Artificial Intelligence and Equity in Access to Palliative Care: Promise or Pitfall?

Palliative care plays a vital role in modern medicine. Its holistic approach addresses not only physical pain but also the psychological, social, and spiritual suffering of patients and their loved ones. The palliative approach aims to be integrated from the time of diagnosis of a serious illness, complementing curative treatments. It includes symptom management, psychosocial support, and advance care planning.

Yet, in practice, this integration often happens too late, or not at all. Many people with serious illnesses, such as cancer, miss out on the benefits of palliative care. Two key barriers persist: delays in identifying who would benefit from a palliative approach and healthcare professionals’ discomfort in initiating end-of-life conversations.

Artificial intelligence (AI) could help change this. In a U.S.-based study, researchers tested an intervention that combined a machine learning algorithm predicting six-month mortality risk for cancer patients with automated text message reminders sent to physicians on the day of the patient’s visit. Physicians also received weekly emails comparing their end-of-life discussion rates to those of their peers.

At first glance, this type of technology could lead to more equitable access to palliative care by identifying patients who could benefit and prompting earlier end-of-life conversations.

AI and the Potential for Earlier, Targeted End-of-Life Conversations

Tools like these could help reduce disparities in palliative care access. Marginalized populations, cultural minorities, rural residents, and people living in poverty, often face unacceptable delays in accessing palliative care. By standardizing referral criteria, AI could help ensure that high-risk patients are identified, regardless of background or social status.

AI may also support clinicians directly. We know that prognostic uncertainty, taboos around death, fear of harming the therapeutic relationship, and optimism bias all contribute to delays in initiating end-of-life conversations. AI-generated predictions, grounded in robust data, could complement clinical judgment and encourage more timely, evidence-based discussions.

But the Equity Risks Are Real

Behind these promises, however, lie significant risks. One of the greatest concerns is algorithmic bias. If an AI model is trained on data from more privileged populations, it may perform poorly when applied to underserved or marginalized groups exacerbating the very disparities it aims to reduce.

Another issue is performance. Many current algorithms correctly identify fewer than 30% of patients who are truly at high risk. Over-reliance on these tools could give clinicians a false sense of security and delay end-of-life conversations even further.

Moreover, initiating end-of-life conversations is not a simple task. It requires sensitivity, clarity, and language that resonates with each patient’s values, often across multiple visits. These conversations cannot be reduced to a single nudge or prediction.

Even once a patient is identified as a good candidate for palliative care, that care doesn’t materialize on its own. The clinician must still take steps to ensure the patient is referred to and supported by the right team. In a context where access to palliative care remains limited, identifying the need is only part of the equation, ensuring actual access is another challenge entirely.

As with many things, AI is neither a magical solution nor an inevitable danger. Its impact depends on how it's implemented, and on which blind spots are acknowledged and addressed. When designed with care, AI can help systematize and normalize early palliative care conversations.

We are still at the very beginning of exploring the potential of artificial intelligence (AI) in palliative care, but equity must remain the guiding principle.