Medical Matchmakers: Startup ComplexDX Helps Specialists Find Hard-to-Diagnose Patients

By Gordy Slack

There are a lot of illnesses out there. Some of them are simple and straightforward, and others are bafflingly complex. The NIH estimates that of the approximately 20,000 diagnosable medical conditions, there are about 7,000 rare ones that are hard to identify. Even common diseases can present in ways that makes them tricky to diagnose. The National Organization for Rare Disorders says that of patients who were ultimately diagnosed with a rare disease, over 36 percent took more than a year to find a diagnosis. Seventeen percent took more than six years.

Brad Kittredge, a UC Berkeley MBA and Public Health student, has sought an explanation for his own hard-to-diagnose symptoms for four years, a journey that has brought him before nearly 20 doctors around the country. While he learned that his own condition is not too serious, the experience gave Kittredge insight into the challenges and frustrations of hard-to-diagnose patients. Last year, he teamed up with a physician and two computer programmers in search of a way to connect such patients with the doctors who are most likely to successfully diagnose their conditions.

Members of the ComplexDX project won second place in this year's Big Ideas competition for their project to give doctors both a mechanism and incentive to find the patients they can help..

Their new company, ComplexDX, is turning the traditional hunt for a diagnosis on its head. Rather than going from doctor to doctor, patients pay ComplexDX to post their cases to an online database, where participating doctors can examine and solve them.

The idea for ComplexDX first arose in the spring of 2008 during a class taught by Ravi Nemana, CITRIS Executive Director for Services and Health Care. “We realized that there was a huge need going unaddressed,” says Kittredge. “If we could just give doctors the tools and incentive to find patients whose symptom profiles fit their specialties, we could save a lot of suffering, time, and money for thousands of patients.”

At first, Kittredge and his partners called the business Hyoumanity—“to represent the idea of connecting one to many to solve these problems”—and emphasized giving medical specialists a powerful financial incentive to identify and solve tricky cases.  The group won a CITRIS Big Ideas Award for $8,000 in 2009. They have raised additional startup funds and plan to launch the business this fall.

ComplexDX will charge each patient $250-500 to have their case listed in the database and is aiming to pay doctors $1,000 for each correct diagnosis. The team is developing software that will help patients describe their symptoms and develop narratives that will allow the doctors reviewing them to find the telltale signatures of hard-to-diagnose illnesses.

“In addition to financial rewards, many doctors are interested in the sheer intellectual challenge of solving cases and pursuing their Hippocratic mission to help patients,” says Kittredge. “Solving cases also helps doctors build their reputations as experts on certain conditions.”

“These patients have already collected a medical record full of information about their cases,” says Kittredge. “Patients will be able to import and input that record into our system using Google Health, and then we will post it.”

The database will help doctors identify relevant cases and zero in on the ones that fall in their bailiwick. For example, a GI doctor who specializes in infant celiac disease would know that many of these patients lack any gastrointestinal symptoms, a fact leading many pediatricians to miss the diagnosis. This doctor could search our database for symptoms like weight loss, yellowed teeth, and mood changes. Other doctors may have considered Graves disease, growth hormone deficiency, adrenal failure, or diabetes, and may never have come up with celiac because it is usually associated with gastrointestinal distress. The patient may never find this specialist because they do not know to look there and their doctors would need to have had the celiac hypothesis to refer them there. 

After identifying interesting cases on which they can shed light, the ComplexDX-affiliated specialists will then suggest diagnoses and make suggestions about further tests or treatment options. But it will be the patients’ primary care physician who will follow up.

“We are limiting ourselves to providing diagnoses at this point,” says Kittredge, “and steering clear of the treatment business for now.”

For legal reasons, the patients and the ComplexDx doctors who review their cases will each be represented, in the database, only by numbers. Their identities remain private.  “The right diagnostic idea is much more important than the credentials or name of the doctor who has it,” says Kittredge. 

Jonathan Hicks, a computer scientist and Master’s candidate in the UC Berkeley School of Information, is the technical lead of the ComplexDX team. The main challenge, he says, is coordinating the standardized medical information in the patient records with a looser, more qualitative patient narrative.

“The trick is responsibly indexing and organizing it in such a way that a physician can look at it and get the necessary information,” says Hicks. Down the road, the project will also put to use the data they have aggregated about this chronically understudied group of patients.

About half a million people in the U.S. have undiagnosed or misdiagnosed diseases, estimates ComplexDX co-founder Elise Singer, a physician who, like Kittredge, is getting her MBA at Berkeley. That number includes the cumulative incidence of rare diseases as well as complicated and misleading presentations of more common ones, such as multiple sclerosis, Lyme disease, and celiac disease, she says.

“No one has studied these patients as a group before,” says Hicks. “It is hard to find any data on this phenomenon. If we can aggregate thousands of these tough cases that elude diagnosis, there will be a lot of value for basic research and clinical trials,” says Hicks.

As associations between the indexed postings and successful diagnoses accumulate, the system itself may be able to begin to flag likely diagnoses. “It is a learning system,” says Hicks. 

Often an accurate diagnosis can lead to quick and effective treatment, after years of fumbling around. “I know what it is like to have an ailment that evades diagnosis,” says Kittredge, “and if we can give relief to long-suffering patients, we can add real value in terms of medical costs, health outcomes, and quality of life.”