How data provides vital insight into the social determinants of health

May 15, 2019 3:37 pm

As providers take on more risk for managing the health of patient populations, many hope to gain an edge by applying data analytics to the social determinants of health. They recognize that social factors such as lack of housing, food insecurity and domestic violence can determine the success of a patient’s treatment and whether the organization wins or loses in a risk-sharing contract.

Some providers use sophisticated analytics programs with natural language processing algorithms that can “read” through the unstructured data in an electronic health record (EHR) to extract terms or concepts related to social determinants of health, says Douglas Fridsma, MD, PhD, president and CEO of the American Medical Informatics Association (AMIA).

Another common strategy is to “link” social determinants to patients through techniques like geocoding, which associates socioeconomic factors (e.g., average income, average housing prices, access to transportation and grocery stores) with specific ZIP codes. Providers can then merge neighborhood data with EHR data and public data sets to create a multidimensional view of a patient, Fridsma says.

“A lot of data analytics is about pattern matching: trying to find patterns in the data to identify patients at risk of having a bad outcome and patients who have had a good outcome to understand the process of care that potentially led to that outcome,” Fridsma says. “If people got the same care, adding in social determinants often can give you a better sense of what makes those two groups of patients different.”

Using data to bridge health and community

At Providence St. Joseph Health, leaders are using analytics to identify social conditions that impact patients’ care, access to care and overall health status, says Rhonda Medows, MD (pictured at right), president of population health management. Leaders have found that the four social factors most likely to correlate to a future medical event are:

  • Homelessness
  • Mental illness
  • Substance abuse
  • Domestic violence

As part of this strategy, leaders leverage patient-level EHR data as well as public-health mortality and morbidity data and screening and prevention data. These sources feed into an integrated data platform that was launched in October 2017. The platform also incorporates “hot spot” analysis to identify specific ZIP codes where factors like employment rates, education levels, income and access to food and mental health centers can contribute to poor health outcomes. 

So far, leaders at Providence St. Joseph Health have focused their analysis on Medicaid patients and dual-eligibles who are high-utilizers and have certain clinical factors, such as active substance abuse, a mental health diagnosis or multiple chronic conditions. 

If patients with high social needs have a primary care “home,” a specially trained care coordinator in the physician practice will follow up with the patient and set up referrals to community partners that can address some of the social issues. If the patient does not have a primary care home, a centralized care manager will reach out to address the individual’s healthcare and social needs and connect her or him with primary care.

Comparing the approach to traditional targeting strategies

Medows and her team plan to compare results from using the analytics platform to the traditional method of using claims data and some clinical and social data to identify patients in need of care management. The comparison will look at how well each approach reduces unwanted utilization, such as avoidable emergency department visits and readmissions, and the associated costs. 

Medows believes the analytics platform can identify the impact of social determinants of health more accurately compared with more-traditional care management approaches. “We will identify more people in need, intervene earlier, and partner with our community and social service agencies well in advance and address issues that usually impede or impair their ability to get care,” she says.

In addition to improving patients’ care management and care coordination, utilizing analytics should help leaders better allocate resources at the practice level. “If a practice has a lot of patients with diabetes, maybe we can have stronger a partnership with a food bank that can provide more nutritional food or a transportation service that can help patients get to their appointment for nutrition counseling,” Medows says.

Matching solutions to needs

While organizations such as Providence St. Joseph Health are using their own analytics tools to identify social determinants of health, others may choose to work with a vendor. In such cases, Fridsma recommends inquiring about the specific data sets used in the platform to ensure the age and sex distribution, socioeconomic mix and other attributes closely match the provider’s own population. If the populations are not well matched, the analytics platform could create incorrect associations or miss subtler associations, he says.

Leaders also should have a clear understanding of the specific data standards and definitions used. “Black boxes are hard to describe, and they are hard to trust,” he says. “Transparency and explainability are important so that you know the algorithms are doing the right thing.”

Read more

For more examples of how hospitals are using analytics to address social determinants of health, see:

How artificial intelligence could help uncover social determinants of health

Testing a model for better coordination of care in the community

Interviewed for this article:

Douglas Fridsma, MD, PhD, FACP, FACMI, president and CEO, American Medical Informatics Association (AMIA), Bethesda, Md.

 Rhonda Medows, MD, president of population health management, Providence St. Joseph Health, Renton, Wash.


googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text1' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text2' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text3' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text4' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text5' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text6' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text7' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-leaderboard' ); } );