Why Programs Addressing Social Drivers of Health Keep Missing the Mark, and What Would Fix Them

Ask anyone who's worked in public health and they'll tell you the same thing: where you live shapes how long you live. A stable home, reliable transportation, access to fresh food, these aren't side issues to medical care. They're often the main event. The research backing this up isn't new or controversial. What's less understood is why so many programs built to address these social determinants of health, known in the field as SDoH, keep falling short of their promise.

The problem isn't the concept. It's the execution, and specifically, the data behind it.

For years, health plans have had to make funding decisions based on educated guesses. They'd look at a population and make broad assumptions about who needed help and why. Once a program launched, there was often no reliable way to track whether it was actually working. Imagine trying to treat a patient without ever checking their vital signs again after the first visit. That's roughly the position many plans have been in.

Part of the issue is definitional. The industry has struggled for decades to pin down what counts as a social determinant of health in consistent, measurable terms. The data that does exist tends to be collected at the neighborhood level rather than the individual level, which is a bit like diagnosing a single patient based on the average health of their zip code. When individual data is available, it usually comes from patient surveys or medical claim codes called Z-codes, a process that's slow, expensive and often incomplete.

Building the Capabilities That Actually Work

Simply pouring more money into these programs won't solve the underlying problem. What's needed instead is a shift toward three specific capabilities that, taken together, can change the trajectory of SDoH work.

The first is making individual-level data the foundation, not an afterthought. Claims data alone only captures billed care, which leaves enormous gaps. To get a real picture of a person's health, that claims data needs to be combined with clinical information, things like lab results, immunization records and prescriptions, alongside social risk data drawn from public and private sources that capture financial strain, food insecurity, housing instability, transportation barriers, limited digital access and social isolation. Broad demographic assumptions simply can't substitute for this kind of granular, person-specific picture. It's the foundation everything else depends on.

The second capability involves using what's sometimes called stacked risk to target interventions with precision. Knowing that food insecurity affects a large population is interesting, but not especially actionable. Knowing that one specific individual is dealing with food insecurity, a high-risk pregnancy and unreliable transportation all at once, that's a different story entirely. These layered, overlapping risks explain why two people with nearly identical clinical profiles can end up on wildly different cost and health trajectories. Once plans can identify these stacked risks, they can group the highest-risk individuals into cohorts and direct resources where they'll have the most impact.

This isn't just theory. Earlier this year, the Centers for Medicare and Medicaid Services released findings from its Accountable Health Communities Model, a program launched back in 2017 to connect beneficiaries with community resources tied to their specific needs. The results were striking: $200 million in net savings across more than a million beneficiaries, driven largely by fewer hospital stays and emergency room visits. Tellingly, the biggest savings came from members with multiple, complex needs, exactly the kind of population that individual-level, stacked-risk data is designed to find.

The third capability is the one most often skipped: closing the feedback loop to prove a program is actually working. Health plan finance teams frequently expect a return on investment within a set timeframe, but they rarely build in the verification step needed to confirm whether that return materialized. Without that check, there's no way to know if a program succeeded or failed. This is, more than anything else, why so many SDoH programs lose funding over time. It's also why plans have increasingly gravitated toward narrower, more measurable efforts, like "food is medicine" programs targeting specific food deserts, rather than sprawling initiatives trying to tackle everything at once.

Even plans that are ready and willing to act face two additional hurdles. One is simply getting the right data to the right intervention program, which requires interoperable systems and pre-established data connections, often called APIs, that allow different platforms to talk to each other. The other is measuring the true downstream impact on a member's health, costs and healthcare use over time. Skip that second step, and even a well-designed, well-targeted program has no way to demonstrate its own value, which makes it nearly impossible to secure continued funding.

Why It Matters

The real advantage of this three-part approach, pairing individual-level data with stacked-risk targeting and a genuine feedback loop, is that it replaces guesswork with confidence. Health plans of any size can know which members need help, understand why, direct resources accordingly, and then actually track whether those efforts moved the needle on health, cost and care use. That's the return on investment payers have been chasing for years but rarely managed to prove. Medicare Advantage plans stand to gain here too, since integrating SDoH data into care management can surface quality improvement opportunities tied directly to reimbursement.

But the people who stand to benefit most aren't the plans or the finance departments. It's the members themselves, the people dealing with food insecurity, unstable housing or transportation gaps who finally get the kind of individualized support that can genuinely change their quality of life. Getting the data right isn't a technical nicety. It's the difference between a program that looks good on paper and one that actually helps someone stay out of the emergency room.