One of the biggest shortcomings of modern healthcare is inequality in access to and standards of care.

Too often, rural and underserved communities receive substandard and inadequate care, and sometimes no care at all if they are located in one of many healthcare deserts. Even in metropolitan areas, the digital divide is a major barrier to patients accessing timely and appropriate  care, leaving millions of Americans disconnected from their healthcare providers. Social determinants of health (SDoH) are the root cause of many of these gaps; left unaddressed, they threaten to exacerbate health inequities.

SDoH refer to the social and economic conditions that affect health and quality-of-life outcomes and risks, such as access to housing, education and technology; clean air, water and nutritious foods; language and literacy skills; and prevalence of racism, discrimination and violence. Medical research has uncovered the widespread ways in which SDoH influence everything from medication adherence and medical literacy, to chronic disease progression and quality of life.

Unfortunately, advanced technologies like artificial intelligence (AI), can unknowingly perpetuate existing SDoH biases rather than counteracting them. This paradox occurs because probabilistic AI algorithms are trained on data that contain human conscious and subconscious biases and then reflect them into the real world. The AI system accepts these biases as truths and propagates stereotypes, leading to further marginalization of underserved communities. Combining artificial intelligence with automation results in what is called intelligent automation,  which may be the key to reversing these inequities. Automation is the completion of a task by an agent without the agent being explicitly instructed to do so — in other words, it is the use of technology to reduce human intervention in processes. In this way, automation can expand the capacity of our human workforce allowing them to accomplish more with less. This additional capacity, if channeled appropriately, can be devoted to the patients and populations that are most disadvantaged by the current healthcare ecosystem.

As a physician, I’ve been disappointed by how significantly social determinants can influence the ways in which minority populations receive substandard care. Yet in my other role as a medical director at a digital health company, I’m optimistic in how intelligent automation can be an instrumental tool in leveling the playing field.

1. The inability to communicate with patients in their native language.

As a hospitalist, I frequently admit patients to the hospital who do not know English and can only communicate in a foreign language. More often than not, the patient’s care team is not fluent in the patient’s native language and handles this language divide by simply having fewer and less thorough interactions with the patient. Interpreters are difficult to schedule, phone translators are clunky to use, and families can be tough to get a hold of. As a result, it is not uncommon for providers on rounds to suggest “circling back when the family can help translate”rather than seeing the patient in person. There are many other examples of how patients who do not speak English receive suboptimal care. For example, consent forms are often only printed in English, which should leave us wondering if a foreign language-speaking patient’s “informed consent” was truly ever “informed.”

Automatic language translation to hundreds of foreign languages is one area where automation can be invaluable in reversing inequities. Training an entire medical staff to speak every foreign language is not just impractical, it’s impossible. On the flip side, training a “digital assistant” to learn the most common languages and automatically present material in a patient’s preferred language is not only possible, it’s trivial.

Automation is well-equipped to intake large amounts of data, manipulate this data instantly (such as translating it to a foreign language), and present it back to users, all without burdening human staff. Automating routine or manual tasks like patient outreach or clinical documentation in a patient’s native language can provide a delightful patient experience while simultaneously allowing clinicians to focus on higher-level tasks.

2. Patients’ fear of judgment by the medical community.

It is well established that some minority populations distrust the medical system, driven by fear of judgment, mistreatment, or stereotyping by healthcare professionals. A consequence of this distrust is that many patients may be less willing to be honest with their providers about sensitive topics like substance use, alcohol consumption and sexual activity.

In my own practice, I’ve often had patients tell me that they were taking a certain medication or that they had stopped abusing a particular drug only to find out later that this wasn’t the case. I don’t blame them for hiding the truth, but not knowing the truth did make it harder to provide them the best care.

Automation can be used to digitally collect patient histories wherein patients answer questions asked by a bot rather than by a human. With automated digital intakes, patients can answer sensitive questions on their own mobile device, in the privacy of their home without the perception of being judged by a physician sitting a few feet away from them. Although this type of digital interaction may seem less “compassionate” at first glance, there are certain cases in which it can lead to more honest, open and substantive discussions and can allow patients to feel more comfortable than they would sharing this information with an unfamiliar physician.

3. Implicit bias permeates our clinical decisions.

In healthcare, care teams strive to provide the best outcomes for every patient, but our unconscious biases often cloud our judgment and cause us to deliver substandard care. In contrast to AI alone which perpetuates these biases, rules-based deterministic automations can be used to standardize care regardless of a patient’s skin color, native language or appearance.

In one recent example, a study found that black and Latinx patients were less likely than white counterparts to be admitted to a cardiology service for heart failure compared to a general medicine service. Being admitted to a cardiology service was associated with fewer readmissions and improved outcomes. This disparity in admission may be a result of hidden biases contained by the admitting providers. In contrast, intelligent automation can be used to establish deterministic rules for when a patient should be admitted to one service or another — for example, any patient with objective signs of heart failure such as an elevated NT-proBNP and imaging findings of pulmonary edema may be automatically suggested for admission to the Cardiology service.

As a healthcare system, we have an incredible opportunity to leverage intelligent automation to combat healthcare inequities. Automation expands the capacity of care teams, allowing providers to accomplish more with less. As a result, automation can offer an unbiased and personalized mechanism to reach patients in the manner that best suits them — whether that’s in the form of their native language, a digital intake, or an unbiased recommendation. It is imperative that we do more to resolve healthcare disparities, and intelligent automation may be an important solution in our toolkit.