Supporting the drive towards equity in evidence-based healthcare
Author: Prof Lisa Whitehead1,2, Prof Ritin Fernandez3,4, Mary Anne Ramis5,6, A/Prof Kate Kynoch5,6,7, A/Prof Irene Ngune2,8, Katrina Lenzie-Smith9,10, Casey Peiris, Prof Janie Brown11,12
1. Edith Cowan University
2. JBI Centre for Evidence Informed Nursing and Midwifery
3. Newcastle University
4. JBI Newcastle Centre for Transformative Nursing and Midwifery
5. Mater Health
6. JBI Queensland Centre for Evidence-Based Health Innovation
7. Queensland University of Technology (QUT)
8. Edith Cowan University
9. Royal Melbourne Hospital
10. JBI Centre Royal Melbourne Hospital Nursing and Allied Health
11. Curtin University
12. JBI Evidence Informed Healthcare Curtin University
Artificial intelligence (AI) has the potential to transform healthcare, but without a deliberate focus on equity, it may also amplify existing health disparities. Members of the JBI Centres based in Australia met to reflect on and discuss the issues around AI becoming increasingly embedded in clinical practice, and the risks we need to anticipate in relation to existing health inequities. This blog reflects our discussion and argues that AI governance, evaluation and implementation, assessed through an equity lens, must become a central pillar of evidence-based healthcare.
It is hard to ignore the growth in digital health and, increasingly, the AI systems that inform the systems that deliver and respond to the management of health and wellbeing. We have found through our work that access to and engagement with digital devices and platforms are not equitable across populations. As a group, we can see that this raises multiple questions: while AI systems are presented as objective and efficient, they are only as robust as the data they are trained on and the governance structures that oversee their use. AI systems are reliant on open-access or readily available evidence, and when consumers, clinicians and researchers lack the resources to make their experience and evidence easily accessible, their work may be less visible to AI systems. Hence, we propose that AI-generated evidence risks underrepresenting geographic regions and populations with limited resources. AI systems are, therefore, prone to risk of bias and error, which can be introduced at the algorithm level, the data input level or within the application being used.
Consider an AI-enabled symptom management system that aims to support people living with cancer to stay at home and avoid the need for hospital or GP visits. If the underlying model has been trained primarily on data from urban or mainstream populations, its predictions around symptom severity and duration, and its symptom management advice, may be less accurate and relevant for people living in rural and remote communities, or individuals from culturally and linguistically diverse backgrounds. We argue that AI may, therefore, reinforce existing disparities by systematically misclassifying risk or overlooking contextual factors such as socioeconomic circumstances. These could include sociodemographic information or details relating to digital access, language or geographical barriers.
This is where equity and systematic validation become essential. The critical question is no longer simply whether technology works, but for whom it works, under what circumstances, who has access to it and who may be left behind. Importantly, an AI system can be technically accurate yet still produce inequitable healthcare if some populations cannot access, understand or meaningfully engage with it. Equity, therefore, needs to be considered across the AI lifecycle, from development and validation through to implementation, monitoring and review.
As a group of researchers and clinicians who conduct systematic reviews, we believe that they play an important role in building and sustaining evidence-based healthcare. In the age of AI, we propose that they can also provide an important mechanism for examining representation, contextual sensitivity, transparency and real-world applicability. By identifying who was included in research, where AI systems were tested, who may have been excluded and whether outcomes differ across populations, evidence synthesis can help reveal where evidence to support equitable implementation is lacking. Tools that can help readers and reviewers to do this include the CONSENSYS instrument, a question-based framework that helps readers consider the cultural, social, organisational, political and practical factors that affect whether research evidence can be successfully implemented in a particular setting, rather than assuming that evidence can be transferred directly from one context to another. Similarly, the CONSORT-AI reporting guidelines aim to improve the transparency and completeness of reporting clinical trials involving AI. The areas covered include information about AI development, data handling, error cases and model performance, which together, can help reviewers critically consider generalisability and potential inequities. Together, these frameworks support us as researchers and clinicians to move beyond traditional measures of effectiveness by examining whether AI tools are developed, evaluated and implemented in ways that support equitable care.
Following evidence synthesis is implementation, and we encourage interrogation around whether evidence of efficacy can translate into equitable implementation. Differences in infrastructure, workforce capability, digital literacy, language, socioeconomic circumstances and access to healthcare may influence whether the benefits of an AI intervention are realised in routine practice. We argue that AI should augment, rather than replace, contextualised clinical judgement. Healthcare professionals remain essential in recognising circumstances that may not be captured by an algorithm and in ensuring that technology supports, rather than undermines, person-centred care.
We all need to continue to ask difficult questions about data bias, representation, transparency, accountability and implementation. Ultimately, evidence-based healthcare is not simply about improving outcomes. It is about understanding whether an intervention is effective, safe, accessible and equitable for the populations and contexts in which it will be used.
As AI increasingly shapes clinical decision-making, equity can no longer be treated as an optional consideration. It must be embedded in governance, evidence synthesis, evaluation and implementation from the outset, and monitored throughout the AI lifecycle. Only then can AI fulfil its promise of reducing health disparities rather than deepening them.
References
Bengough, T., Sommer, I., & Hannes, K. (2023). The CONSENSYS approach: An instrument to support CONtextual SENsitivity in SYStematic reviews. Research Synthesis Methods, 14(2), 266–282. https://doi.org/10.1002/jrsm.1615
Liu, X., Cruz Rivera, S., Moher, D., Calvert, M. J., Denniston, A. K., & The SPIRIT-AI and CONSORT-AI Working Group (2020). Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: The CONSORT-AI extension. Nature Medicine, 26, 1364–1374. https://doi.org/10.1038/s41591-020-1034-x
Osonuga, A., Osonuga, A. A., Fidelis, S. C., Osonuga, G. C., Juckes, J., & Olawade, D. B. (2025). Bridging the digital divide: Artificial intelligence as a catalyst for health equity in primary care settings. International Journal of Medical Informatics, 204, 106051. https://doi.org/10.1016/j.ijmedinf.2025.106051
Resnik, D. B., & Hosseini, M. (2025). The ethics of using artificial intelligence in scientific research: New guidance needed for a new tool. AI and Ethics, 5(2), 1499–1521. https://doi.org/10.1007/s43681-024-00493-8
Whitehead, L., Robinson, S., Arabiat, D., Jenkins, M., & Morelius, E. (2024). The report of access and engagement with digital health interventions among children and young people: Systematic review. JMIR Pediatrics and Parenting, 7, article e44199. https://doi.org/10.2196/44199
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
