Using the novel approach of analyzing blog posts, Pomona College and USC researchers have discovered that female stroke patients report more mental status changes—disorientation, confusion or loss of consciousness—than do male patients.

Found in first-person blog accounts, this sex difference in reporting non-traditional stroke symptoms could help lead to better medical treatment for women who suffer strokes and demonstrates the Internet's value as a rich source of large-scale medical data, according to Pomona College Professor of Linguistics and Cognitive Science Deborah Burke, one of the authors of an article released today in the Journal of Medical Internet Research.

Burke points to evidence from some hospital studies that women stroke patients spend more time in triage than men pending a diagnosis, that their strokes are treated less aggressively then men's, and their outcomes are more negative than men's.

"One nontraditional symptom that we found women patients tended to report more was a change in mental state, like feeling confused. Because that description doesn't fit the traditional profile for a stroke, that can make the stroke harder to diagnose," says Burke.

The research team included Burke, Pomona College Professor of Computer Science Sara Sood, Sukjin Koh '12, Christopher Wienberg '10, USC Professor Andrew Gordon and Stephanie Morley CMC '12.

Gordon and Wienberg (who is a USC graduate student), with contributions from Sood, built a program called StoryUpgrade, which identifies personal stories from millions of posts and then developed a system to identify personal stories about specific health issues.  StoryUpgrade collects and indexes English language blog stories posted each day, coupled with user interfaces designed to support targeted searches of these collections. Stories were coded for first or third person narrator, traditional and non-traditional patient symptoms, type of stroke, patient sex and age, delay before seeking medical assistance, and delay at hospital and in treatment.

Blogs with first-person narrators (the patients) included more nontraditional symptoms during the stroke for both males and females. For the specific symptom of mental status changes— disorientation, confusion, or loss of consciousness—first person female narrators reported more such occurrences than did male narrators in their blogs. There was no such sex difference in reporting for third person narrators.

"This is important," says Burke, "because it suggests that the identity of the narrator affects the symptoms reported. This new finding has implications for hospital studies which do not typically analyze the effect of narrator. Whether the narrator is first or third person is an important variable to be examined in future studies because it affects what symptoms are reported and whether sex differences are found."

Burke says the new research replicates some previous results from hospital studies demonstrating the reliability of weblogs as a source of data about illness, at least regarding strokes. What's more, Burke says, the study identified an influential variable (narrator) that should be analyzed in future hospital studies: "Our findings recommend stories on the internet as a significant source of data on illness, with an advantage over hospital studies in terms of cost and speed of completion."

The idea for the project was born out of a conversation Burke had with Lewis Morgenstern (Pomona College Class of 1984) several years ago. Morgenstern, a neurologist and the director of the stroke program at The University of Michigan Medical School of School of Public Health, shared with Burke the problems that arise because women more than men report stroke symptoms that are not traditional stroke symptoms.

"Since treatment for stroke must be given within the first few hours after symptom onset, time is essential and anything that slows down rapid diagnosis and treatment can lead to worse outcomes. If someone presents with non-traditional symptoms it can distract from the real diagnosis of stroke and delay or confuse the diagnosis," says Morgenstern.

Sood, whose research is focused on the expression and impact of emotion in online communications, points to the tremendous rise of blogs at the turn of this century as a vast source of information, commentary, opinions, narrative and stories, posted by everyday people. Health-related blogs are a huge portion of personal blogs, largely as a chronicle of one's experiences with a disease, she says.

While Sood acknowledges that the use of blogs for medical studies seems unorthodox, "We hope to make it a commonplace, inexpensive way for one to test medical hypotheses before planning a full-scale medical study. This work will contribute to both artificial intelligence and medical research by making an untapped resource, the personal narratives of millions published in blogs, accessible for reliable analysis."

Wienberg was advised by Sood while a student at Pomona.

"It was really nice to be able to collaborate with Pomona faculty after I had graduated. It was nice to keep those connections open, even after I moved onto a new university. Additionally, this collaboration is likely to factor into my dissertation work, which is on learning about people from social media data," Wienberg says.

Koh was recruited by Burke to work on this project his junior year, and his senior thesis was based on this work. Now, that collaborative work is garnering attention at Oregon Health and Science University, where Koh is attending medical school—and one of his professors wants to apply these findings to a study of psoriasis.

"Looking back, it is incredible to see all that we were able to do starting from a simple idea," says Koh. "Both Professors Burke and Sood helped me learn how to design a study and carry it out from start to finish. It was truly proactive learning, an experience that I could not have gotten from any specific class at Pomona."