Health data is at the heart of modern healthcare innovation and policymaking, promising advances from personalised medicine to pandemic preparedness. The European Health Data Space (EHDS) Regulation, effective as of early 2025, has made significant strides in creating a unified framework that allows individuals greater control over their health data while enabling secondary uses, such as research and innovation, across the European Union (EU). Yet, as the power over health data increases, so too must our vigilance regarding who controls, benefits from, and is included in these systems. Applying a Data Feminism lens – a framework that critically examines power imbalances and systemic inequities in data – reveals that despite positive steps, significant challenges remain. These challenges must be urgently addressed to prevent social hierarchies from being reinforced or deepened by digital health infrastructures.

What is Data Feminism?

Data Feminism considers data “both in terms of their uses and limits, informed by direct experience, commitment to action, and intersectional feminist thought”. It is an approach to data science that critically examines power structures and challenges existing within unequal data practices. It uses feminist theory to analyse how data is collected, analysed, and used, emphasising the need to question who benefits from these practices and whose interests are being served. However, the notion is not limited to gender issues, but has more to do with power, specifically examining who holds it and who does not.

Power structures in health data governance

Health data governance structures tend to prioritise the interests of a handful of actors over those of patients and vulnerable groups, leading to inequities in data access, representation, and outcomes. For example, a study by Sharon and Gellert (2023) on global health data governance has revealed that funding organisations have exerted considerable influence over priority-setting. This results in different transgressions, one involving reshaping spheres in line with their own “values” and “interests”: According to Sharon and Gellert, the more collaboration technology actors engage in with traditional actors, the more significant their agenda-setting role becomes. This can occur when the personal interests of corporate leaders motivate their involvement in a particular sphere or disease, or when there are value conflicts between tech and domain experts, with the former prevailing. An empirical paper which studied Twitter data from 20 global health actors found a strong link between funders’ priorities and agendas of implementing institutions.

Data feminism enables us to examine who holds power in decision-making, who benefits from data collection and analysis, and who is excluded from these structures. For example, it challenges the dominance of certain groups in data science, which can lead to biases in health data analysis and policymaking. Applying such a lens can help critically examine how these power dynamics shape data-sharing practices, legislative decisions, and the distribution of health benefits across the board. For instance, diagnostic algorithms for cardiovascular diseases, which are developed using data skewed towards male symptoms, tend to fail to detect atypical presentations more common in women. Data feminism highlights how these so-called neutral algorithms encode historical inequities. Ignazio and Klein state that objectivity in goods and services flowing from data is objective only because those who produce them are considered the “default”. However, this objectivity is becoming a focus of study as algorithms tend to be sexist, racist, or flawed.

i. Access to health data

The EHDS Regulation aims to facilitate the exchange of electronic health data across Europe, ensuring that individuals can access their health records and share them with healthcare providers when necessary. The governance framework of the EHDS assigns distinct roles to various stakeholders, including data subjects, data controllers, processors, and bodies governed by law. The former three are similarly mentioned in the General Data Protection Regulation (GDPR). According to TEHDAS (2023) data subjects are “natural persons whose data concerns health, falling within EU jurisdiction, is specified as personal data related to the physical or mental data of a natural person, and/or reveals information about one’s health status”. The group of actors designated as data controllers is heterogeneous and ranges from “actors involved in the provision of health care to national health agencies and the private sector”. Data processors are “actors deployed by the data controller to process data on their behalf”. The last group of actors includes “bodies governed by public law with either supervisory competencies over the use and protection of personal (health) data or other tasks relevant to the governance of the health-data landscape”.

Such a framework is designed to ensure compliance with the GDPR whilst maintaining trust and transparency in health data sharing. However, this highlights that the decision-making power still rests within certain entities, raising questions about how health data is shared and used. For example, while patients are empowered to manage their health data regarding primary use specifically, the overarching control remains with the Health Data Access Bodies, which may have different priorities from those of the patients or individuals themselves. Furthermore, both primary and secondary use involve first accessing the data. Under Article 2(2)(d), the primary use of health data refers to the processing of electronic health data for the provision of healthcare services. Whereas, under Article 2(2)(e), secondary use of health data refers to processing of electronic health data for purposes other than the initial purposes for which they were collected. The mechanisms to access this data can create barriers for those on the fringes. If access is mainly granted to those with the most resources or networks, there is a risk of perpetuating inequalities within healthcare and its research benefits.

ii. Quality of health data

The influence of the private sector in health data management raises questions regarding who benefits from data access and who does not. This is because they often prioritise profit-making over societal interests. While the public sector’s attitudes towards healthcare have not been without faults, the private sector’s profit-driven model and lack of democratic accountability pose unique issues about inequitable data access and outcomes.

Furthermore, private entities developing Artificial intelligence (AI) systems for healthcare often rely on datasets that may not accurately represent diverse populations, thereby reinforcing existing biases. Although not entirely perfect and functional, public data sets, such as national health surveys, face stricter requirements in terms of representation. At the same time, this has generally been studied in the context of digital healthcare. It remains to be seen whether today’s dynamic EU regulatory framework regulates infrastructures more effectively. The EHDS is touted to be an open-data initiative aiming to standardise sharing, posing an interesting study to understand whether its implementation truly allows for equitable access to health data. What remains to be studied is whether there is sufficient representation of different stakeholders and communities in deciding what data is collected and processed, and how this would impact the understanding of other health needs and outcomes.

Conclusion

The data governance framework must confront these existing biases to prevent the reinforcement of systemic inequities in healthcare, while recognising how socio-economic factors intersect to shape individuals’ experiences within the healthcare system. To achieve this, the EU could operationalise Article 78(6) of the EHDS Regulation by mandating bias examinations and impact assessments for high-risk AI systems under Article 10 of the AI Act alongside requirements for diverse stakeholder consultations during dataset curation on permitted secondary uses. Integrating socio-economic proxies into fairness metrics for AI-driven health analytics could further align with EHDS Recitals, emphasising equitable data access and quality. Through these targeted measures, the EU can establish a more equitable framework that serves the diverse needs of stakeholders across the healthcare landscape.