Syphilis has made a comeback in the U.S., and health researchers are using Google searches and tweets in an effort to predict the next outbreak.
After decades of steady decline, infection rates for the sexually transmitted disease started to rise after hitting a historic low in 2000 — from 2.2 cases per 100,000 people to 8.7 cases per 100,000 in 2015 and 2016, according to the Centers for Disease Control and Prevention.
Researchers at the University of California at Los Angeles working with the CDC have developed an artificial intelligence program that can identify with high accuracy where and when a syphilis outbreak is likely to occur based on Google search terms and social media comments related to risky sexual behavior.
Lead researcher Sean Young, executive director of the University of California Institute for Prediction Technology, said the technology will allow public health officials to respond immediately to crises and eliminate a lag time of up to five years between an outbreak and its detection.
“Oftentimes, once the CDC finds out about disease outbreaks, there have been so many cases that have already been spread and transmitted that it’s just become a disaster,” Mr. Young told The Washington Times. “So if we can get ahead of the curve, that would really help.”
In creating the artificial intelligence program, researchers relied on syphilis incidence reports that hospitals and medical providers across the country have been required to submit to health officials since the 1940s.
They fed the program data on syphilis cases reported from 2012 to 2014 and had it compare this data with search terms and tweets referencing risky sexual activity during the same period. The computer program “learned” to detect patterns in social media posting and search queries as indicative of an STD outbreak, pinpointing vulnerable areas at the county and state levels.
The Google search portion of the research was published last week in the journal Epidemiology, and the Twitter portion is in the April issue of Preventive Medicine.
To protect the privacy of study participants, the researchers didn’t identify the specific keywords they mined on Google but generally described search queries including “sex,” “STD help,” “sex without a condom,” “do I have an STD,” “symptoms of STD’s” and “how to find sex right now.”
In the Twitter study, the researchers identified more than 8,500 tweets, looking for “colloquial terms for intercourse” that were often crude mentions of genitals or sex.
“They will tell you that they are drunk driving, they’ll tell you the types of drugs they’re using, they’ll tell you who they’re having sex with,” Mr. Young said of Twitter users his team observed.
“People are very comfortable putting anything and everything out there, and they don’t really question it. … It’s become a social norm that ‘I can and do share,’ and that’s the way, especially younger people, connect with each other,” he added.
Syphilis manifests in four stages: first with genital sores, then with a skin rash, swollen lymph nodes or fever. A period of no symptoms can last from a few weeks to several years. But if the infection is untreated, it can cause severe problems in the heart, brain or other organs. When caught early, syphilis is treated with penicillin.
The researchers decided to test that the artificial intelligence had accurately identified a pattern between the syphilis reports and the internet data. They selected a time frame for which they had data about syphilis cases and of which the program was not aware. They then fed it 144 weeks’ worth of social media comments and found that the program had a 90 percent accuracy rate in identifying where and when syphilis cases occurred.
“It’s not just identifying some association,” Mr. Young said. “[We asked], can it be used to predict the future? And found that it was able to.”
Scouring user data to predict an adverse event has science-fiction overtones, and Mr. Young said it’s hard to discount comparisons to “Minority Report,” the 2002 action film in which people are arrested before they can commit crimes.
“It’s definitely Big Brother-ish … but we’re living in a time where everything is Big Brother-ish,” he said.
Mr. Young, who is also an associate professor of family medicine at UCLA, said his team is aware of the delicate balance between helping the public and invasion of privacy. Follow-up interviews with research subjects showed that they largely approved of using their data for public health purposes, he said.
“Generally… people are saying, ‘Companies are already monitoring everything we do, so if researchers and public health officials can apply these same methods — but to promote public health and social good — then we support you,’” he said.