More on Analytics

Using zip codes to find at-risk patients to determine the social determinants of health

Why don't patients do what they know is good to stay healthy? The answer often comes down to the social determinants, expert says.

Susan Morse, Managing Editor

As providers and payers do more to incorporate the social determinants of health into their care management plans, there has been an increased interest in determining the populations at risk and in need.

Dr. David Nace, CMO for Innovaccer, who also held that title for UnitedHealth Group and Aetna, formerly worked in family medicine. He wanted to know then, he said, why patients didn't do what they knew to do to stay healthy.

"'I should walk more, eat better,'" they told him. "When I asked why they didn't, they looked perplexed. I went back to do a residency in psychiatry. The problem wasn't just behavior, it was their environmental conditions."

HIMSS20 Digital

Learn on-demand, earn credit, find products and solutions. Get Started >>

The Centers for Disease Control and Prevention has an algorithm to estimate the social vulnerability index based on every census-tract in the U.S., Nace said, but this often results in an over-estimation of social vulnerability.

Nace decided to apply machine learning to the algorithms to determine the social vulnerability at the zip code-level of the population.

There was a lot of data, he said, and it was taking a long time to move the data around, yet he saw that Google and Amazon used massive amounts of data and made it look easy.

"I spent time with engineers with Amazon and Google, finding out why we're using technologies that are antiquated," Nace said.

The result was to build a new data platform that can determine over 55 factors of the social determinants of health, drilled down to the individual county level, or zip code.

Where the CDC used 15 social and economic factors, Nace and researchers looked at a broader set of risk factors and did a regression analysis. Innovaccer then used Google maps at the county level to pull out healthcare data such as rehospitalization rates and HEDIS scores.

"What we found was our new index was validated," he said. "Then we did the same process down to the level of the same zip code."


Providers and payers working in regions of vulnerable populations need to address the social determinants of health to improve care outcomes and reduce cost.


Work such as what is being done by Nace goes beyond the data warehouses for actionable data. Apps on smartphone or a computer in a waiting room can be used, for example, for better patient generated data than what can be found in the EHR.

"Once we have that all together, we can then help to provide specific information to a care manager."


"Each person is different, each county level zip code is different," Nace said. 

Nace and colleagues released a research paper they were discussing during HIMSS19, "From Myth to Reality - Revolutionizing Healthcare with Augmented Intelligence and Social Determinants of Health."

Other researchers included Dr. Vibhuti Agrawal, director of Analytics at Innovaccer; Dr. David Nash, founding Dean, Jefferson College of Population Health; Dr. Glenn Steele Jr., vice chair of Health Transformation Alliance and former Geisinger Health System President, and CEO; and Dr. Stephen Klasko president and CEO, Thomas Jefferson University and Jefferson Health.

Twitter: @SusanJMorse
Email the writer: