If you want to stay healthy, it’s a good idea to avoid people who are sick. But little is known about humans’ skill at identifying people who are under the weather.
Now, however, a new study offers evidence that humans can detect whether someone is sick by looking at their faces. That new info may help reduce the spread of infection, and eventually help doctors better diagnose patients, researchers say.
The new study — published Tuesday in Proceedings of the Royal Society B — found that people were sometimes able to discern sick from healthy volunteers by looking at photos of their faces. The key determinants, researchers found, were paleness, swelling, and hanging eyelids.
But the judgements were far from perfect — 30 percent of photos of healthy people were deemed sick. The findings matter because past studies have shown that people might avoid others who appear to be sick.
“How we perceive other people’s faces is really, really important,” John Axelsson, a neuroscientist at Sweden’s Karolinska Institute and the study’s lead author, told STAT. “People who are attractive always get favored. People who are sick are being stigmatized in a sense.”
The researchers started by taking photos of 16 people when their faces looked “healthy.” From there, they injected those people with a shot of lipopolysaccharide, a molecule found on the surface of many pathogenic bacteria, which induced an immune response. Two hours after the injection, researchers took another round of photos to capture subjects’ visual symptoms of looking “acutely sick.” In both photos, the subjects wore plain white T-shirts and makeup, and looked straight into the camera with a neutral expression.
After that, two different groups each comprised of roughly 60 people were asked to look at the photos. The first group rated whether each person in the photo looked sick or healthy. The second group rated on a 1 to 7 scale how tired or alert each person looked, followed by answering questions about specific facial cues like paleness or redness of the skin. Observers could only look at each photo for a maximum of five seconds.
The researchers found that, for 13 of the 16 sick individuals, observers performed better than chance at detecting their illness. They also found that pale skin and hanging eyelids were the strongest predictors of apparent sickness.
Still, Mark Schaller, a psychology professor at the University of British Columbia who wasn’t involved in the study, pointed out that of the 1,215 times observers categorized a photo as “sick,” 440 were false alarms.
“They are only semi-accurate,” Schaller said. “It’s a useful reminder of the fact that when we humans use superficial characteristics as sickness cues … those superficial features often lead us astray, with the consequence that we may often respond to healthy people as though they are sick.”
Prior to this study, research related to sickness detection largely focused on animals, Axelsson said. Recent research suggests that acute inflammation-induced sickness is linked to changes in body odor and gait. But the latest findings, he said, suggest that the concept of sickness is connected to human preferences for attractiveness.
The study’s size and scope were limited. Axelssson said the study didn’t look at the visual symptoms of sickness past a few hours’ time. He also noted the background of the subjects — largely comprised of young white adults — meant the findings were not generalizable. Future studies should look at a more diverse group of people from different ethnic and age groups, as well as across a broader variety of diseases, he said.
“Is [detecting sickness] something you know genetically or learned?”Axelssson said. “It’s probably learned growing up around sick people, but needs to be studied.”
Despite the limitations, Axelsson believes the findings could have a broad clinical implications. He believes that being better able identify when someone is sick could help some doctors improve their ability to diagnose patients, and it could help improve technology used to monitor how sickness spreads across populations.