A carregar...
Publicação
Impact of social media data on postmarket safety evaluation of medicines: a literature review on automatic mining initiatives of adverse drug reactions
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.35 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Introduction/Background: The impact of adverse drug reactions (ADRs) on public health and national
healthcare systems is substantial. The current pharmacovigilance method is time-consuming,
incomplete and prone to data loss. Also, due to characteristics inherent to the reporting process, a
great portion of ADRs are never reported. Social media (SM) data, due to its volume and immediacy,
shows promise for a patient centered way of reporting, and has received increasing attention over
the last few years.
Objectives/Methodology: In this research project the author proposes to evaluate how can ADR
automatic detection from social media contribute for pharmacovigilance, through a systematic
literature review. The review included articles published over the last five years, accounting to 33
publications that were retrieved and reviewed in detail.
Discussion: Several aspects have proven to be critical when developing SM based ADR mining - the
main purpose of the analysis (detection of posts containing ADRs and the extraction of specific ADRdrug
pairs), the approach (lexicon or machine learning based), and the type of platform used (healthfocused
or general use). The studies have shown a prevalence of machine learning (ML) based
approaches, from which supervised learning is the most popular method, despite the rising trend
against the need for costly and time-consuming annotation of data. Mixed approaches have often
been used as they seem to derive better performance, whether in combining data sources from
general platforms and disease forums, or using distinct sources of annotated data sets, such as
biomedical corpus to increase algorithms strength, and even the combination of ML approaches with
lexicon based features.
Conclusions/Limitations: The end goal of ADR mining from social media is to be able to identify drugs
that are either frequently related to ADRs, or those that are associated with previously unknown
ADRs. Combining data from multiple sources will contribute to prevent the impact of serious or
previously unknown ADRs, focusing on the issues most pertinent to patients, and will provide a
broader safety profile of any medication, with benefits for patients, health systems, companies and
regulatory agencies. SM data comes with its specificities (informal language, semantic confusion and
ambiguity) that lead to analysis hurdles; hence the method and approach used must be adapted to
the purpose of investigation and resources available. Norms and agreed practices to guide these
efforts are needed, considering ethical issues, data quality and governance. The progress in
information technology and the need to consider patient experience should motivate future research
on social media surveillance for complementing conventional pharmacovigilance with patient centric
and real time ways of reporting.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business Intelligence
Palavras-chave
Social Media Real World Data Pharmacovigilance Medicines Machine Learning Adverse drug reactions