How Google searches can help determine population health

Dive Brief:

  • New research by The New York Times sorting every county in the United States by how "hard" they are to live in found that in hard places to live, people search for much different things on Google than those that live in "easier" places. In harder places to live, searchers focused in part on free medical care, medicine and basic health information.
  • For example, searches for diabetes, lupus, blood pressure, 1,500-calorie diets and "SSI disability" make the list of terms correlated with harder-to-live-in counties. This trend correlates with gaps in health and life expectancy between easier and more difficult places to live over the same period.
  • Meanwhile, in the easier-to-live-in counties, popular search terms included camera models, Baby Bjorns, "best cupcake," Apple devices like the iPod Nano and foreign vacation destinations such as Machu Picchu, New Zealand and Switzerland.

Dive Insight:

An important word on method: The terms isolated by the study were those that linked to the index researchers use to determine quality of life, which include factors such as income, education and life expectancy. The very top searches on Google, such as Oprah Winfrey or the Super Bowl, are popular nearly everywhere. 

It's also worth noting that the searches that correlated most strongly with the index were strikingly different from easy to hard places to live, with popular terms on one list virtually invisible on the other. For example, in the hardest places to live, health problems, weight loss diets, guns, video games and religion are all common search topics, but they rarely appear in the searches of those living in easier counties. The Times suggests that this analysis points to the rise of two very different Americas, and that signs suggest that the inequality hinted at here may continue to grow.

As the rise of ACOs and other value-based care models push providers towards better population health analysis, hospitals may benefit from taking a closer look at these disparities and how they can be addressed. Stakeholders across the industry have started to view non-traditional data—like geographic and socioeconomic data—as the pathway to risk assessment. For example, Duke Medicine recently announced that it will integrate geographic data into its electronic health record data to allow physicians to predict diagnoses in real time within a given population.

Brad Sitler, the principal industry consultant at the SAS Center of Health Analytics and Insights, told Healthcare Dive that creating a population database that is predictive of patient outcomes is dependent on leveraging nontraditional data: Socioeconomic data combined with EMR data and payer data. And Bill Davenhall, who is a health and human services expert, recently gave a TED Talk asserting that providing physicians with accurate geographic and environmental data on patients can significantly improve outcomes.

But "non-traditional data" may soon come to mean an amalgamation and analysis of digitally-sourced information as well. Remember, an online algorithm that analyzes tens of thousands of local news sites, social media outlets, government websites and infectious-disease physician networks was able to predict the Ebola outbreak nine days before traditional outlets. While the Times' analysis of Google search trends may be rough now, this kind of study is only going to be refined over time as researchers learn how best to organize and contextualize the information. 

Want to read more? You may enjoy this article that discusses how big data can be used to produce an accurate risk analysis of an ACO population

Filed Under: Health IT Payer Policy & Regulation Practice Management