Author: Daliya Rizvi
Department of Biology
Duke University, Durham, North Carolina
A physician I surveyed for a research project said it plainly: “Cultural curiosity is not a major consideration for the majority of the healthcare workforce.” He wasn’t confessing a failure. He wrote this sentence matter-of-factly, the way you’d describe a hospital’s parking situation. That was the moment I understood the depth of the problem. I had spent months conducting literature reviews and surveying palliative care providers in the United States and Pakistan, trying to understand why end-of-life care so consistently fails patients from non-Western, non-white, and marginalized backgrounds. I expected to find gaps in resources, in staffing, in access. What I found instead was something more fundamental. Palliative care as a discipline was built around a single cultural assumption: that a good death means an autonomous individual making informed, documented choices about their own care. That assumption is Western, not universal. And when you treat it as universal, you don’t just fail patients from other cultures. You make their values invisible.
Consider the Ojibwe elders. In Ojibwe culture, caregiving is communal and death is understood as a natural continuation of life, not a medical event to be managed. When researchers studied end-of-life attitudes among Ojibwe elders, they found that many saw no need for advanced directives or preemptive planning, not because they lacked awareness, but because their cultural framework already had death covered. Their younger relatives would care for them. A peaceful death was expected, not negotiated. The palliative care system reads this as non-compliance. It is not. It is a coherent, dignified approach to dying that the system was never designed to recognize.
The failure is not always philosophical. Sometimes it is as concrete as a meal. Research on Jewish and Muslim patients in Western palliative care settings reveals that patients who have no philosophical objection to end-of-life care cannot fully access it because the system is culturally illiterate at the most basic level. Jewish patients cannot get kosher food. Muslim patients cannot get a halal meal, or an imam to recite the prayers their faith requires at the moment of death, or a bed oriented toward Mecca. These are not luxury requests. They are the difference between a death that affirms a person’s identity and one that erases it. The system is not actively hostile to these patients. It is simply indifferent to them.
Then there is the case of Black Americans, whose lower rates of advance care planning and palliative care engagement are routinely framed in the literature as a problem of awareness or mistrust to be corrected. This framing is itself a form of harm. Black Americans’ reluctance to engage with a medical system that conducted the Tuskegee Syphilis Study, that has documented racial disparities in pain management, that refers Black patients to palliative care at lower rates than White patients- is a rational response to a system with a documented record of failure. When we treat it as a communication problem to be solved with better outreach, we avoid the harder question: what would it take for the system to actually deserve that trust?
The scale of this failure is staggering. Every year, approximately 56.8 million people around the world need palliative care. Only 14% receive it. The people least likely to receive it are overwhelmingly the people whose values, beliefs, and histories the system was least designed to serve– immigrants, racial and ethnic minorities, Indigenous communities, the unhoused, the poor. We have built an elaborate infrastructure for a good death, and then made it structurally inaccessible to most of the world.
Even worse, the problem is about to get harder to fix. Across medicine, artificial intelligence is increasingly being used to guide clinical decision-making: predicting prognosis, flagging patients for palliative referral, shaping treatment recommendations. In theory, AI promises consistency and objectivity — a system that sees every patient the same way. In practice, it promises something far more dangerous: the laundering of existing bias into the appearance of neutrality. AI systems do not generate wisdom from scratch. They learn from existing data, which means they learn from existing failures. A system trained on decades of palliative care practice — practice that already under-refers Black patients, already ignores cultural and spiritual needs, already treats Western assumptions about autonomy as universal facts — will not correct for those patterns. It will encode them, accelerate them, and make them harder to challenge because they will no longer look like human decisions. They will look like data. And the scale compounds the problem. A single biased physician makes biased decisions for their patients. A biased algorithm makes biased decisions for every patient in every hospital system that deploys it — simultaneously, invisibly, at a speed that makes correction nearly impossible before the damage is done. We are not approaching a future where AI might inherit our assumptions about death. We are already building it. If we have not reckoned with whose vision of a good death palliative care was built around, we are not ready to hand that vision to a machine. Right now, we are doing so regardless.
Palliative care was not designed with most of the world in mind. It was built around a particular vision of a good death (peaceful, autonomous, documented, and individual) that reflects the values of the culture that created it. For the billions of people whose understanding of dying is communal, spiritual, fatalistic, or simply different, that vision is not neutral. It is a wall. Until we acknowledge that palliative care has a cultural foundation rather than a universal one, we will keep building systems that ask patients to die on our terms instead of their own. Everyone deserves a graceful death. In today’s system, most people don’t get one.
Summary:
Palliative care was built around a Western, individualist vision of a good death, and treating that vision as universal actively harms patients whose values, beliefs, and cultures it was never designed to serve.
The failures are not accidental or fixable through awareness training. They are structural, embedded in the design of the system itself.
AI is about to automate and scale these biases at a speed and opacity that makes correction exponentially harder, and we are deploying it before we have reckoned with whose vision of death it is encoding.
References:
1.Bazargan M, Bazargan-Hejazi S. Disparities in Palliative and Hospice Care and Completion of Advance Care Planning and Directives Among Non-Hispanic Blacks: A Scoping Reviews of Recent Literature. American Journal of Hospice and Palliative Medicine®. 2020 Dec 8;38(6):104990912096658.
2.Dennis MK, Washington KT. “Just Let Me Go”: End-of-Life Planning Among Ojibwe Elders. The Gerontologist. 2016 Dec 7;58(2):gnw151.
3.Klitzman R, Di Sapia Natarelli G, Garbuzova E, Sinnappan S, Al-Hashimi J. Muslim patients in the U.S. confronting challenges regarding end-of-life and palliative care: the experiences and roles of hospital chaplains. BMC Palliative Care. 2023 Mar 27;22(1).
4.Moreman CM, Chatterjee A. A Survey of Jewish Attitudes and Experiences Relating to End-of-Life Care and the “Right to Die.” The American journal of hospice & palliative care/The American journal of hospice & palliative care (Online). 2024 Mar 30;42(2).
5.Muishout G, van Laarhoven HWM, Wiegers G, Popp-Baier U. Muslim physicians and palliative care: attitudes towards the use of palliative sedation. Supportive Care in Cancer [Internet]. 2018 May 8;26(11):3701–10. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6182360/
Declaration of Interests
I have read and understood the BMJ Group policy on declaration of interests and declare the following interests: none.