How AI can solve healthcare’s equity dilemma | Viewpoint

There was a collective gasp recently when the Medicare Advantage Star ratings dipped for the third consecutive year.

That makes aligning focus with CMS priorities – especially newer ones around promoting health equity – more critical than ever for sustained success.

That’s easier said than done. Meeting patients and members where they are has been an edict for years, but for far too long “strategies” to do just that are list-pulls and batch-sends to thousands of people who have been targeted simply by care gaps or need for an annual physical.

Traditional methods of outreach scoped and designed by marketing teams or clinical staff are more suited for the health system or health plan performing them, versus those who need care.

The issue is that healthcare is about the furthest thing from one size fits all that you can get. “Universal” messaging doesn't take into account the language, type of outreach, or imagery that makes each of us tick, a creation process that’s as inherently biased as most clinical research and trials.

Taking into account genomes and geography

The spotlight on – and funding for – personalized medicine is projected to maintain its current velocity for years to come. But what about truly personalized messaging that takes into account that what we think and feel is as important as what our DNA sequence is?

Value-based care models in the public and private sectors, long dependent on disadvantaged and rural populations for success, could turn the tide on accessibility and profitability by adopting innovations that no longer pit man against machine but instead unite man with machine.

AI already has proven value in the exam room for clinician decision support, for instance. Now, AI needs to be incorporated into patient journeys long before that, merging artificial intelligence and behavioral science not as a black-and-white roadmap but as a GPS that iterates and evolves for all of a patient’s twists, turns, and bumps in the road.

Just as you wouldn’t use the same language, modality, or images to get an 80-year-old grandma to engage in her health outcomes as her 50-year-old daughter (or, let’s face it, what time an email or text is sent), there’s tremendous power – and profit – in true personalization.

Personalizing healthcare has been shown to increase care quality by up to 25 percent, reduce costs by as much as 10 percent, and improve patient/member experience by 10 percent. Given the weight of patient experience when it comes to Star ratings, it’s time to prioritize the customized communications that can lead to longitudinal engagement.

Introducing new technology and new terms

“With health systems worldwide grappling with the moral and financial costs of health inequities, the case for using AI to close the gap could not be stronger,” shares Harvard Business Review. “AI’s capabilities in leveraging multiple different types of data to predict and intervene at all stages of a patient’s journey make it uniquely well placed to address the main causes of health inequities.”

Artificial intelligence allows us to evaluate floods of data that would otherwise deluge health systems with spreadsheets and disparate analytics platforms while also performing agile outreach informed by behavioral science – the study of cognitive, social, and environmental drivers and barriers that influence health-related behaviors. This is where the healthcare ecosystem is at an inflection point: one where designing outreach for equity must be accounted for not as an afterthought but as a priority.

While large language models (LLMs) can support patient engagement efforts, they fall short in helping organizations effectively move patients to take specific actions that lead to better health outcomes.

The reason is simple: when it comes to patient and member engagement, language is only part of the equation. Understanding human behavior is the rest. That’s where Large Behavioral Models (LBM) come in to closely observe individuals’ health journeys, understand their behaviors within their unique contexts, and draw from past experiences to accurately predict likely future actions and potential health impacts.

Research shows, for example, that not all patients with risk factors for conditions like hypertension – which disproportionately affects Black people in the United States, at a younger average age – will take the same actions with the same outreach.

Race, digital access, digital literacy, geographic and nutritional factors, and more social determinants of health must all be considered cumulatively by behavioral scientists like myself.

By doing so, we can use AI to truly deploy the right messaging at the right time, and in the right place, benefiting patients and the organizations responsible for their care.

Sarah Deedat, PhD, is the vice president of behavioral design at Lirio.