BioPharm International - October 2021

BioPharm-October 2021-Regulatory-Sourcebook

Issue link:

Contents of this Issue


Page 23 of 33

24 BioPharm International eBook October 2021 ogy," said Former FDA Commissioner Stephen Hahn (7). For example, with the advances in data collection and management, the agency has been emphasizing the use of real-world data (RWD) and real-world evidence (RWE). RWD is gathered from a variety of sources that can pro- vide information about a person's health status. RWE is the clinical evidence derived from analyzing RWD regarding the usage, benefits, and risks of a medi- cal product. The real-life clinical perfor- mance of a medical product can be more clearly demonstrated through RWD/ RWE because a controlled clinical trial often cannot evaluate all applications of a product in clinical practice across the full range of potential users (8). Obtaining a suff icient amount of RWE for regulatory use and other deci- sion-making processes requires a large quantity of RWD. The data then needs to be collated into an analyzable format to ensure accuracy and reliability. AI has become invaluable for collecting and analyzing data. In 2018, researchers conducted a study of myopia in Chinese school-aged children. They collected RWD from electronic medical records across eight ophthalmic centers. The study involved the use of an ML algo- rithm that would predict myopia among children as young as three years old (9). CHALLENGES WITH AI With the increased use of AI, data are rapidly becoming an organization's most valuable asset and a catalyst for nearly unlimited possibilities to create business value. Still, because data are an essential component of modern technology and innovation, there are growing concerns regarding the responsible use of it. Given that the technology is highly data-cen- tric, biased data will lead to biased con- clusions, resulting in decisions based on skewed information. Therefore, a signif- icant challenge with AI is data quality. The technology cannot be of value if it is provided incomplete, biased, or inac- curate data. The World Health Organization (WHO) published a report that points out that opportunities and risks are l inked, a nd it caut ions about t he unethical collection and use of data (10). To illustrate the global impact of the technolog y, the organization warns that AI systems trained primar- ily on data collected from individuals in high-income countries may not per- form well for individuals in low- and middle-income settings. Moreover, AI technology might not be available for a specific population. In this scenario, no data or poor-quality data could dis- tort the performance of an algorithm, resulting in inaccurate outcomes. Compiling data in resource-poor set- tings is difficult and time consuming. It also imposes an additional burden on the workers in the health care commu- nity. Sufficient data for generating accu- rate and unbiased outcomes is unlikely to be available on the most vulnerable or marginalized populations, includ- ing those where health care services are lacking. Data may also be difficult to collect because of language barriers or mistrust, leading people to provide incorrect or incomplete information. Mishandling data can undermine trust in the entire digital ecosystem. Ethical data science requires careful consideration of data's potential and impact in the various contexts where data are gathered and used. Therefore, WHO advises that AI systems be care- fully designed to reflect the diversity of socio-economic and health care settings and be accompanied by digital skills training and community engagement. STRATEGIES FOR SUCCESSFUL AI IMPLEMENTATION Effective data stewardship is a scenario where people and technology work in unison to foster a culture of quality stan- dards and best practices. The Healthcare Information and Management Systems Society outlined the following core prin- ciples addressing the ethical use of data in AI technologies (11): • Reliability and safety—Data han- dling could impact research and clinical decision-making, which may possibly result in different outcomes. Regardless of the tech- niques used for gathering and maintaining data, effective A I depends on quality data collected from reliable sources. • Fairness and inclusivity—AI systems should not only treat patient data in a balanced and fair way, but also not impact similar groups of people in different ways. To eliminate bias in research and clinical practice, inclu- sivity must be incorporated into the design of AI systems. • Transparency and accountabil- ity—If AI systems are used to help make health care-related decisions, understanding how those decisions are made must be transparent to the key stakeholders. Explainable AI makes it easier to identify and raise awareness of potential bias, errors, and unintended outcomes. Specifically, the aim of AI systems is to carefully consider the quality of data and ensure the data clearly convey what the AI team needs to learn. This requires focusing on data that cover important cases and are consistently labeled so that the AI can learn from these data what it is supposed to do. A common scenario that hamstrings data quality is siloed data. When disparate business units maintain their own data- bases, stakeholders are unable to effec- tively share and access data. This leads to issues such as duplicate, incomplete, and unreliable information. Having a single source for data augments a com- pany's ability to ensure data reliability, inclusivity, and transparency. AI REGULATORY OVERSIGHT GOING FORWARD U ltimately, globa l reg ulator y bod- ies will define and publish guidelines regarding data handling, which will include procedures for holding com- panies accountable should they fail Regulatory Sourcebook Data Integrity

Articles in this issue

Links on this page

Archives of this issue

view archives of BioPharm International - October 2021 - BioPharm-October 2021-Regulatory-Sourcebook