Issue link: https://www.e-digitaleditions.com/i/1494960
14 Pharmaceutical Technology ® Quality and Regulatory Sourcebook March eBook 2023 PharmTech.com Pharmacovigil ance involves ongoing evidence surveillance, bi-monthly search updates, and review updates every six months as needed. This approach enables pharmacovigilance professionals to keep abreast of the latest research, identify potential safety concerns and emerging risks associated with medicinal products, and make in- formed decisions about the potential benefits of using automation technologies for data extraction (1). Auto- mated data extraction can streamline the process of literature monitoring by improving the efficiency of identifying relevant research and reducing the time and effort required for manual screening. Recent advances in computational processing speed and data storage have led to the development of var- ious data mining models and algorithms, which are now being used in the field of pharmacovigilance literature monitoring. These models and algorithms can help pharmacovigilance professionals to quickly identify and extract relevant information from the literature and make more informed decisions about potential safety concerns and emerging risks associ- ated with medicinal products. Utilizing diverse and representative data sources. One approach to addressing data gaps in pharmacovig- ilance is to use diverse and representative data sources. These sources can include electronic health records, patient registries, and social media, in addition to tra- ditional sources such as clinical trial data. However, it is essential that these data sources contain minimum safety information (MSI) and medical assessment for causality. Such information is crucial for building robust and reliable ML models for pharmacovigilance. Having MSI and Medical Assessment for Causality in the data helps to make the ML models more robust and able to work in real-world scenarios for pharmacovigilance; for example, when training ML models to make causality predictions, the causality decision model chosen for an individual case safety report (ICSR) will be applied to all cases (subjectively). If the decision model is not applied correctly, data gaps will result. If the ML model is im- properly trained, it may produce inconsistent results, es- pecially in post-marketing surveillance. This is because it allows the models to identify patterns and trends that may not be apparent in traditional data sources, such as clinical trial data, which is typically collected under controlled conditions. Using diverse and representative data sources that include MSI and Medical Assessment for Causality can help to provide a more complete pic- ture of the real-world population and circumstances in which drugs are used, which can ultimately improve the safety of drugs and protect public health. Employing data augmentation and sensitivity anal- yses techniques. Another strategy for closing data gaps is to employ data augmentation techniques. These techniques involve generating synthetic data that can be used to augment existing data sets, helping to fill in missing or incomplete data. Data augmentation can be particularly useful in cases where it is difficult or impossible to obtain additional real-world data. Sensi- tivity analyses can determine the extent of the impact of the missing data (2). In-stream supervised learning. Another strategy for closing data gaps in ML for pharmacovigilance is using in-stream supervised learning, where data are tagged in real-time as they are being collected. This approach can help ensure that data are captured and known in a timely and accurate manner, helping to fill in gaps that may otherwise arise. In-stream su- pervised learning can be particularly useful in situ- ations where data are collected through ongoing pro- cesses or activities, such as electronic health records or patient registries. By tagging data in real-time, it is possible to ensure that the data used to train ML models is up-to-date and representative of the cur- rent population and circumstances. Practices to enhance the quality and integrity of data Implementing data governance practices, such as data governance frameworks, data catalogues, metadata management systems, data quality management, data lakes, and data virtualization, is essential for address- ing data issues and ensuring the reliability and integ- rity of data used in organizations. Data governance frameworks. Data governance frameworks play a critical role in ensuring the quality of data used in pharmacovigilance. These frameworks provide a set of policies, procedures, and standards that guide the management, usage, and protection of an organization's data. They help to ensure that data are accurate, consistent, and compliant with regulations. There are several data governance frameworks that can be used for pharmacovigilance, including: • T he Ident i f icat ion of Med ic i n a l P roduc t s (IDMP) set of guidelines and standards devel- oped by the International Organization for Standardization (ISO) • The International Council for Harmonisation (ICH) E2E Guideline on Data Governance • The Data Governance Institute (DGI)'s The DGI Data Governance Maturity Model • The Open Data Institute (ODI) Data Governance Canvas • The Data Governance Institute (DGI) DGI Data Governance Capability Model. By implementing these frameworks, organizations can improve the quality of data used in pharmacovigi- lance and support the development of ML models. The frameworks provide guidance on how to collect, man- age, and exchange information about medicinal prod- ucts in a consistent and compliant manner. They also help organizations to identify and correct errors and in- consistencies in data and monitor data quality over time.