Sifting the Searches

Searching Google can provide information about many different things. But it turns out that examining Google searches themselves can also provide some important insights. For instance, having a deep understanding of the searches on its service helps Google and its advertisers make money.

But less cynically, Google searches can also provide information about the spread of disease, as with Google's Flu Trends service. Flu trends works in part by tracking online queries for influenza-related terms in order to estimate flu activity.

In this spirit, let's continue investigating global Google search activity for the term “ebola”, beginning with a look at search patterns in the United States.

Searchin’ USA

Figure [1] above shows that when an important Ebola-related event happened in the United States, Google search activity for Ebola spiked immediately or shortly thereafter (mouse over or tap on an event line to see its title).

Spikes were more dramatic for Ebola infections discovered in the United States than they were for aid workers returned to the US for treatment. Interest peaked after nurse Amber Vinson was diagnosed with Ebola on October 15.

Media criticisms, both foreign and domestic, of the American response to Ebola have ranged from calling it a “epic, epidemic overreaction” to labelling it as “disease-provoked hysteria”. But is this fair? How does it compare to the response in other nations? Before investigating this further, let's get a better sense of just how intense search activity for Ebola in the United States was.

Comparing Activity

The musician Taylor Swift released the album “1989” in America on October 27, 2014. It sold nearly 1.3 million copies with the biggest sales week for an album in over 10 years.

Americans celebrated the popular annual holiday Halloween on October 31.

Figure [2] plots search activity in the USA for the terms “Halloween”, “Taylor Swift” and “1989” against the search activity for Ebola.

Peak search activity was much higher for Ebola than for Taylor Swift or her album, but only half that of Halloween.

The nations with the consistently highest search activity for Ebola are the stricken nations in West Africa – Liberia, Sierra Leone, and Guinea. Figure [3] plots their search activity relative to that of the United States.

The Global Response

Let's resume our earlier question about the character of the response in other nations around the globe. In examining the data, the excited response of the United States to Ebola within its borders appears to be typical of how other nations reacted as well.

Spain

As of this writing, Spain is the only other non-West African nation to report an Ebola case contracted within its borders. On October 6, a Spanish nurse contracted the disease after treating a man brought to Spain for hospitalization after becoming sick with Ebola in Sierra Leone.

And on October 16, Spain had another Ebola scare when several patients who later tested negative were hospitalized with potential Ebola symptoms. Figure [4] illustrates these events.

Chile

The event in Spain on October 16 suggests that even the threat of Ebola within a nation's borders is enough to measurably raise search activity. Some nations experienced very dramatic surges around a suspected Ebola case that later proved not to be Ebola.

Chile experienced an Ebola scare on October 12th when a viral video spread around the internet. The video, taken at a hospital, features a voice on a loudspeaker saying, “Can I have your attention, please. We have a patient who is suspected to have Ebola. Please leave the room and go to another hospital”.

Sweden

Not all nations responsed equally to suspected Ebola cases.

In Sweden, patients were hospitalized for suspected Ebola on August 31 and October 19. These events produced a perceptable rise in interest, but at a smaller scale than in many other nations.

Good Neighbors

On March 31, shortly after Ebola was first detected in West Africa, Senegal closed its borders to neighboring Guinea. And in a moment noted for its irony, Mexico and Belize denied an American cruise ship permission to dock for fear of Ebola.

This pattern is reflected in the search data as well. When a nation confirms or is suspected to have a case of Ebola, search interest in neighboring countries spikes. Mexico and Canada's Ebola search activity, for example, closely mirrors the United States (figure [7]). Mexico diverged for a period shortly after August 6 when a student in Mexico City was suspected of Ebola.

And when a man was suspected of having Ebola on October 10th in the southern Brazilian state Paraná, which borders Paraguay, Ebola search activity in Paraguay actually eclipsed the activity in Brazil (figure [8]).

Looking at Policy

Governments around the world have issued a variety of different policies in attempting to keep Ebola out, or once it enters, to remove it. Over thirty nations have issued some form of travel restrictions against those arriving from ebola-stricken nations. Australia and Canada have banned the issuance of visas to travellers from Liberia, Guinea, and Sierra Leone.

A major focus of the policy debate in the United States has been on the use of mandatory quarantines imposed on aid workers returning from West Africa after Dr. Craig Spencer was diagnosed with Ebola. A central figure in this debate has been nurse Kaci Hickox, who protested quarantines imposed on her first at Newark airport, and then in Maine. Arguments for and against the quarantine policies play on the idea that they affect citizens' level of Ebola-related fear.

Figure [1] above showed that frightening Ebola-related events were soon followed by dramatic changes in search activity for Ebola.

Figure [9] overlays policy-related events on the graph of American Ebola search activity. It does not provide strong evidence to say whether Americans were assuaged, frightened, or even indifferent to the evolving quarantine policies.

An NBC/Wall Street Journal poll taken on October 30th stated that 71% of Americans favored mandatory quarantines for health care workers returning from treating Ebola in West Africa.

Policy in Australia

Australia has not yet had an Ebola case within its borders, but it has instituted tough restrictions against travellers from West Africa.

On September 28, Australia issued a policy to prevent its health care workers from assisting in Ebola outbreak zones. On October 27, it stopped issuing visas to residents of Liberia, Guinea, and Sierra Leone.

Figure [10] shows that Australia had consistently lower Ebola search activity than the United States.

Policy in Cuba

Cuba is unusual in both its Ebola search activity and its policy decisions. While other Caribbean island nations including Jamaica, Haiti, and the Dominican Republic have banned visitors from stricken West African countries, Cuba has sent “the largest foreign medical team from a single country” to fight Ebola in West Africa.

Figure [11] charts search activity for Ebola in Cuba against the United States. Search activity data was not available for the entire period under study, and activity is marked at 0 during empty periods. Ebola search activity data for Cuba was only available at one-week sampling intervals.

Nigerian Success

Nigeria's response to and containment of its Ebola outbreak has been widely praised. On October 20th a man arrived in Africa's most populous city – Lagos, Nigeria – with vomiting and diarrhea from an Ebola infection. On August 4, Nigeria confirmed several new infections among the health care workers who treated the sick man. The situation looked dire, but Nigeria managed to contain the outbreak with no new cases occuring after August 31.

In a statement on Nigeria's response, the World Health Organization credited Nigerian public awareness campaigns as an important contributor to success. “Interest spiked after health authorities announced the first confirmed case of Ebola infection of a Nigerian on Aug. 4—and the local media coverage exploded.” Figure [12] depicts an extreme spike in Nigerian Ebola search activity shortly after this August 4 date. I could find no single event to explain the resumption in search activity around the date of October 4.

Google Flu Trends, cited above as an example of the power of search data in predicting flu outbreaks, has actually been strikingly inaccurate in its estimates of the number of flu cases. To attempt to fix the problem, Google is incorporating into their service the official, traditional, and human-gathered CDC reports.

In West Africa, human-gathering and dissemination of information remains very difficult. As new admissions to ebola treatment centers drop, health care workers there say that it is too early to celebrate. Indeed, as far back as May, headlines were claiming success in combatting the outbreak when in fact the crisis would go on to take a dramatic turn for the worse.

Information travels fast these days, but good information can still be hard to find.

Data and Methods

Data about the spread of the ebola virus were gathered from the World Health Organization Global Alert and Response and Situation Report listings. Ebola case and death numbers reported above are the combined confirmed, suspected, and probable numbers issued by the CDC.

Data about ebola searches were gathered using Google Trends. All data were normalized against the United States, with the peak US search activity for the period under study fixed at a value of 100. For an explanation of the Google Trends data, visit the about page. Search data prior to July 30th was obtained at a weekly sampling interval and interpolated. Search data on and after July 30th was sampled at a daily interval.

The source code and raw data for this project are publicly available on Github