As population health initiatives grow, technology is looming as an enabler of far-ranging efforts to achieve results.

For example, the Baylor Scott & White Quality Alliance, an accountable care organization serving 275,000 patients in north Texas, recently received a grant to hire 30 community health workers to reach out to some of the 63,000 patients in its Medicare Shared Savings Program. The targeted patients are those the organization has identified, via their medical records and claims data, as being at the highest risk of needing inpatient care or other costly interventions.

The workers, members of the same communities as the patients, and closer to them in age than most of BSWQA's clinical employees, will connect them with social services that can help improve their health and reduce their risk. They'll also gather information on the patients' living situations, access to transportation, level of literacy, and other details that will help the ACO in its efforts to keep them as healthy as possible.

The alliance uses an analytics platform from Explorys, which has commercialized population health tools developed at the Cleveland Clinic. Explorys, acquired by IBM in April, currently works with 23 health systems that represent about 360 hospitals, and its database contains de-identified clinical and claims data on 55 million patients, which its clients can use to analyze their own patient populations and sort them into categories based on risk.

Rounding out the record

It's an excellent tool as far as it goes, but it doesn't know everything. Brandon Pope, director of analytics for BSWQA, hopes to use the community health workers to capture elusive pieces of information, such as patients' degree of frailty or risk of becoming malnourished. These are details that can't always be accurately assessed during an office visit or derived from a database.

"We know a lot of these things are underappreciated in their impact on patients' health," Pope says. He plans to have the workers enter social information into the organization's EHR systems, in discrete fields that can be added to the population health database and analyzed for their impact on overall risk.

Many determinants of health-clean air and water, access to nutritious food, a sturdy winter coat, dependable shelter, reliable heat, a place to exercise, transportation to doctor appointments-are things that healthcare providers traditionally don't control. In a fee-for-service environment, that lack of influence didn't matter particularly. If patients ended up riding an ambulance to the emergency room because they couldn't afford bus fare to get to the doctor, the hospital still got paid. In fact, it got paid more. In most cases there was no financial incentive to delve into a patient's social situation, although some providers would do so anyway as part of their mission or public obligation.

But as the focus of reimbursement changes toward value-based accountable care, and hospitals and health systems face financial consequences from avoidable admissions and readmissions, they need to deliberately look beyond their own walls to identify and address the social issues that affect the health of the patients covered by those risk-sharing arrangements.

For example, connecting patients with a food bank, or making sure they have a friend who checks on them, could nip potentially expensive health problems in the bud, or at least keep them from escalating to catastrophe.

Social challenges to health aren't always solely economic, like living in a food desert or not owning a car. Living alone is the single factor that best predicts both initial hospital admissions and readmissions, says A.G. Breitenstein, chief product officer for provider markets at analytics vendor Optum Health.

"We can get that data," she says. "FICO [the credit report database] knows how long you've been in your house, whether you live alone, whether you own a car. You have to contextualize it with the specifics of the patient."

Breitenstein cites a personal example. "My parents are 80 and on the precipice: their ability to walk has declined significantly," she says. "Their social data haven't changed, but their context has." Their particular combination of clinical, claims and social data should flag them as a fall risk, and a home visit would quickly confirm that danger. Buying and installing a few grab bars could be a wise investment for a healthcare provider trying to keep them out of the hospital.

Slow start

Many organizations are challenged in embracing a population health approach because they still work mostly in the fee-for-service world, where that approach doesn't make economic sense, says Frank Williams, CEO of Evolent Health, a population health analytics company that grew out of work at the University of Pittsburgh Medical Center and now works with providers in 25 markets, managing almost a million patients.

"A lot of organizations are doing pilots with payers and the government, but they need to move in a more integrated way to get to scale," he says. "Five thousand lives aren't going to move the dial."

Vanderbilt Health Affiliated Network is an Evolent client. It encompasses Vanderbilt Medical Center and a network of 50 hospitals and 3,500 physicians that operate independently but participate in the network's risk-sharing contracts.

VHAN is currently at risk for about 100,000 patients, mostly Vanderbilt employees and students, but Jason Grant, vice president of information technology, expects that number to grow dramatically in the next few years. When it does, he'll be ready.

Evolent's population health platform pulls data from all the EHR systems used by VHAN's providers and combines it with claims data from health plans. It then identifies patients most likely to develop costly health problems without intervention and puts them on a list to be contacted by the person who can best help them. That's usually a nurse care manager who's part of the patient's primary care team, but in some cases it's a pharmacist, social worker or behavioral health specialist.

Grant hopes to expand the system's capabilities to include social components-for example, whether a patient is homeless-and figure out ways to address those dire issues. "You can't just put them back on the street, but you might be able to put them up in a hotel if they don't need acute care," he says.

Just ask

Claims data and electronic health records are indispensable for stratifying populations into risk groups, but they can't tell the whole story.

Many analytics vendors, including Explorys, are busy incorporating census data, air quality data, real estate information, consumer credit scores, drivers' license data and information from other sources that can be used as surrogates to flag patients who may be in precarious financial circumstances or live far from a grocery store with a decent produce section.

But sometimes it's easiest just to ask.
The American Academy of Pediatrics, citing the dangers of hidden malnutrition for children's health and growth, recently issued a recommendation that its members routinely ask their patients whether they're hungry, as well as two other standard questions aimed at finding out whether they have a secure food supply.

The Boston-based not-for-profit organization Health Leads has had a lot of practice asking the right questions. Started in 1996 at Harvard as a service project, Health Leads trains student volunteers to connect patients with social services.

The organization has grown rapidly in recent years, fueled by providers' growing interest in understanding their patients' social needs, and it now operates in Massachusetts, New York, Maryland, Washington, D.C., and Berkeley, Calif. A dozen hospital partners include Massachusetts General in Boston and Bellevue Hospital in New York.

Health Leads has also hatched an analytics arm, headed by Sara Standish, who left international consulting firm Bain & Co. to pursue the opportunity. "This is a unique moment, where there is evidence that clearly links social and medical needs, and hospitals are under increasing pressure to understand how to address those needs holistically," she says.

That understanding begins with a simple questionnaire, part of the paperwork a patient fills out at a primary care visit. Standish says the questionnaire is tailored to the needs of a hospital's patient population and what the hospital wants to know.

Topics may include food insecurity, ability to pay for prescriptions, whether the patient lives in a safe house and has electricity and heat, whether a baby has enough diapers, and whether the family understands and has access to transportation to get to the hospital or to appointments.

The clinician talks the patients through the questionnaire and refers them to the Health Leads desk if necessary. There, patients provide further information and receive an action plan to connect them with help. "They walk out with a list of resources within a mile of their home or work to address their most urgent needs," Standish says. The volunteer follows up at least every 10 days, usually more often, to make sure things are working out.

The information also ends up in a database. "As we understand more about patient populations, we can more successfully segment and target patients who need a heavy versus a light touch," Standish says. "We track how many days it takes to fulfill a particular need, how many times we followed up with a patient, and how many needs they have."

Starting from scratch

With expansions in Medicaid coverage and patients newly insured under the Affordable Care Act, providers may increasingly find themselves taking on risk-based contracts for patients who have essentially no health history, says Mike Myers, executive vice president of solution delivery for AxisPoint Health.

"If we have access to pharmacy data and even a year of encounter claims, we can get close to risk stratification and predictive analytics," says Myers, whose company provides analytics and chronic care management services, including a nurse advice and triage line for all patients, to government payers, commercial insurers and providers.

"The problem is that so many of these populations have no history, and so you have to start from scratch," says Myers, who describes one client in the western states that hired AxisPoint to develop information on its post-expansion Medicaid population. Of almost 300,000 patients, 87 percent had no affiliation with a primary care provider.

For those mystery patients, AxisPoint's services include doing detailed assessments over the phone, gathering not only health histories but also information to create a picture of the patient's economic and social needs. The inquiry includes questions about access to transportation, presence of extended family or close friends in the household or nearby, and other factors that might affect the patient's physical and mental health or ability to access care. Interviewers are native speakers of the patient's language.

If the phone assessment reveals a potentially high-cost patient, especially one with mental health or substance abuse issues, he or she may get a visit in person from AxisPoint's assessment team. It's not necessarily a home visit, because some of the patients are homeless.

"The nurse will coordinate a meeting at a community health center, and then just hope the patient shows up," Myers says. "You can tell a lot from being in front of a person."


Here's one possible sequence for connecting patients with the social services they need and capturing the relevant data. As the patient's needs are met, the organization can measure the impact on his or her health, and can refine its procedures to better help future patients.

1. Patient fills out a social needs screening questionnaire along with other paperwork. Responses are included in patient's record.

2. Clinician reviews questionnaire, flags possible needs, and asks further questions during patient interview.

3. Clinician writes "prescription" for relevant social services, refers patient for follow-up with appropriate support staff (for example, care manager, social worker, or community health worker), and notes the referral in the patient's record.

4. Support staff conducts further assessment with patient, connects patient with appropriate resources, and notes the details for inclusion into the patient's record.

5. Support staff follows up with patient to make sure needs have been met, and makes further referrals and connections. Repeat as necessary.

6. Support staff updates clinic team prior to patient's next visit.

Register or login for access to this item and much more

All Health Data Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access