Issue link: https://www.e-digitaleditions.com/i/1494960
PharmTech.com Quality and Regulatory Sourcebook March eBook 2023 Pharmaceutical Technology ® 15 Pharmacovigil ance For example, the IDMP framework can be used to standardize the identification of medicinal products and ensure consistency in the data collected. The ICH E2E guideline on data governance can be used to en- sure that data governance practices in the organization meet regulatory requirements and support the develop- ment of ML models. The DGI Data Governance Maturity Model can be used to assess and improve data gover- nance practices in the organization and identify areas for improvement. The ODI Data Governance Canvas can be used to design and implement data governance processes that support the development of ML models. Data catalogues and metadata management sys- tems. Data catalogues and metadata management sys- tems are practices in data management that can help to improve the quality and reliability of data used in pharmacovigilance. A data catalogue is a centralized repository of information about data assets, such as data sources, data quality, and data lineage. It enables pharmacovigilance, professionals to easily discover, un- derstand, and access data assets, and it helps to improve data governance and compliance by providing a single source of truth about data. A metadata management system, on the other hand, is a tool that can be used to manage the metadata asso- ciated with data assets. It enables pharmacovigilance professionals to capture, store, and manage meta- data in a consistent and standardized way. Metadata includes information such as data definitions, data lineage, data quality, and data governance policies. It enables pharmacovigilance professionals to under- stand the context and quality of data, and it can be used to improve data governance and compliance, and to ensure that data are used correctly. By using data catalogues and metadata management systems, pharmacovigilance professionals can improve the quality and reliability of data used in pharmacovigi- lance. They can easily discover, understand, and access data assets, improve data governance and compliance, and ensure that data are used correctly. Furthermore, by having accurate metadata, pharmacovigilance pro- fessionals can trace the lineage of data and understand the context and quality of data which can be used to demonstrate the reliability and validity of ML models and identify any issues with the data that may impact model performance. Using data lakes. Data lakes can be particularly use- ful for ML in pharmacovigilance as they function as centralized repositories that store structured and un- structured data in its raw form, allowing for easy access and analysis by multiple users. The advantage of this approach is that it enables a wide range of use cases and improves data reusability. One example of how a data lake can be used to improve data quality and integrity in pharmacovig- ilance is by storing and curating electronic health re- cords (EHRs) and integrating them with other data sources such as clinical trials and social media. This allows for better management of patient data and improved accuracy of ML models, leading to a more comprehensive view of patient information. Another example is the use of a data lake in clinical trials to store and process large amounts of data gener- ated from the trials, including both structured and un- structured data such as patient records, lab results, and imaging studies. By using a data lake, one can ensure that the data are properly curated and governed, which will improve the quality and integrity of the data. This will allow for more accurate analysis of the data and improved decision making in pharmacovigilance. A data lake can also be used to monitor and analyze social media data for potential adverse reactions to drugs. By using natural language processing tech- niques, the data lake can extract relevant informa- tion from social media posts and analyze it in re- al-time. This can help identify potential safety issues with drugs and allow for more efficient monitoring of the drug's safety profile. However, managing large amounts of raw data in a data lake presents its own set of challenges, including data modeling, indexing for efficient querying, and inte- grating suitable and scalable computing capabilities (3). Data virtualization. Data virtualization is a prac- tice that allows users to access and analyze data from multiple sources as if it were stored in a single loca- tion. In the context of pharmacovigilance, data virtu- alization can be used to support the development of ML models by integrating data from various sources, such as electronic health records, clinical trial data, public data sets, and internal data sources. While the use of data virtualization offers many benefits, it can be a complex and challenging process. Companies often find it difficult to implement data virtualization due to high costs, heavy information technolog y (IT) and cloud technolog y resource re- quirements, the need for domain experts, and the fact that it needs to be an enterprise-wide initiative with executive leadership commitment. It can also take two to three years for any results. Given these challenges, companies are looking for vendors who can simplify the process and deliver re- sults quickly without disturbing business as usual. They also want vendors who can provide cost-effec- tive solutions. Data virtualization vendors are focus- ing on developing solutions that can be implemented quickly and easily, and that do not require extensive IT resources or domain expertise. It's worth noting that while these practices have been around for some time, the real challenge is in curating the data and investing in the systems and resources to maintain a data governance framework. This is something that needs to be done at an enter-