Skip to main content
Premium Trial:

Request an Annual Quote

Med Data Quest Prepares to Launch NLP, Modeling Software to Improve Disease Diagnosis

Premium

NEW YORK (GenomeWeb) – San Diego-based Med Data Quest is hoping to build a business around providing natural language processing and statistical modeling tools that can extract and use information from electronic medical records to help physicians accurately diagnose diseases and suggest more effective treatments.

Yuan Gao, president and CEO of MDQ, told BioInform this week that the firm hopes to launch a first version of its platform later this year. The software will be available as a service and MDQ is still discussing commercialization and pricing models, contemplating options such as revenue-sharing models where the software is made available for free and users only pay if the system helps them generate revenue, he said.

Gao formed MDQ and began developing its platform in 2009 with co-founder Feng Hu, who worked for Google prior to joining the company. Gao and Hu launched the company with an undisclosed amount of innovation funding provided by Kaiser Permanente, which has also provided office space for the company. Gao also has a position in Johns Hopkins Department of Biomedical Engineering. He took a leave of absence from his post there last year to focus on developing, validating, and launching MDQ's platform. Elliot McVeigh, who is the chair of Hopkins' BME department, serves on MDQ's scientific board.

Kaiser Permanente is collaborating with MDQ to develop and test its software solution, which is internally referred to as the MDQ analytic engine as it does not have an official name yet. The partnership gives MDQ access to the comprehensive EMR infrastructure that the managed care consortium has in place including patients' disease and family histories, laboratory tests, physician notes, previous treatments and medical procedures, imaging data, and more.

The MDQ software analyzes the EMR data and generates statistical models of each disease recorded there, Gao explained. Then when a patient comes in to see the doctor, the system compares the information he or she provides to the models that it has created for each disease and looks for similarities with the patient's reported symptoms, test results, and so on. It then assigns a probability score for each disease that it thinks the patient potentially has, he said.

It's an improvement on the way physicians currently diagnose disease, which is often based on personal experience and training, Gao said. The physician still has to make the final call but at least is aware of all the possible conditions that the patient could be suffering from, based on years' worth of data collected from numerous patients, and can make more informed decisions about next steps. Other features of the system include tools to check that the correct ICD-9 codes are entered for billing purposes so that the right reimbursements are made, according to Gao. This is also of benefit to insurance companies who can use the MDQ platform to check for cases of medical fraud, he added.

Gao hopes to make the first version of the MDQ software available this year once it's been tested and validated at Kaiser Permanente, which is using the technology alongside its own clinical language processing pipeline to improve disease diagnosis internally. Kaiser Permanente has also done a study where MDQ's solution was used to detect adverse drug reactions from physician and nurse's notes.

In later iterations of its solution, the company hopes to add the ability to suggest the most effective treatments based on what worked in patients with similar symptoms, he said. For example, if a patient comes in with an infection that requires antibiotics and the physician could use the software to search for similar patients, he might find that with one particular drug, only 80 percent of patients recovered, while with another treatment, 90 percent of patients improved. They also plan to enable the system to provide cost estimates for suggested treatments.

Further down the road, Gao hopes that MDQ's platform will have the ability to take patients' genetic information into account when suggesting treatments instead of simply relying on statistics of what's worked at the population level. Right now, however, the focus is on bringing to market a solution that physicians can use now to provide the best care for their patients with limited side effects based on information that's available now, he said.

Other future plans include working on methods of using next-generation sequencing data to identify disease-causing pathogens faster than traditional laboratory testing which can take days to return results, he said. This would be helpful particularly for treating conditions where the time to treatment is crucial, for example in cases of sepsis infection which have about a 40 percent mortality rate, he said. Another plan is to work on developing methods for personalizing cancer treatments based on patients' genomic profiles. MDQ and Kaiser Permanente are currently holding discussions about these projects, Gao said.

The company is also currently hiring, hoping to at least double its current headcount of seven by next summer.