3 use cases for public search data in the health sector

Search can be one of the most informative sources of publicly available data. If harnessed correctly this data could provide the health sector with a plethora of predictive insights.

Recently we saw how query data from search engine Baidu was being used to determine forward-looking microeconomic indicators in China. We also saw how search data can be used to predict rising consumer trends, like the rise of ‘Pumpkin Spice’.  

More than ever, with advanced AI, we’re able to analyze massive amounts of this type of data and identify more specific rising trends we could never before evaluate. With access to these insights comes new possibilities, especially when it comes to the healthcare space.

According to the Google Knowledge Graph, 1 in 20 Google searches are for health-related information. As a result, in 2015 Google started displaying medical facts next to relevant health searches in an effort to offer verified results for customers. Each of these facts has been checked by medical doctors at Google and the Mayo Clinic for accuracy.

As long as a Google search remains more accessible than a visit to a doctor, there’s a good chance most consumers will continue to consult their search bar leading up to a conversation with a professional. If analyzed in aggregate, this search data can be extremely helpful for the professionals looking to better serve this population.

Here are a few use cases that speak to ongoing efforts to leverage search data analysis in the healthcare and research space.

Mapping information: using search data to understand health needs in Africa

Analysts like Rediet Abebe, an advisor to the National Institutes of Health, are working to demonstrate how AI and machine learning can be “better integrated into biomedical and clinical research,” according to Venturebeat.

Abebe, along with researchers from Cornell University, SUNY Stonybrook, Microsoft Research, and the Rockefeller Foundation Fellowship, recently published a report in which they analyzed 18 months of Bing search results across all 54 African nations to assess queries related to HIV/AIDS, malaria, and tuberculosis. AI then categorized words and topics from these searches that are associated with specific conditions or diseases. It was able to expose patterns in health information needs by demographic groups and country.

According to the report, “Search results demonstrated that women and users aged 18-24 are more concerned about stigma than other groups, and natural cure searches were highest in the 35-49 age group.”

However, a problem they uncovered was that often the web pages users found when searching for natural cures had “serious issues with accuracy, effectiveness, and relevance,” Abebe said. They also discovered a correlation between the rate of stigma-related searches and high rates of HIV, as well as between general searches around a disease and high incidence of that disease.

Source: https://arxiv.org/pdf/1806.05740.pdf

While this data is helpful due to its open-ended nature, which provides insights into the mindset of those performing searches, Abebe and others caution that it is high level and should not be used in itself to make predictive decisions. According to the study, the real use case here is to “uncover and gain insight into health information needs in Africa.” In the end, the study suggested that “search data can help…inform discussions on health policy and targeted education efforts both on- and offline.”

The authors suggest such anonymized search engine data could be used as a greater platform for public health data.

Tracking the impact of vaccines

In another use case, access to search and social media data can help researchers understand the effectiveness of a targeted health campaign, such as the introduction of a vaccine in a particular area.

In 2015, Vasileios Lampos, Elad Yom-Tov, Richard Pebody and Ingemar J. Cox published a study that looked at the influenza vaccination program launched in England during the 2013/14 season. They observed millions of geo-located search queries within the Bing search engine and posts on Twitter.

Naturally, people were far more likely to search terms related to the flu and flu-like symptoms when they or their children were experiencing them. Taken in aggregate this data was indicative of areas where flu was in fact spreading. The authors of the 2015 study were able to compare the results in this region after distributing the vaccine to a previous control study to determine whether searches and social posts related to flu had signified the prevalence of flu itself, and through the act of distributing vaccines the flu had in fact decreased.

Related studies that look to predict the spread of flu, such as Google’s Flu Trends project, have in some instances been debunked as ineffective methods of tracking the spread of this type of illness. According to a deep dive by WIRED, however, often these cases have less to do with the potential value of the data itself, but rather stressed the importance of developing air-tight algorithms that enable researchers to use the data correctly and effectively.

Predicting a visit to the Emergency room

A new study by researchers at Penn Medicine found that “in the week before an emergency room visit, patients conduct nearly three times as many health-related searches, or 16 percent of their total.”

This has led them to believe that people research their health conditions extensively online just before they determine whether it’s necessary to visit the emergency room. Insights like this can help hospitals better prepare for the demands of incoming patients.

As well, this data can reveal where there might be information gaps between what doctors have told their patients and the way their conditions developed. For instance, “one patient searched, ‘how big is a walnut?’ and ‘what is a fibrous tumor?’ The researchers said that the patient had been informed by a doctor that she had a fibrous tumor the size of a walnut. She apparently didn’t understand her condition, and she had to turn to Google to look it up.”

These use cases highlight the incredible potential that can be unleashed when search data is harnessed effectively, giving us a micro glimpse into the public’s mindset, fears, concerns, curiosities and more. Access to this vast data set has opened medical research to far larger sample sizes than ever before. The healthcare space is primed for greater impact from AI technologies, and the introduction of alternative datasets like public search is only the beginning.

Recent Articles