A Tool to Customize Drug Recommendations for Patients

Columbia Engineering students (left to right) Aleksandr Makarov, Harshit Saxena and Mayur Saxena took first place in the Columbia Venture Competition's technology division finals on April 28. (Kim Martineau)

With more than a hundred diabetes drugs to choose from, doctors don’t always know which one will work best for a particular patient. Some drugs may trigger allergic reactions, interact with other medications, or simply have no effect.

To remove some of the guesswork, a team of Columbia Engineering students has developed an application that estimates which drugs are less likely to cause side effects in specific patients. Their platform, Droice, took first place in the Columbia Venture Competition's technology division, competing against five teams to win the $25,000 prize.

“Pharmaceutical reps play a major role in deciding which drugs doctors prescribe to their patients,” said Aleksandr Makarov, a master’s student at Columbia’s Data Science Institute. “Not surprisingly, their focus is on the drug’s effectiveness, and not the risk of side effects.”

The U.S. Food and Drug Administration estimates that drug side effects cost $136 billion a year in complications resulting in illness or death. To reduce these risks, the FDA has made more clinical trial data and related information public. The Columbia team populated its tool with open data from ClinicalTrials.gov, openFDA, and the Medicaid-National Average Drug Acquisition Cost (NADAC) data sets, among others.

Initially they focused their attention on Type 2 diabetes, a disease that affects 30 million Americans. They designed their app to filter hundreds of clinical trial results for the 19 most popular diabetes drugs on the market. Then they developed profiles of patients likely to have a bad reaction to one drug or another. They added in data on each drug’s price and availability.

After winning the $2,000 grand prize at a Cornell Tech health hackathon in February, they began approaching doctors to test their platform. Nearly all of the 20 doctors they surveyed recommended making the tool compatible with their electronic medical record system as well as voice-activated, allowing them to keep their hands free.

“Doctors will use this tool only if they can fit it seamlessly into their current way of doing things,” said team member Mayur Saxena, a graduate student in biomedical engineering at Columbia Engineering.

In March, they entered an improved version of their platform into a second hackathon sponsored by Cornell and MIT, and won the $1,800 grand prize. They recently built a website to promote the tool and are lining up doctors to participate in a pilot study.

They are also expanding the platform to include the mostly commonly prescribed drugs for treating childhood asthma, another type of medication associated with a high number of side effects.

“Everyone has the potential to benefit from this application,” said team member, Harshit Saxena, who is studying for a master’s in computer science at Columbia Engineering. “Doctors save time, patients get better care and the U.S. healthcare system saves money.”

“If we could properly analyze the oceans of unstructured medical data out there we could make American healthcare much more efficient,” added Mayur Saxena.

— Kim Martineau


550 W. 120th St., Northwest Corner 1401, New York, NY 10027    212-854-5660
©2017 Columbia University