Artificial intelligence as a medical device is increasingly being applied to healthcare for diagnosis, risk stratification and resource allocation. However, a growing body of evidence has highlighted the risk of algorithmic bias, which may perpetuate existing health inequity. This problem arises in part because of systemic inequalities in dataset curation, unequal opportunity to participate in research and inequalities of access. This study aims to explore existing standards, frameworks and best practices for ensuring adequate data diversity in health datasets. Exploring the body of existing literature and expert views is an important step towards the development of consensus-based guidelines. The study comprises two parts: a systematic review of existing standards, frameworks and best practices for healthcare datasets; and a survey and thematic analysis of stakeholder views of bias, health equity and best practices for artificial intelligence as a medical device. We found that the need for dataset diversity was well described in literature, and experts generally favored the development of a robust set of guidelines, but there were mixed views about how these could be implemented practically. The outputs of this study will be used to inform the development of standards for transparency of data diversity in health datasets (the STANDING Together initiative).

Artificial intelligence (AI) has been heralded as one of the key technological innovations of the 21st century. Within healthcare, much attention has been placed upon the ability of deductive AI systems to analyse large datasets to find patterns that would be unfeasible to program. Generative AI, including generative adversarial networks, are a newer type of machine learning that functions to create fake data after learning the properties of real data. Artificially generated patient data has the potential to revolutionise clinical research and protect patient privacy. Using novel techniques, it is increasingly possible to fully anonymise datasets to the point where no datapoint is traceable to any real individual. This can be used to expand and balance datasets as well as to replace the use of real patient data in certain contexts. This paper focuses upon three key uses of synthetic data: clinical research, data privacy and medical education. We also highlight ethical and practical concerns that require consideration.

Introduction: The COVID-19 pandemic has made well-fitting face masks a critical piece of protective equipment for healthcare workers and civilians. While the importance of wearing face masks has been acknowledged, there remains a lack of understanding about the role of good fit in rendering protective equipment useful. In addition, supply chain constraints have caused some organizations to abandon traditional quantitative or/and qualitative fit testing, and instead, have implemented subjective fit checking. Our study seeks to quantitatively evaluate the level of fit offered by various types of masks, and most importantly, assess the accuracy of implementing fit checks by comparing fit check results to quantitative fit testing results. Methods: Seven participants first evaluated N95 and KN95 respirators by performing a fit check. Participants then underwent quantitative fit testing wearing five N95 respirators, a KN95 respirator, a surgical mask, and fabric masks. Results: N95 respirators offered higher degrees of protection than the other categories of masks tested; however, it should be noted that most N95 respirators failed to fit the participants adequately. Fit check responses had poor correlation with quantitative fit factor scores. KN95, surgical, and fabric masks achieved low fit factor scores, with little protective difference recorded between respiratory protection options. In addition, small facial differences were observed to have a significant impact on quantitative fit. Conclusion: Fit is critical to the level of protection offered by respirators. For an N95 respirator to provide the promised protection, it must fit the participant. Performing a fit check via NHS self-assessment guidelines was an unreliable way of determining fit.

Artificial intelligence (AI) is widely recognised as a transformative innovation and is already proving capable of outperforming human clinicians in the diagnosis of specific medical conditions, especially in image analysis within dermatology and radiology. These abilities are enhanced by the capacity of AI systems to learn from patient records, genomic information and real-time patient data. Uses of AI range from integrating with robotics to creating training material for clinicians. Whilst AI research is mounting, less attention has been paid to the practical implications on healthcare services and potential barriers to implementation. AI is recognised as a “Software as a Medical Device (SaMD)” and is increasingly becoming a topic of interest for regulators. Unless the introduction of AI is carefully considered and gradual, there are risks of automation bias, overdependence and long-term staffing problems. This is in addition to already well-documented generic risks associated with AI, such as data privacy, algorithmic biases and corrigibility. AI is able to potentiate innovations which preceded it, using Internet of Things, digitisation of patient records and genetic data as data sources. These synergies are important in both realising the potential of AI and utilising the potential of the data. As machine learning systems begin to cross-examine an array of databases, we must ensure that clinicians retain autonomy over the diagnostic process and understand the algorithmic processes generating diagnoses. This review uses established management literature to explore artificial intelligence as a digital healthcare innovation and highlight potential risks and opportunities.

Synthetic data, generated through artificial intelligence technologies such as generative adversarial networks and latent diffusion models, maintain aggregate patterns and relationships present in the real data the technologies were trained on without exposing individual identities, thereby mitigating re-identification risks. This approach has been gaining traction in biomedical research because of its ability to preserve privacy and enable dataset sharing between organisations. Although the use of synthetic data has become widespread in other domains, such as finance and high-energy physics, use in medical research raises novel issues. The use of synthetic data as a method of preserving the privacy of data used to train models requires that the data are high fidelity with the original data to preserve utility, but must be sufficiently different as to protect against adversarial or accidental re-identification. There is a need for the development of standards for synthetic data generation and consensus standards for its evaluation. As synthetic data applications expand, ongoing legal and ethical evaluations are crucial to ensure that they remain a secure and effective tool for advancing medical research without compromising individual privacy.