For years, health care providers have struggled with the reality that they can do little within their roles to directly address the social factors often at the root of their patients’ health issues. They discharge a homeless patient back onto the winter streets, while fearing he may soon be wheeled into the ED with frostbite. They encourage a working mother with diabetes to eat healthier meals, while recognizing that her income level and limited local grocery options make that near impossible. They request multiple follow-up visits to adjust blood pressure medications, all the while knowing that the patient doesn’t own a car and isn’t near a bus line.
Last month marked what appears to be a major turning point in that struggle, one that alters how clinicians and health care organizations define their roles in addressing social determinants of health. In prepared remarks, Health and Human Services Secretary Alex Azar suggested that his agency may begin allowing some federal health programs to directly address social factors.
"What if we gave organizations more flexibility so they could pay a beneficiary's rent if they were in unstable housing, or make sure that a diabetic had access to, and could afford, nutritious food?" Azar said. "If that sounds like an exciting idea... I want you to stay tuned to what [the Center for Medicare and Medicaid Innovation] is up to."
By law, Medicaid cannot cover rent. While several types of waivers have been used to provide clinical services and housing-related initiatives, these have generally required that the beneficiary have either a mental or behavioral health condition that qualifies them for Medicaid services delivered within supportive housing, such as California’s 1915(i) waiver, or a clinical condition that requires skilled nursing care in a residential facility, as in Louisiana’s 1915(c) waiver. Simply being “in unstable housing” or requiring access to “nutritious food,” as Azar phrased it, has not historically been enough to qualify for support—despite the numerous studies demonstrating the impact such social factors have on overall health. It’s a tantalizing concept with the potential to affect the progression of an individual’s overall health and wellbeing, and prevent the need for more expensive services down the line.
To be clear, Azar wasn’t suggesting an Oprah-like “you get a home” giveaway of rent checks from the federal government. Rather, he suggested that by aligning the government’s health expenditures with investments in social determinants of health, we can improve patients’ outcomes and reduce overall spending. He also highlighted that value-based payment systems, in which providers are rewarded for improving outcomes and reducing costs, would be the ideal proving ground for this work.
Although we are waiting for a concrete proposal, Azar’s remarks are welcome news to those of us who have emphasized the importance of addressing social determinants in concert with health. We can no longer attempt to solve health problems in isolation from social problems and expect a positive long-term outcome, as I told Modern Healthcare.
Yet Azar’s remarks also hint at the challenges of “treating” social determinants. If we were to approach housing, healthy food or transportation on the same level as clinical interventions, how do we ensure that we deliver those services in the same evidence-based manner that we deliver medications and therapies today? How do we find the right individuals at the right time and deliver the “social treatments” in a way that will have a lasting impact? There’s little room for guesswork, particularly in programs that capitate reimbursements and put service providers at financial risk for overages. We’ve already seen Medicare tying its payments to measurable health outcomes; its logical their sibling’s programs would pursue such an avenue, as well.
At Evolent Health, we’ve been studying these challenging questions for the last seven years, trying to understand how work on social determinants can be operationalized. In our work with several provider-sponsored Medicaid plans that are stepping up their efforts to address social factors, we’ve identified three major challenges that health care organizations, communities and caregivers will face together.
Identifying Patients At Risk Of An Adverse Health Event Due To Social Determinants
Based on the surge of interest in social determinants, one might believe that our industry has mature data sources for identifying individuals who face housing instability, lack of access to food or transportation, or other social barriers. That’s simply not the case. What social data we do have are sparse and disorganized.
Working with our partners at Passport Health, a Medicaid plan serving 300,000 individuals in Kentucky, we set out to identify those patients experiencing homelessness or housing instability. Of those identified, just 11 percent had an ICD-10-CM Z-code for homelessness. We found others by matching their mailing address to HUD’s public housing database (63 percent), by identifying those with “couch surfing,” “place to place” or similar phrases in the address field of their membership eligibility files (24 percent) and by using natural language processing to scan unstructured notes in the electronic medical record (2 percent). Relying on any one of these sources would miss a sizable chunk of the population. Effectively integrating all of these sources was not easy, but it was necessary to provide the most comprehensive picture of those patients most at-risk.
Not surprisingly, these individuals have pronounced health care needs. Among Passport’s patients with high scores on Evolent’s Social Needs Index—a weighted measure that includes housing, transportation, food access, literacy and finances—monthly health care costs, based on internal analysis, were nearly twice that of those with low index scores. These social determinants must be combined with other clinical data, such as diagnoses, lab results, recent admissions and prescriptions, to develop a more complete picture of the patient’s risk for poor outcomes and high spending.
We’ve found that such data, when gathered in near real-time, can help us to identify patients before they incur lengthy hospitalizations, expensive visits and costly care. For example, 50 percent of the patients at the highest risk of a behavioral health adverse outcome, such as opioid misuse and dependency, were identified to have high social needs according to our index.
Integrating “Social Therapies” With Clinical Care
Once we can reliably identify those patients whose social circumstances increase their risk of poor health outcomes, we need to determine the best combination of approaches for treating their social and medical needs in tandem. Those “social therapies” must target specific health goals—they can’t stand alone as solutions in themselves.
Take, for instance, patients with transportation barriers that prevent them from making their appointments or going to the pharmacy to refill prescriptions. Through our analysis, we’ve found that within a specific population of high-risk patients, those with a transportation barrier had a 63 percent higher risk of readmission. However, merely offering rides will not reduce that risk; in one study, Medicaid patients who were offered free transportation missed appointments at the same rates as those that who did not receive the offer.
Such transportation initiatives have value, but only if they are part of a coordinated care plan in which a medical professional is helping to arrange those follow-up visits and follow the care regimens their physicians provide. Compliance isn’t always an option for a patient who lacks a reliable way of being contacted, or childcare or eldercare services to cover during an appointment. A patient who can’t reliably charge a phone, or who lives in a motel without a refrigerator or whose COPD is aggravated by an apartment full of mold, won’t benefit much from a transportation effort that begins and ends at the curb.
When we at Evolent addressed transportation barriers in a timely manner through a comprehensive care plan, our analysis showed that 30-day readmission rates were additionally reduced by 30 percent as compared to rates for patients where transportation needs were not addressed.
Similarly, a Medicaid program that pays for housing would need to be integrated with the care coordination network to ensure people who are lifted out of problematic housing situations are empowered and enabled to address the other overlapping and underlying factors impacting their health and wellbeing. A RAND Corporation study of homelessness in Los Angeles this year added to the body of evidence supporting the idea that housing, in conjunction with clinical and behavioral services, can reduce health care costs and utilization. In that study, participants spent 75 percent less time in the hospital and had 70 percent fewer ED visits in the year after moving into supportive housing.
Building “Social Decision Support”
With the spread of electronic medical records, physicians have come to accept clinical decision support. Patient-specific information is fed into a clinical intelligence engine that recommends evidence-based treatments to achieve the best clinical outcome.
What if we could create the social equivalent of clinical decision support—an intelligence engine that understands the relationship between social determinants and health outcomes, and then recommends targeted social therapies to the appropriate members of the care team, as part of a larger coordinated care plan? While a number of companies and applications have popped up to help connect those in need with social community services, the field is still waiting for an organization to use artificial intelligence to determine which services are most important for preventing hospitalizations, ED visits or other adverse outcomes.
At Evolent, we are piloting such a solution in January with one of our partner organizations that serves hundreds of thousands of Medicaid beneficiaries in a large, diverse metropolitan area. As part of this pilot, we will be taking fragmented social determinants data, transforming it and integrating it in a way that an AI algorithm can use it for social decision support. It has been a painstaking, complex task. Yet we see that this data, when harnessed, provides new insights that could help us deliver social support more effectively.
Let’s take the experience of an actual Medicaid patient, who we’ll call Bill. He lives alone within a census tract that has a median household income of $40,000. Together with his comorbid diseases—which qualify him for our Complex Care program—social criteria identified Bill as high risk for a hospitalization. Machine learning technology also determined that by providing transportation to appointments, we could reduce Bill’s risk of admission by 72 percent.
These are the kinds of actionable insights that providers need if they’re going to invest federal dollars wisely in addressing social determinants of health. A pipeline of government funding for social needs interventions could help to further mature the science and move these capabilities from pilot projects into the mainstream.
Taking a step back, it’s important to remember that the United States invests just 56 cents in social services for every dollar that is spent on health care. That pales in comparison to other countries in the Organization for Economic Cooperation and Development, where the average is $1.70. Clearly, this is a challenge that health systems, providers and health plans alone cannot overcome.
However, Medicare and Medicaid health plans and providers—the ones who will potentially be rewarded for effectively deploying social treatments—are in an excellent position to lead the way. Through their innovations, successes and failures, they can inform our broader health system and society on the best path forward.