Language in Facebook
posts may be able to predict whether someone will develop diabetes and other conditions including depression, anxiety, alcohol abuse, sexually-transmitted diseases, and drug abuse better than demographic information like age, sex, and race.
People who often use the words “God” and “pray” in their Facebook posts are 15 times more likely to develop Type 2 diabetes than people who rarely use those terms on the platform, a new study from the University of Pennsylvania School of Medicine finds.
Type 2 diabetes, also known as adult-onset diabetes, occurs when the body develops a resistance to insulin, leading to high blood-sugar levels. Type 1 diabetes, or juvenile diabetes, occurs when the body stops or dramatically slows insulin production.
Merchant’s team of researchers from Penn Medicine and Stony Brook University collected old Facebook posts from 999 adults who agreed to be part of the study. More than three-quarters — 76% — of study participants were women, 71% were black, and 70% were 30 years old or younger.
Black Americans have a higher rate of Type 2 diabetes, but the study compared those who used the words “pray” and “God” in their posts versus those who did not. The former were more likely to have Type 2 diabetes.
“This is a population that traditionally isn’t studied in early research, so we are glad to have studied this group,” Merchant said.
It’s also a persistent and growing problem: Nearly half of the U.S. population will be obese by 2030, according to one study published by the peer-reviewed journal, the American Journal of Preventive Medicine.
What is behind the connection between prayer and obesity?
Raina Merchant, the lead author of the study and the director of Penn Medicine’s Center for Digital Health, said she didn’t know exactly why “God” and “pray” were linked to diabetes.
However, a 2011 study from Northwestern University found that those who begin regularly attending religious services while young are more likely to become obese by the middle of their lives. (Obesity is one of the main factors that can lead to Type 2 diabetes, and both are considered contributors to early mortality.)
There is a large archive of research that suggests church-going Americans are more likely to be obese, particularly among women. One theory: The researchers in the Northwestern study pointed to high-calorie meals often present at religious celebrations as one potential reason behind the results.
“African Americans and persons with limited education are more likely to be obese and are also more likely to be affiliated with Baptist or fundamentalist religious groups,” according to one 2006 study published in peer-reviewed Journal for the Scientific Study of Religion.
Another possible explanation: More socially conservative people and those who are very religious may find food as a source of “earthly pleasure.”
In his book, “Whitebread Protestants,” Daniel Sack wrote, “When I was young, church meant food. Decades later, it’s hard to point to particular events, but there are lots of tastes, smells, and memories such as the taste of dry cookies and punch from coffee hour-or that strange orange drink from vacation Bible school.”
The authors of the Journal for the Scientific Study of Religion study add, “Many religions in the U.S. place priority on constraining sins such as smoking, excessive alcohol consumption, and sexual promiscuity. Gluttony does not receive the same level of pastoral or congregational condemnation in most denominations, perhaps indirectly creating an ‘accepted vice.’”
However, they added that those with lower incomes may also have less access to healthier foods. “The relationship between religion and social stratification is important to consider while interpreting these finding. Indeed, it may be argued that social inequality and racism are the more antecedent causes of the higher incidence of obesity among those most religiously inclined.”
They said two questions remain unanswered: “Is religion related to the prevalence of obesity?” Or, “Is religion related to the incidence of obesity?” Answers to these questions will help identify whether religion is a “risk or protective factor” in understanding the growing incidence of obesity in the U.S., they added.
Could Facebook predict depression, drugs and alcohol abuse?
Perhaps less surprisingly, the words “drink” and “bottle” predict alcohol abuse, and expletives suggest drug abuse. “Social media posts are often about someone’s lifestyle choices and experiences or how they’re feeling,” Merchant wrote.
Merchant’s team conducted a similar study last year and found that Facebook posts could predict a diagnosis of depression three months before the diagnosis at a clinic.
“It’s not surprising that someone who is depressed may be posting about their mood,” Merchant told MarketWatch in an interview. “But we have less knowledge of the language of, say, diabetes, so it’s something we wanted to look into.”
Merchant is hopeful that social-media posts could one day help doctors diagnose diseases like diabetes early or prevent them altogether, but there’s still more research to do before your doctor begins analyzing your status updates. Merchant plans to conduct a large study later this year that shares social-media information directly with health providers.
She’s also hoping to find ways to summarize the social-media data. “There is so much data and we, as providers, aren’t trained to interpret it ourselves,” she wrote.
“I don’t foresee physicians ever looking through someone’s social media. I do think, however, an interesting area for next steps is to think about how we can synthesize this information and then ask the patient if they want to share that with their physician,” she said.
Analyzing social-media posts presents obvious privacy concerns, but the ability to predict or prevent diabetes could save Americans a lot of money.
Americans with diagnosed diabetes spend an average of $7,900 each year to treat the condition. And adjusted for age and sex, people with diabetes spend 2.3 times more on health care than people without the illness. In 2012, the total direct and indirect costs of medical expenditures on diagnosed diabetes in the U.S. amounted to $245 billion, the Centers for Disease Control and Prevention estimated.
In the latest study, the researchers reviewed 949,530 Facebook status updates from between March 2009 and October 2015 from these participants. They then analyzed the data they gathered from these posts against information on 21 health conditions.
Looking at the patients’ current medical records, they found the data to be predictive of all of these conditions and more predictive than demographic information for 10 illnesses.
For those worried about privacy in the latest report, Merchant says it’s a top priority. “We made it very easy for patients to decide they no longer wanted to participate anymore, and we didn’t look at any data from their friends. This would be an opt-in process, and privacy needs to be part of the conversation,” she said.
Shares of Facebook are up 44.5% year-to-date, compared to a rise of 15.3% in the S&P 500
and an increase of 13% in the Dow Jones industrial Average