Rutgers computer science researcher to create search tool for online patient-written content
Everyone gets tired from time to time, but Heather Reiner was experiencing what she describes as crushing fatigue, along with headaches, memory problems, and back pain. “We didn’t know what was wrong with me,” says the East Brunswick woman, who was 23 and a Rutgers graduate student when her symptoms began in the summer of 2008.
After consulting with two physicians who could not come up with a definitive diagnosis, Heather felt discouraged. “They were kind of giving me the runaround,” she says, “so I started searching online for what disease matched my symptoms.”
Her search brought her to lymenet.org, a nonprofit foundation’s website focused on the prevention and treatment of Lyme disease. Through online discussions and support groups, current and former patients use the site to share talk about their symptoms, medications, experiences, physicians, and more. Reiner gained medical information, made online connections with other patients, and even located a doctor, recommended by other site users.
Reiner’s decision to go online and learn from other patients’ reports is an increasingly common occurrence, according to Amélie Marian, assistant professor in the Computer Science Department, School of Arts and Sciences. Where once a patient might have heard advice about a health condition only from a friend, neighbor or relative who had a similar diagnosis now it’s possible to find hundreds of tips and opinions with just a few mouse clicks.
In their search for information, patients are going beyond traditional medical websites from health care institutions and turning to online health communities where others like themselves write and post content on forums, blogs, and mailing lists. Such messages are read for many reasons, from understanding a diagnosis and gaining emotional support to decision-making about treatment or ongoing care.
Marian is a principal investigator of a new research study, funded by the National Science Foundation, that may help change that situation. The study, “Gaining Knowledge from Other Patients: Structuring and Searching the Content of Health-Related Web Posts,” will tackle the first steps of analyzing online patient-generated content and creating tools to improve searches.
Physicians and other health professionals don’t always feel comfortable with the quality of information and answers patients gain from each other, but “it seems they don’t have much of a choice. Patients are going online anyway,” Marian adds.
Using anonymous data from patient-authored material on a breast cancer organization’s website, Marian and a Rutgers computer science doctoral student, Gayatree Ganu, will be working on the components for building a search, scoring data, and developing an intelligent search engine for posts.
A collaborator from the Department of Biomedical Informatics at Columbia University, Noémie Elhadad, will guide the annotation and evaluation of data as well as the automatic processing of the text using natural language processing techniques. “On our side, it’s more computer-science based –designing a basic search interface with a little more functionality,” Marian says.
By analyzing the text and adding structural information, the project can make searching in forum postings easier and more effective than a simple keyword search.
“You could navigate through the currently very shallow structure so you can have more relevant information,” says Marian. “It’s hard to get this information in real life.”
The researchers envision that the tool they build would permit advanced searches by topic as well as by the writer’s emotional expression. The search tool also could be used in conjunction with expert-written health portals, which generally do not include personal patient experiences.
Online health communities contain a staggering number of postings. In just the website the researchers will be looking at, there are 50,000 registered members who may contribute to 55 different forums – producing about 800,000 posts on 37,000 threads.
Marian believes that providing better access to the information contained in those postings through improved and flexible search tools will serve several purposes. It will help patients find what they’re looking for more quickly, aid researchers who want to study the effects of online communities on health behaviors and provide health care professionals with greater insight into patient needs and emotions.
For Heather Reiner, the personal quality of patient-written postings helped her during the difficult time when she was suffering so severely that she had to take a year off from school. Now receiving treatment that she says has helped, she’s a full-time Psy.D. student in the Graduate School of Applied and Professional Psychology at Rutgers.
“I don’t go on it (the website) as frequently as before,” Heather says. “But if I’m having a bad day, I’ll go and read the success stories.”