoring and evaluation of the safety and efficacy

of pharmaceutical products post-market approval.

The application of rigorous risk assessment

frameworks within pharmacovigilance becomes a

linchpin for identifying potential risks associated with

medication use. This includes adverse drug reactions,

unexpected side effects, and any other safety

concerns that may arise during the course of patient

treatment.

In the field of pharmacovigilance, the challenges

and limitations of current risk assessment

methodologies are multifaceted and underscore the

evolving nature of risks associated with

pharmaceutical products.

One prominent challenge lies in the dynamic

nature of risks. Traditional risk assessment

methodologies often struggle to keep pace with the

rapidly changing landscape of pharmaceuticals,

where new drugs are continually introduced, and their

effects may only become apparent after widespread

use. The inherent complexity of biological systems

and the variability in patient responses contribute to

the dynamic nature of risks, necessitating a more

adaptive and responsive approach to risk assessment.

Moreover, the inadequacy of traditional

approaches becomes apparent when addressing

emerging threats. Conventional risk assessment

models may not adequately account for novel and

unforeseen risks that can emerge as a result of

evolving scientific knowledge, changes in patient

demographics, or the introduction of innovative

therapeutic modalities. These emerging threats may

include previously unknown side effects, drug

interactions, or unexpected patient populations

susceptible to adverse reactions.

Additionally, the global interconnectedness of the

Skali, A.

An Epistemological Approach to Risk Assessment in Pharmacovigilance and Mitigation Through Artiﬁcial Intelligence.

DOI: 10.5220/0012869800003854

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st Inter national e-Health Forum (IeHF 2023), pages 29-36

ISBN: 978-989-758-711-5

pharmaceutical market poses challenges to traditional

risk assessment methodologies. The widespread

distribution of pharmaceutical products across

diverse populations and regulatory environments

requires a more comprehensive and globally aligned

approach to risk assessment. Traditional models may

struggle to capture the nuanced variations in risk

profiles across different regions and demographic

groups, potentially leading to incomplete risk

assessments.

The reliance on spontaneous reporting systems for

adverse drug reactions is another limitation. Such

systems heavily depend on healthcare professionals

and patients voluntarily reporting adverse events,

leading to underreporting and a potential lag in

identifying risks. This limitation hampers the real-

time assessment of risks associated with

pharmaceutical products.

In the context of pharmacovigilance, the existing

literature on risk assessment and mitigation provides

valuable insights into the challenges and

advancements in ensuring drug safety. However,

there are noticeable gaps and areas where

epistemological perspectives are underexplored.

The current literature predominantly focuses on

the technical and methodological aspects of risk

assessment, such as signal detection, data mining, and

statistical modeling. While these approaches are

crucial, there is a paucity of literature delving into the

underlying epistemological foundations that shape

our understanding of risk in pharmacovigilance.

One evident gap lies in the exploration of the

ontological and epistemological assumptions inherent

in risk assessment methodologies. Understanding the

nature of knowledge and reality as it pertains to drug

safety is crucial for refining risk assessment models.

For instance, the ontological status of adverse events,

whether they are discrete entities or part of a complex

network of interconnected factors, remains a topic

that warrants deeper philosophical exploration.

Furthermore, there is limited literature on the

epistemic uncertainties associated with

pharmacovigilance data. Epistemological

perspectives can shed light on the inherent

uncertainties in observational data, the reliability of

different sources, and the interpretative challenges in

discerning causality. Addressing these epistemic

uncertainties is pivotal for improving the accuracy

and reliability of risk assessments.

The potential contributions of integrating

epistemology into risk management practices are

substantial. Epistemological perspectives can inform

the development of more robust risk models by

providing a foundation for understanding what counts

as evidence, how causality is established, and the

nature of knowledge production in pharmacovigilance.

This integration can enhance the transparency and

accountability of risk assessment processes, as it

encourages a critical examination of the assumptions

and values that underpin decision-making.

Moreover, incorporating epistemological

considerations can foster interdisciplinary

collaboration between experts in pharmacovigilance,

philosophy, and other relevant fields. This

collaboration can lead to a more comprehensive and

holistic approach to risk assessment, considering not

only the technical aspects but also the epistemological

underpinnings that shape our understanding of drug

safety.

In the context of pharmacovigilance, the

integration of effective risk management strategies

aligns with the broader goal of fostering a culture of

patient safety. By systematically identifying,

assessing, and mitigating risks associated with

pharmaceutical products, the healthcare industry can

uphold the highest standards of patient care and well-

being. This interconnected approach underscores the

symbiotic relationship between robust risk

management practices, patient safety, and the

integrity of the pharmaceutical industry.

This paper embarks on the ambitious journey of

unraveling the intricate relationship between

epistemology and risk assessment in

pharmacovigilance. Its overarching goal is to

contribute a nuanced understanding that enriches the

current discourse on drug safety by delving into the

philosophical underpinnings and epistemic

dimensions inherent in the field. The primary focus of

this exploration is on meticulously examining the

epistemological facets of risk assessment in

pharmacovigilance.

This endeavor involves unraveling the intricacies

of how information regarding drug safety is

perceived, interpreted, and validated. Within this

focus, particular emphasis will be placed on

elucidating both the ontological essence of adverse

events and the epistemic processes that govern

knowledge production in the realm of

pharmacovigilance.

The scope extends beyond the superficial layers,

aiming for an in-depth examination of the

philosophical foundations that shape current risk

assessment methodologies. This includes a critical

analysis of how philosophical perspectives influence

our conceptualization of adverse events. Within this

extended examination, the exploration encompasses a

comprehensive analysis of the nature of knowledge

production in pharmacovigilance. This sheds light on

IeHF 2023 - International e-Health Forum

how epistemological assumptions contribute to the

construction of narratives surrounding drug safety.

The paper actively advocates for interdisciplinary

collaboration between experts in pharmacovigilance,

philosophy, and related fields. While promoting such

collaboration, it acknowledges inherent limitations in

providing exhaustive analyses of the technical

intricacies within the pharmacovigilance domain.

The proposed framework for integrating

epistemology into risk management practices will be

thoroughly discussed within the specific context of

pharmacovigilance. This includes practical insights

into how epistemological considerations can enhance

transparency, accountability, and accuracy in the

assessment of drug safety. By navigating through

these interconnected realms, the paper aims not only

to shed light on the philosophical dimensions of risk

assessment in pharmacovigilance but also to advocate

for a collaborative and informed approach towards

ensuring drug safety.

2 A SIMPLIFIED FRAMEWORK

FOR DECISION MAKING

In contemplating the seemingly disparate realms of

investing and pharmacovigilance, one might readily

dismiss any potential correlation based on their

ostensible divergent objectives. However, upon

closer examination of their core essence and

methodological approaches, an unexpected similarity

emerges.

Pharmacovigilance, in its essence, constitutes the

scientific discipline and set of actions dedicated to the

vigilant monitoring of medicine safety and the

proactive management of any issues that may arise in

this domain. The World Health Organization

encapsulates this concept as encompassing activities

directed at detecting, assessing, understanding, and

preventing adverse reactions to medicines and other

medicine-related problems. In essence,

pharmacovigilance is a continuous process of

monitoring medicine safety, with the overarching aim

of reducing risks and optimizing benefits.

Conversely, the foundational objective of

investing is rooted in the strategic utilization of

available information and analytical tools to make

judicious decisions. These decisions are not solely

oriented towards maximizing profit; they necessitate

a comprehensive evaluation of associated risks and

the implementation of strategic measures to mitigate

these risks effectively.

While the goals of investing and

pharmacovigilance may appear divergent at first

glance, a closer examination reveals a shared pursuit

– the reduction of risks and the enhancement of

positive outcomes. Both disciplines, albeit operating

in distinct domains, converge on the fundamental

principle of informed decision-making to achieve

outcomes that are not only focused on obtaining

benefits but also resilient in the face of inherent

uncertainties.

In the realm of pharmacovigilance, a succinct

framework guiding the evaluation of whether a drug

should remain on the market involves several key

steps:

Data Collection and Reporting Signal Detection

Risk Assessment Benefit-Risk Evaluation

Risk Minimization Strategies

What unites these procedural steps is a

fundamental principle shared with diverse domains

such as investing, marketing, and management.

Despite apparent differences in decisions, they

collectively adhere to a basic yet robust structure—

the problem-solving structure—which forms the

bedrock of the scientific method.

At the core of these decision-making processes

lies a simple and universal structure, akin to the

scientific method. When embarking on drug usage,

individuals are essentially testing a basic hypothesis:

"Is this drug sufficiently beneficial to justify potential

associated risks?"

"Do we possess comprehensive information to ensure

the accuracy of our decisions?"

Figure 1: Basic framework for the scientific method.

In its essence, this structured decision-making

process maintains a constant framework, its iterations

adapting to our evolving comprehension of the

situation or the issues at hand. This universal

approach extends beyond specific domains,

encompassing diverse fields such as marketing,

investing, management, sports, and, as asserted in this

context, pharmacovigilance. At the core of effective

problem-solving lies the validation or affirmation of

a hypothesis grounded in experimental data (Figure

1). This involves exploring correlations or causations

among various elements to make informed

predictions and actively working to minimize the

An Epistemological Approach to Risk Assessment in Pharmacovigilance and Mitigation Through Artiﬁcial Intelligence

likelihood of recurring errors.

The debate surrounding the applicability of this

streamlined decision-making model to

pharmacovigilance may require further discusiont.

However, the valuable perspective gained by

examining the field through this lens provides clarity,

fostering a nuanced understanding of its intricacies.

3 THE FRAMEWORK AND

PHARMACOVIGILANCE

In the context of pharmacovigilance and the

delineated steps, applying this straightforward model

could be represented as follows:

DAE, AE, or SAE → Information Gathering →

Judgment

Where Adverse Drug Effects (ADEs) constitute a

comprehensive category encompassing any harmful

or unintended effect resulting from drug usage,

Adverse Drug Reactions (ADRs) constitute a subset

of ADEs, specifically referring to unwanted and

harmful effects caused by a medication when taken at

normal doses during the regular course of treatment,

and Spontaneous Adverse Effects (SAE) denote

unintended, harmful reactions to a drug occurring

without any apparent cause or known pattern.

And the element that we always have to take into

account when approaching risk assessment is the

distinction arises between the "real risk" and the

"assumed risk." During the process of risk

assessment, there's a common inclination to believe

that the available information is comprehensive

enough to ensure accurate decision-making.

A compelling analogy that encapsulates this idea

is the iceberg metaphor. What's visible above the

waterline represents only a fraction of the entire

structure, with a substantial portion hidden beneath

the surface. This concept is reminiscent of the

challenges faced by ships navigating near glaciers.

Initially, ships assumed that wooden hulls were

sufficient to navigate these icy terrains, leading to

numerous sinkings. The realization that the actual

risk, the unknown factors, outweighed the assumed

risk prompted strategic measures. Ships began

reinforcing their hulls with metal, establishing

specific routes and timings to navigate through

glaciers. While they couldn't precisely determine the

size of each glacier, they made educated guesses

about potential risks, enabling them to accomplish

their goals despite incomplete knowledge.

Three essential elements emerge for informed

decision-making across various domains:

understanding the known, which involves delving into

the philosophy of knowledge (epistemology);

identifying main risks by discerning primary risks

based on existing knowledge and formulating initial

technological solutions; and establishing a feedback

loop, creating a continuous learning mechanism for

iterative improvements over time, fostering

adaptability and enhanced decision-making

capabilities.

3.1 Reducing the Gap Between Real

and Perceived Risk

All risk reduction strategies are rooted in two

fundamental principles: augmenting our

understanding of causal/correlation

relationships and mitigating errors associated with

human judgment. The empirical evidence from the

notable investor Ray Dalio and his hedge fund,

Bridgewater Associates, underscores the

transformative impact of enhancing these two

principles on decision-making. The initial imperative

is to amass more pertinent information while ensuring

its accuracy and relevance. In the realm of

pharmacovigilance, diverse avenues exist to gather

information on potential adverse effects, adverse drug

effects, and spontaneous adverse effects.

Figure 2: Application of the epistemological framework to

pharmacovigilance.

However, the underlying questions remain

consistent. Pertinent inquiries include assessing the

severity, frequency, reversibility, and likelihood of

potential adverse drug reactions (ADRs). These

questions bear substantial weight, influencing critical

IeHF 2023 - International e-Health Forum

decisions such as whether to proceed with drug

commercialization or delay it. The pivotal factor lies

in the reliability and trustworthiness of the amassed

information. Consequently, a refined framework for

pharmacovigilance can be articulated (Figure 2).

In addressing these inquiries, additional

considerations come to the forefront:

Determining the specific information essential for

making informed decisions is a foundational step.

Evaluating the reliability of the information received

is paramount, questioning its accuracy and relevance.

Scrutinizing the trustworthiness of the

information source, whether it be a program or an

individual, becomes crucial. Assessing the

believability of the person or the reliability of the tool

involves questioning when it was last validated and

calibrated in a similar context. Assessing the

credibility of the person providing information and

the trustworthiness of the tools used to obtain

information is imperative. Interrogating the methods

employed and the tools utilized in information

acquisition is essential for ensuring their reliability.

Gauging the truthfulness of the information and

its compatibility with a usable format forms a critical

component of the evaluation.

The significance and prioritization of these

questions vary depending on the context. The

applicability of these inquiries differs substantially,

whether in the domain of pharmacovigilance during

the clinical trial phase or the post-marketing phase.

Due to constraints of brevity, this discussion

primarily aims to provide an overarching perspective

on these concepts and elucidate how digital tools can

contribute to enhancing the safety of our endeavors.

4 MAIN CAUSES FOR ERRORS

IN THE DATA COLLECTION

AND REPORTING STAGE

During the data collection phase, various challenges

can significantly impact the reliability and

comprehensiveness of gathered information:

Underreporting is a prominent issue involving the

failure of healthcare professionals, patients, and

pharmaceutical companies to report adverse events

related to medications. This leads to incomplete

safety profiles and hinders the identification of risks.

Conversely, overreporting can occur, attributing

unrelated events to medication use, requiring

meticulous sorting to discern genuine concerns.

Factors such as ignorance, lethargy, complacency,

diffidence, insecurity, and the absence of feedback

contribute to underreporting.

Delays in reporting adverse events pose a serious

challenge, hindering the ability to take timely

corrective actions and assess the overall impact on

patient safety. Inconsistent and non-standardized

reporting practices across healthcare institutions,

regions, or countries complicate data collection and

analysis, impeding the identification of trends and

patterns in adverse events.

Patient selection bias arises when participants in

pharmacovigilance studies are not representative of

the general population taking the drug, potentially

skewing results. For instance, studies may include

only patients with specific medical conditions or

those taking the drug at high doses.

Poor data quality, including inaccuracies,

duplications, or missing data, undermines the

reliability of pharmacovigilance databases, leading to

erroneous conclusions. Incomplete information and

confounding factors, such as multiple medications or

underlying health conditions, complicate data

interpretation.

Effective communication between regulatory

agencies, pharmaceutical companies,

pharmacovigilance teams, healthcare providers, and

patients is crucial. Gaps in communication hinder the

timely exchange of safety information and

collaborative efforts to mitigate risks. Addressing

these challenges is essential for enhancing the

accuracy and utility of pharmacovigilance data.

4.1 Main Causes for Error in the Case

of Signal Detection in

Pharmacovigilance

Pharmacovigilance systems encounter various

challenges that can impede their effectiveness. One

crucial aspect is sensitivity, representing the proportion

of actual adverse drug reactions (ADRs) reported to the

system. A low sensitivity raises concerns, indicating

that a considerable number of ADRs may go

unreported. This limitation compromises the system's

ability to comprehensively capture and address

potential risks associated with medications.

Another significant challenge is the Low Positive

Predictive Value (PPV), which denotes the proportion

of reported ADRs truly caused by the drug. A low

PPV introduces noise and potential confusion into the

pharmacovigilance system, as a significant portion of

reported ADRs may not be directly attributable to the

drug in question.

The difficulty in detecting rare ADRs is also a

noteworthy challenge. Due to their infrequent

occurrence in a limited population, rare ADRs are

An Epistemological Approach to Risk Assessment in Pharmacovigilance and Mitigation Through Artiﬁcial Intelligence

often underreported, making it challenging for the

pharmacovigilance system to identify and address

these less common but potentially severe adverse

events.

Long-term ADRs pose a distinct challenge, as

their manifestations may be delayed for months or

even years. This delayed onset makes it difficult to

associate these ADRs with the medication, leading to

underreporting and hindering the timely

identification of such adverse events.

Cultural influences play a significant role in ADR

reporting practices. In Japan, a cultural norm

discourages complaining, extending to the reporting

of ADRs. This cultural inclination results in

reluctance among both patients and doctors to report

ADRs, contributing to underreporting in the country.

Similarly, traditional Chinese medicine reflects a

belief that side effects are inherent and necessary for

the medicine to be effective. This belief may

discourage patients from reporting ADRs, viewing

them as integral to the healing process. Cultural

variations also impact the types of ADRs reported,

with a preference for reporting skin-related ADRs in

Asian countries compared to liver-related ADRs in

Western countries.

These challenges underscore the intricate nature

of pharmacovigilance and emphasize the need for

nuanced strategies to effectively address them. A

comprehensive approach that considers cultural

factors, enhances sensitivity and PPV, and tackles the

difficulties in detecting rare and long-term ADRs is

essential for ensuring the robustness of

pharmacovigilance systems.

5 NEW TECHNOLOGY AND AI

TO HELP GET REDUCE THE

BIASES AND ERRORS

In addressing the challenges of measuring real risk in

decision-making, strategies include specific training

and hiring experts in risk assessment and

epistemology. This discussion focuses on using

digital solutions, emphasizing AI/ML's role in data

ingestion, including duplicate detection and anomaly

identification. Machine learning aids in detecting

ADRs, performing safety surveillance, and managing

signal detection, such as automating the classification

of first-person reports of ADRs in social media. It

offers advantages in detecting ADRs not captured by

medical professionals, processing data quickly, and

utilizing personal information in social media posts

related to ADRs.

Machine learning is also employed to classify

ADRs, determining the seriousness of patient cases

through different algorithms based on precision,

recall, and accuracy. Clinical trials, crucial for drug

approval, face structural limitations, and post-

marketing monitoring through AE reports in

pharmacovigilance is not error-proof due to biases

like underreporting, especially for rare events and

drug-drug interactions. Machine learning aids in

streamlining adverse event reports, comparing rule-

based queries and semi-supervised machine learning

against a reference standard.

In pharmacoepidemiology, ML predicts adverse

events, facilitating early quality assurance measures.

Its use in signal detection and analysis automates

processes, adapting to patients presenting with

multiple disease states, medications, and ADRs.

Institutions like Connecticut Children’s Medical

Center leverage machine learning to streamline

adverse event reports. Additionally, ML in

pharmacoepidemiology studies drug interactions in

real-life conditions, predicting adverse events

promptly for patient safety.

The utility of machine learning is underscored in

its application to screen and analyze voluminous

datasets of adverse event reports through

sophisticated algorithms and text mining. Specific

implementations include the development of

algorithms like "AwareDX," exhibiting the capacity

to predict sex-specific risks of adverse drug effects

with a remarkable degree of precision, and the

identification of targeted patient populations

vulnerable to specific toxicities. Machine learning

further aids in predicting drug side effects during

post-marketing surveillance, leveraging knowledge

extracted from literature to enhance the efficacy of

spontaneous reporting system methods.

Artificial intelligence makes significant strides in

integrating prediction uncertainties into patient safety

through the deployment of deep learning-based

computer-aided diagnosis, yielding more dependable

results in cases fraught with ambiguity. However, the

seamless integration of AI into existing

pharmacovigilance systems raises potential

challenges, potentially amplifying workload and

complexity. The judicious implementation of AI/ML

in PV is recommended, specifically when it

streamlines workload, simplifies complexity, or

optimizes budget allocation, enabling more effective

resource deployment for critical aspects ensuring

patient safety.

The integration of AI/ML into pharmacovigilance

encounters legal challenges in both Europe and the

United States, particularly concerning the liability for

IeHF 2023 - International e-Health Forum

errors arising from artificial intelligence technology.

Despite these legal impediments, the promising

potential of AI/ML in pharmacovigilance remains

evident, prompting a critical examination of how to

harness these technologies effectively to construct a

future fit for purpose. Establishing a seamlessly

connected system for the flow of inputs and outputs

across diverse data systems emerges as a critical

imperative. Such a system would not only foster an

interactive continual learning solution but also

enhance the understanding of the benefit–risk profiles

of medicines and vaccines. Additionally, it would

empower prescribers, patients, and other stakeholders

to obtain pertinent information and pose inquiries as

needed, thereby contributing to a more informed and

responsive healthcare ecosystem.

REFERENCES

Caster, O. (2018). Benefit-Risk Assessment in

Pharmacovigilance. In A. Bate (Ed.), Evidence-Based

Pharmacovigilance (Methods in Pharmacology and

Toxicology). Humana Press.

https://doi.org/10.1007/978-1-4939-8818-1_13

Coleman, T. S. (2012). Quantitative Risk Management: A

Practical Guide to Financial Risk, (B. Litterman,

Foreword).

Deng, L., & Chang, Y. (2022). Risk Management of

Investment Projects Based on Artificial Neural

Network. Wireless Communications and Mobile

Computing, 2022, Article ID 5606316, 13 pages.

https://doi.org/10.1155/2022/5606316

Huysentruyt, K., Kjoersvik, O., Dobracki, P., Savage, E.,

Mishalov, E., Cherry, M., Leonard, E., Taylor, R.,

Patel, B., & Abatemarco, D. (2021). Validating

Intelligent Automation Systems in Pharmacovigilance:

Insights from Good Manufacturing Practices. Drug

Safety, 44(3), 261-272. https://doi.org/10.1007/s40264-

020-01030-2

Pascarella, G., Rossi, M., Montella, E., Capasso, A., De

Feo, G., Botti, G., Nardone, A., Montuori, P., Triassi,

M., D'Auria, S., & Morabito, A. (2021, July 8). Risk

Analysis in Healthcare Organizations: Methodological

Framework and Critical Variables. Risk Management

and Healthcare Policy, 14, 2897-2911.

https://doi.org/10.2147/RMHP.S309098

Hamid, A. A. A., Rahim, R., & Teo, S. P. (2022).

Pharmacovigilance and Its Importance for Primary

Health Care Professionals. Korean Journal of Family

Medicine, 43(5), 290-295.

https://doi.org/10.4082/kjfm.21.0193

Expert Group on Clinical Trials. (2017, April 25). Risk

proportionate approaches in clinical trials:

Recommendations of the expert group on clinical trials

for the implementation of Regulation (EU) No

536/2014 on clinical trials on medicinal products for

human use.

Randeep, V. C., Divya, P., Susmita, A., Sushmitha, P.,

Ramya, Ch., & Chandini, K. (2022, December 12).

Automation in pharmacovigilance: Artificial

intelligence and machine learning for patient safety.

Journal of Innovations in Applied Pharmaceutical

Science (JIAPS), 7(3), 118-122.

https://doi.org/10.37022/jiaps.v7i3.37 Hillson, D.

(2023). The Risk Management Handbook: A Practical

Guide to Managing the Multiple Dimensions of Risk

(6th ed.). Routledge.

Ito, S., & Narukawa, M. (2022). Estimation of the Under-

Reporting of Suspected Serious Adverse Drug

Reactions in Japan Using An Interrupted Time Series

Analysis. Therapeutic Innovation & Regulatory

Science, 56(2), 358–365.

https://doi.org/10.1007/s43441-022-00379-z

García-Abeijon, P., Costa, C., Taracido, M., Herdeiro, M.

T., Torre, C., & Figueiras, A. (2023). Factors

Associated with Underreporting of Adverse Drug

Reactions by Health Care Professionals: A Systematic

Review Update. Drug Safety, 46(7), 625–636.

https://doi.org/10.1007/s40264-023-01302-7

Drukker, K., Chen, W., Gichoya, J., Gruszauskas, N.,

Kalpathy-Cramer, J., Koyejo, S., Myers, K., Sá, R.C.,

Sahiner, B., Whitney, H., Zhang, Z., & Giger, M.

(2023). Toward fairness in artificial intelligence for

medical image analysis: Identification and mitigation of

potential biases in the roadmap from data collection to

model deployment. Journal of Medical Imaging

(Bellingham), 10(6), 061104. doi:

10.1117/1.JMI.10.6.061104.

Sartori, D., Aronson, J. K., & Onakpoya, I. J. (2020).

Signals of adverse drug reactions communicated by

pharmacovigilance stakeholders: Protocol for a scoping

review of the global literature. Systematic Reviews,

9(1), 180. doi: 10.1186/s13643-020-01429-z.

Garcia-Agundez, A., García-Martín, E., & Eickhoff, C.

(2022, May 20). Editorial: The Potential of Machine

Learning in Pharmacogenetics, Pharmacogenomics and

Pharmacoepidemiology. Frontiers in Pharmacology,

13, 928527. https://doi.org/10.3389/fphar.2022.928527

Wewering, S., Pietsch, C., Sumner, M., Markó, K., Lülf-

Averhoff, A. T., & Baehrens, D. (2022, June). Machine

learning approach to identify adverse events in

scientific biomedical literature. Clinical and

Translational Science, 15(6), 1500-1506.

https://doi.org/10.1111/cts.13268

Yang, S., & Kar, S. (2023). Application of artificial

intelligence and machine learning in early detection of

adverse drug reactions (ADRs) and drug-induced

toxicity. Artificial Intelligence Chemistry, 1(2),

100011. https://doi.org/10.1016/j.aichem.2023.100011

Kim, H. R., Sung, M., Park, J. A., Jeong, K., Kim, H. H.,

Lee, S., & Park, Y. R. (2022). Analyzing adverse drug

reaction using statistical and machine learning

methods: A systematic review. Medicine (Baltimore),

101(25), e29387.

https://doi.org/10.1097/MD.0000000000029387

Lee, J. E., Kim, J. H., Bae, J. H., et al. (2022). Detecting

early safety signals of infliximab using machine

An Epistemological Approach to Risk Assessment in Pharmacovigilance and Mitigation Through Artiﬁcial Intelligence

learning algorithms in the Korea adverse event

reporting system. Scientific Reports, 12, 14869.

https://doi.org/10.1038/s41598-022-18522-z

IeHF 2023 - International e-Health Forum