HEALTH EQUITY ADVOCATE BLOG POST

Zainab Garba-Sani

Vice Chair, Sickle Cell Society.

Founder and Executive Director,

ACCESS AI.

Affiliate Research Scholar,Stanford

University.

Zainab Garba-Sani is an award-winning Stanford-affiliated Research Scholar and Commonwealth Fund Senior Harkness Fellow working at the intersection of emerging science (AI, genomics and regenerative medicine) and community-centred policy and practice. She is Founder & Executive Director of ACCESS AI and Vice Chair of the Sickle Cell Society. As one of the first adults in the UK to receive a pioneering stem cell transplant to cure sickle cell, she brings a rare personal and professional depth to her advocacy.

LinkedIn:  linkedin.com/in/zainabgarba-sani

Why this matters to me?

As a Black, Muslim woman who recently received one of the UK’s first adult stem cell transplants to cure sickle cell, I have seen firsthand the systemic impact of health inequity. Simultaneously, as a researcher, policy practitioner and global health advocate, I have the opportunity (and the responsibility) to bridge the gap between community and practice. My goal is to ensure that as we embark on the era of AI, genomic and regenerative medicine, equity is not just an aspiration, but a fundamental foundation.

What is the problem you are addressing?

Emerging technologies such as artificial intelligence (AI) and genomic medicine have the potential to transform healthcare and redress longstanding inequities. They offer the prospect of earlier intervention, more precise treatment, and care that reaches populations historically left behind. However, they also risk amplifying disparities if not implemented responsibly and intentionally, with communities at the heart.

A landmark study by Obermeyer and colleagues illustrated this risk: a widely used algorithm exhibited significant racial bias because it used healthcare expenditure as a proxy for need. Because systemic barriers meant less was historically spent on Black patients, the AI interpreted lower spend as lower risk, thus denying Black patients access to the same additional support as White patients with the same level of illness. Essentially, when AI is trained on data reflecting existing disparities, it learns to replicate them at scale.

The risk extends beyond AI. In the absence of inclusive datasets, medical advances fall short. Historically, African participants have made up only a fraction of genetic studies, while  samples from populations of European ancestry have accounted for over 90% of genomic databases. This lack of diversity has real-world consequences. For example, while the sickle cell community sits at the frontier of regenerative medicine, the existing primary treatment, Hydroxyurea, fails to work in a proportion of patients. Alarmingly, the drug’s mechanism of action and pharmacogenetics (how a patient’s genes dictate their response) remain poorly understood. The problem is that emerging gene therapies are keen to harness these same biological pathways. Therefore, without a more representative evidence base, we risk inheriting these same blind spots and scaling them into the next generation of care.  

What sits beneath the surface?

A fundamental driver of health inequity is the absence of meaningful community engagement across the research and innovation pipeline. Beyond the representation challenge, we have complex underlying drivers including longstanding exploitation, structural discrimination, and legitimate mistrust. These barriers hinder genuine partnerships that allow communities to shape research questions, outcome measures, and healthcare priorities.

History is rife with marginalised communities being exploited in medical research. Where biological material, clinical data, and lived experiences have been extracted, yet the communities are left with little control over how the information is used or who benefits. From the extraction of Henrietta Lacks' cells without consent to the Havasupai Tribe's case (where DNA collected for diabetes research was used for studies the community never agreed to), exploitation has defined the relationship between marginalised communities and medical science for generations. Even today, many in the sickle cell community express a live and legitimate fear of being "guinea pigs" for innovations badged for them but not primarily designed to serve them.

Disengagement is therefore a logical response to a system where data flows away from communities to serve interests that are not their own. It is unsurprising when a community’s experiences continue to be miscoded, needs ignored, concerns dismissed, and treatments developed from their biology, priced beyond their reach. Most concerningly is when institutions use this resulting disengagement as a justification for continued exclusion from the datasets and decision-making processes that shape the community’s care.

Why should this matter to decision-makers?

The case for equity in research and innovation is not only about ethics, but also quality, safety and system efficiency. Take AI as a prime example: training data that is systematically skewed leads to inaccurate model performance in underrepresented populations. This produces unreliable outputs precisely for the very patients who face the greatest burden of disease and have the fewest alternatives. As seen in the previously mentioned Obermeyer study, models that use biased proxies (like expenditure) under-refer vulnerable communities for care. This not only perpetuates inequality but also leads to inaccurate resource prioritisation and potentially waste. Similarly, an algorithm trained on predominantly lighter skin tones, may fail to detect skin cancer in darker skin, directly hindering early diagnosis, worsening outcomes, and increasing the cost of care.  

For industry, the exclusion of sickle cell and other racialised communities from clinical trials represents both a scientific blind spot and a missed opportunity. Gene therapies and disease-modifying treatments for sickle cell represent a significant and growing market, yet the evidence base informing regulatory decisions, reimbursement frameworks, and clinical practice remains thinner than it should be. This is partly because the patients who need these treatments most have not been meaningfully included in generating the knowledge surrounding them.

For healthcare leaders, the downstream cost of underserved chronic disease is substantial. It manifests in avoidable emergency admissions, long-term complications, workforce absences, and an erosion of trust that makes communities less likely to engage with the health system at all. Ultimately, investment in genuine co-production should be viewed as an essential tool for risk management and clinical effectiveness.

From Awareness to Action

The A.C.C.E.S.S. AI framework offers a practical model for interdisciplinary stakeholders to collectively engage communities, identify population-specific barriers, and design health AI and innovation initiatives that actively advance equity rather than replicate existing disparities.

Furthermore, evidence-based frameworks such as the AHSN Network's EnACT principles reinforce this approach. They make clear that meaningful engagement is not a “tick-box” exercise but something that requires representative reach, fair compensation and acknowledgement, real transparency, and genuine accountability.

In developing these two models (as lead author and contributor, respectively), it became clear that equity cannot be retrofitted at the end of a development process. Instead, it must be a design requirement from the outset. The table below builds on this foundation with some practical recommendations.

What Must Change	Who Has Responsibility	What Progress Should Look Like
Behavioural Change
Move from consultation to co-design	Policymakers, Regulators, Commissioners, Industry, Researchers, Healthcare Providers	Engagement starts at the "problem identification" phase, and is maintained throughout the development pipeline, not just after the product is built.
Prioritise equity as a primary metric	Policymakers, Regulators, Commissioners, Industry, Researchers, Healthcare Providers	Equity is treated as a technical requirement (like accuracy) rather than a "nice-to-have" social good.
Commit to real transparency	Industry, Researchers, Healthcare Providers	Openly sharing where models fail or where data is unrepresentative, reducing "black box" secrecy.
Policy and Systems Change
Introduce mandatory equity audits	Regulators, Commissioners, Industry, Healthcare Providers	Demographic performance breakdowns are a mandatory condition for market approval; and publicly published and monitored post deployment.
Establish participatory governance practices	Regulators, Healthcare Providers	Governance structures that allow communities to "pause" or "pivot" a rollout if bias is detected.
Ringfence community benefit funds	Funders, Industry	Mandating a set percentage of project budgets for community compensation and capacity building.
Establish fair compensation framework	Policymakers	Standardised PPI (Patient and Public Involvement) payment scales that value community expertise like professional consultancy.
Enable flexibility within consent and data sharing agreements	Policymakers	Implementing differential consent and data protection models that prioritise community benefit and authority over profit.

A question for the sector

If the communities most harmed by health inequity were given genuine authority (not just a seat at the table) over how health innovation is designed, implemented, governed, and evaluated, what would have to change about who holds power in this sector, and are we prepared to make that change?

Key Publications:  

· A.C.C.E.S.S. AI: A New Framework For Advancing Health Equity In Health Care AI

· How to involve and engage patients in digital health tech innovation: An Evidence Based Guide 

· Attitudes Toward Personal Health Data Sharing Among People Living With Sickle Cell Disorder, Exemplar for Study of Rare Disease Populations 

Walk a mile in my shoes: perspectives towards sharing of health and experience data among individuals living with sickle cell disorder