Some Apple a Day: How we can improve search engine diagnosesPosted: May 6, 2015
‘Achy neck’, ‘poor attention’, search. I’m Googling my current symptoms. I suspect it’s because my current office chair/monitor configuration isn’t optimal, but let’s see. First hit suggests “poor posture, working at a desk for too long without changing position”. Interesting.
Unless I have undiagnosed fibromyalgia or an impending attack of bacterial meningitis (the next two suggestions), this diagnosis was not too sensational. But Dr Google doesn’t always give such tempered results, no, they can often be both unreliable and disconcerting. The good doctor isn’t here to assuage our fears, and cyberchondria is a real and potentially dangerous thing — especially when patients misdiagnose themselves, leading to mistaken self-medication.
Our researchers have teamed up with the Queensland University of Technology and the Vienna University of Technology, Austria to improve search engine processes to help individuals find more reliable online health advice, and to help health professionals treat patients more effectively.
“We’ve been looking at the use of search engines by everyday people to find medical advice online,” said Dr Bevan Koopman, one of our applied health informatics researchers. “People are using search engines all the time and there is a lot of information about illness and disease.
“This is easy when you’re searching for conditions that you already know, that are clear and simple. But it’s a bit harder when things are less well described and you have to tease out what particular symptoms you have, or just general health information. These might not be that well expressed, and so it’s difficult to return good results from a search engine.
“There are two parts to our research: one is evaluating how people are using systems and how effective they are, [and the second] is being done in order to develop new systems.”
The team did a preliminary evaluation of popular search engines, like Google and Bing, using participants and sample queries.
The results showed only 3 out of the first 10 results to be helpful in diagnosing the sample diseases, and only half of those 10 were partially relevant. Many results led to rarer, more fatal cyberchondria-inducing diagnoses, a phenomena which occurs because of the inherent mechanics of search engines, the researchers suggest.
“Our research goes beyond these results” said Bevan, “we’re also developing new methods for search engines to improve the way we can search for health information. For example, methods that take into account how readable the results are.”
The team has developed algorithms for search engines to determine which results are appropriate for the self-diagnosing searcher. The algorithms can determine if a page is jargon-heavy, it’s likely intended for an expert and not layperson.
In the longer term, the team aims to connect the ‘patient’ not only with appropriate webpages but with appropriate health services.
The research also has applications for healthcare professionals.
“We’re doing a lot of research into search engines to help doctors search patient records more effectively.
“For example, we’ve developed a system that allows radiologists to quickly look up common conditions,” said Bevan.
Through clinical partnerships we are delivering technology that improves the safety, quality, and efficiency of health services to create a better health system for all Australians. Find out more about our digital health research.
Right, finished the article. Maybe it’s time for a stretch and walk around the block.