Diagnosis is different than just search
If you are trying to diagnose a rare disease, plugging symptoms into Google will probably not be very helpful. In fact Google’s algorithm is actually biased against rare diseases. Because pages are weighted based on the number of times they are linked to by other heavily weighted pages, the diseases most likely to show up on search results pages are the diseases that are most common.
Wouldn’t it be great if there was a search engine just for rare diseases? There is. It’s called FindZebra.com The project grew out of a collaboration with the University Hospital in Copenhagen, Denmark, and was designed as an alternative to PubMed and Google searches which are too broad to be helpful in the specialized world of rare diseases. We did a blog post on the search engine when it launched last March. As part of our series on diagnosis, we called Ole Winther, PhD, an associate professor of DTU Informatics at the Technical University of Denmark and one of the founders of FindZebra, to get an update and see what new initiatives they’re planning.
How FindZebra works
Like Google, FindZebra uses a simple interface with a lot of data behind it. The landing page has a field with directions to enter symptoms. Advanced search functionalities allow you to perform much more sophisticated queries. For example, you can exclude genetic diseases from your search and focus specifically on rare diseases. Or you can search on the index that includes UMLS metadata associated to the title of each article.
The power of the tool comes from the databases it draws on. The search engine currently indexes over 31,000 medical articles focused on rare and genetic diseases from Orphanet, Wikipedia, NORD’s database, the Genetic and Rare Diseases Information Center (GARD), Online Mendelian Inheritance in Man (OMIM), m-Power Rare Disease Database, and half a dozen others. It uses the open source information retrieval tool Indri to search this index and suggest possible diagnoses, based on the match between the query and disease articles in the database.
FindZebra has already had quite an impact. There have been more than 60,000 visits, with 35% returning users. About 55% of visitors are from the US. “Most people are from US and English speaking countries now because the data is in English, but we have access to Orphanet and Wikipedia data. Eventually the goal is to collect everything in different languages as well,” Winther says.
The project has been featured in over 100 media outlets and they’ve been invited to present at many conferences and hospitals. It is also proving to be very effective for people that use the tool. The most striking case is a man from the UK who had been undiagnosed most of his life. After using the search engine, he went to his doctor with the results and received treatment that was effective immediately.
“I know that my colleagues in the medical world also use FindZebra as a tool for deciding what genomic test to run,” Winther says. “I sense a good deal of excitement.”
Patient and Physician Audiences for Find Zebra
When you visit FindZebra, you’ll see a banner across the top of the page that says, “WARNING! This is a research project to be used only be medical professionals.” But that doesn’t mean that patients can’t use and benefit from the site. It just means that physicians and researchers are the intended users.
“If you are not qualified to perform the search—which I feel that I am not—then you will definitely not get the full use of it,” Winther notes. “So even though I am completely behind patient empowerment and even though every physician can confirm that patients find their own diagnoses in some cases, I really hope to get more professional users.”
Winther’s most immediate goal is to improve the interface. “From a technical algorithmic point of view there are still a lot of things we can and will do. Diagnosis is different than just search. We want to develop novel technology to highlight relevant and discriminatory information while preserving the ease of use that we think is essential for attracting users.”
His team is working on specialized versions of FindZebra for specific disease categories which will be attractive to specialists who know roughly what they are dealing with. Work is also underway on a FindZebra API, so that the search functionality can be integrated into different sites, and a Find Zebra app.
Partners in the Rare Community
“One of the things that is very central to us for the continued development of FindZebra is to build alliances with rare disease patient interest groups. We see this as a way to raise people’s awareness of our existence and as a means to improve the information we provide by linking back to information provided by patient groups.”
Winther is also open to partnerships with pharmaceutical companies. “Initially I wanted to run FindZebra entirely as a research project, but due to a lot of public and commercial interest from pharmaceutical companies, the Technical University of Denmark and I decided to start a company commercializing FindZebra. “
“I see this as a way to fund making the system better and thus helping more people. We intend to make the commercial contents discrete and also making it clear what are sponsored links. This will hopefully mean that we will still be seen as a great entry point for accessing high quality information on rare diseases.”
“I think systems like ours are keys to improved diagnosis. The work of Mark Graber points to the fact that diagnostic errors often happen because of premature closure and looking at the disease in a wrong context. As Graber and his coworkers point out, suggesting alternatives to your current favorite hypothesis might be the best way to get you into a less biased state of mind.”
This post was co-authored by Stephanie McDonald. As Assistant to the President, Stephanie interfaces with all the departments at Siren. She has been with Siren for three years and has a background in marketing, logistics and business development. Stephanie has an MBA from the University of Illinois.