This AI Startup Raised $15 Million To Help Patients Transcribe Doctor Appointments

^Imagine your doctor diagnosed you with pyelonephritis. Sounds scary and you might tune out some of the explanation where you’re told it’s a kidney infection. Wouldn’t it be great to have an automatic transcript of the conversation, so you could go back and read it later? Enter Pittsburgh-based startup Abridge. The company has combined a consumer app with artificial intelligence to securely transcribe any medical conversation that a patient chooses to record. Abridge, which announced its official launch and $15 million in funding on Tuesday, already has 50,000 users, mainly garnered through word-of-mouth. 

“Technology has this potential to help so many more people than I could ever see in my weekly clinic,” says Abridge cofounder and CEO Dr. Shiv Rao, 41, a cardiologist who still occasionally sees patients.

The problem the company aims to solve is this: nearly two-thirds of patients surveyed by Abridge forgot more than 40 percent of the conversations they’ve had with their doctors, including key details, like what to do next to stay healthy. The app highlights key information from those conversations to help them. “It's helping them understand the details of their health from the high-level care plan down to the details of diagnoses, procedures, or medications that are discussed,” says Rao. Patients can also share the transcripts with caregivers or family members or other clinicians. 

It’s the focus on the consumer—and using technology to bridge the gap between patients and clinicians—that attracted lead investor Union Square Ventures. “Most healthcare entrepreneurs attempt to solve things from the healthcare side,” says Andy Weissman, managing partner at Union Square Ventures. But Rao and his cofounder and CTO Sandeep Konam, 25, took the opposite approach. They focused on building what Weissman calls a “buttery user experience” for the consumer, which was key to other technology investments made by the firm, including Twitter, SoundCloud and Kickstarter. “It was rare for us to see people talking about that in healthcare,” says Weissman. Bessemer Venture Partners and the University of Pittsburgh Medical Center also participated in the $10 million Series A round. The company previously raised $5 million in a seed round, which included Aneesh Chopra, the U.S. Chief Technology Officer under President Obama, and investor Esther Dyson.

^In the era of “Dr. Google,” where patients often turn to the internet to get answers to pressing medical questions, Abridge has defined over 400,000 medical terms in easy to understand language, pulling from resources like the Mayo Clinic and the National Library of Medicine. Konam and his team spent two years developing algorithms that could understand medical jargon, surface key terms and detect next steps. The technology skips over discussions of topics like the weather and only focuses on the medically relevant parts of the conversation. “Our thesis has always been that we should focus on the conversation between a patient and the doctor,” says Konam. “Because that conversation is upstream of all the diagnostics, therapeutics, labs and other services.” 

What Abridge has built is more complicated than the technology that underpins voice-controlled systems like Alexa or Siri, since those devices have corrections through voice commands. The company’s algorithms had to be able to recognize complicated names of prescriptions, such as atorvastatin for cholesterol or lisinopril for hypertension, on the first try. Abridge has also partnered with the drug pricing company GoodRx (which went public in September) to include information about available coupons next to the name of prescriptions. Privacy and security of protected health information was paramount to the platform design, says Konam. Another big challenge was being able to interpret varying degrees of microphone quality and people talking over each other. 

Abridge was founded in 2018, but the cofounders met a couple years earlier, when Konam was pursuing a master's in robotics at Carnegie Mellon University in Pittsburgh, where Rao was a cardiologist at the University of Pittsburgh Medical Center. The company’s ties to both institutions still run deep. The algorithms were developed using a de-identified and consented research dataset from UPMC, which is also an investor, and Abridge continues to work with machine learning professors at Carnegie Mellon. The company has 15 full-time employees and plans to use the funding to expand on the engineering and product side to continue developing new features.

“The most important part of the health journey often happens between visits,” says Rao. That is the next step for Abridge, using nudges to get patients to take the next steps—fill their prescriptions, schedule a follow-up visit, get their imaging done. While the main users are patients, the main customers Abridge hopes to attract are health systems and health insurers, who are willing to pay for the use of a platform that could help improve the long-term outcomes for their patients. “Our ambition here isn't to automate humans away,” says Rao. “That's not where we're coming from. We're coming up with an angle of using cutting edge technology, like machine learning to augment humans and the relationships between them.”