Tool helps hospitals access patient data during electronic outages, speeding up appointment scheduling and supporting faster, more informed treatment decisions.
SEATTLE — Last summer, a bungled software update led to canceled medical appointments and surgeries across the country when a national electronic health record system went down for 24 hours.
Incidents of electronic health record (EHR) downtimes are increasing, putting patient care at risk and delaying appointment scheduling. But computer science students at Northeastern University have developed an AI-integrated tool to store and analyze medical records so they can be retrieved — even when an EHR system is down.
When paired with an AI voice receptionist, the tool — named Samantha by the students — has the potential to reduce the time patients have to wait to get an appointment with a doctor, which can add up to a month or even longer.
“There’s a huge wait period because hospital staff have to analyze the documents of incoming patients,” says Isaac Premkumar, a graduate computer science student on Northeastern’s Seattle campus. When EHR systems are down, Premkumar says, making an appointment will take longer or may not be possible at all.
Working with fellow student Ishan Chaudhary, Premkumar developed MedQGraph, which extracts patient information from a hospital’s electronic health record system to help providers better understand patient medical histories when making treatment plans.
Together, MedQGraph and Samantha function as an interface for patients who want to schedule appointments over the phone.
When patients call their doctor’s office, Premkumar says, instead of being put on hold they can speak with Samantha, an AI agent integrated through MedQGraph to the EHR database. After describing their problem, Samantha would access information about past diagnoses, tests and hospital visits in order to speed up a treatment plan — even if the hospital’s EHR system is down.
MedQGraph received the judge’s choice award at Microsoft’s Open Source AI hackathon in April. Developed as a research project with other students in the Insight X Research Lab on the Seattle campus, MedQGraph was initially designed as a tool for hospital staff to summarize patient data, including records and transcripts of patient visits.
These records are often unstructured, Premkumar says, making it difficult to extract and analyze patient history. MedQGraph uses knowledge graphs — a way of storing interconnected data sets — to make doctor transcripts structured and searchable, he says.
“This software enables health care professionals to ask complex medical questions spanning multiple patient histories and time periods,” he says, “unlocking deeper insights for diagnosis, treatment, and research.”
Premkumar, who is pursuing a master’s degree in data analytics engineering, is testing MedQGraph on large hospital data sets to refine its functionality.
As an undergraduate, he focused on developing solutions for agriculture and sustainability.
“That’s where I gain my interest,” he says, “applying AI and machine learning to real problems.”