Students Develop Electronic Medical Record System Using AI to Improve Clinical Outcomes

Graduate students at the University of Miami Miller School of Medicine and a computer science student at the Georgia Institute of Technology are joining forces to develop an electronic medical record system that is more user friendly and clinically accurate. What sets the new platform apart from most others is its enhancement with artificial intelligence and machine learning capabilities.

Graduate students Saim Bhimji, Junaid Ali Muhammed and Timothy Guerard have developed a user-friendly and clinically accurate EMR system enhanced with AI and machine learning capabilities.
Saim Bhimji, Junaid Ali Muhammed and Timothy Guerard

Timothy Guerard, a Miller School master’s student in biomedical science, and his colleagues hope the technology will provide the information clinicians need to make accurate diagnoses and optimal treatment plans to improve clinical outcomes for patients.

An EMR System for Community Care

They are starting with the Medical Mission Abroad clinic sites, which rely on paper-based records. “There are a lot of limitations with that — we needed an EMR,” said Guerard, who established the Miller School’s local chapter of Medical Mission Abroad, which brings preventive medicine to underserved communities and has recently expanded.

Because no existing EMR system met their needs, “we decided to start from scratch,” he said. “It’s going to improve the clinics because it allows us to manage patient data — de-identified, of course — and enhance the continuity of care.”

The EMR project, called Agnodice, is named after the first female medical provider in ancient Athens. Like Hippocrates but less well known, Agnodice is considered the mother of medicine, Guerard said.

The initiative began as a web-based EMR. “Then we realized there’s a lot of machine learning capabilities or capacity that we could incorporate within our EMR, like pattern recognition, database research and decision making,” said Junaid Ali Muhammed, an M.B.S. candidate at the Miller School who collaborated with Guerard on the project.

Python-based computer algorithms will help Guerard, Muhammed and about 75 of the University’s undergraduate, postdoctoral and medical students helping at the Medical Mission Abroad sites to identify individual and population risk factors more easily. The algorithms are being trained on publicly available clinical datasets now, but the plan is to fine-tune the data with patient information moving forward.

System Will Flag Areas for Additional Attention

The hope is that the AI and machine learning will flag certain patients or presentations that might need additional attention or workup, not that it will replace physician skill or judgment. The system will start with basic information like age, sex, vital signs and in some cases ECG results before getting more sophisticated. Because the system can store vast amounts of data across specialties, the hope is that it will recognize less-common presentations and risks that a busy provider might not have time to consider.

The system can incorporate clinical readings like weight and blood pressure with individual risk factors and social determinants of health to make predictions. For example, does a person smoke? How much alcohol do they drink? What is their typical diet?

Built-in Privacy Protections

The algorithms do not use personally identifiable data such as name, date of birth or social security number. Instead, factors like clinical measurements, family history of disease and age go into the clinical calculations, said Saim Bhimji, another collaborator who Guerard described as “the coder behind everything.” Bhimji is a third-year student specializing in software development at Georgia Tech.

“This in no way identifies the patient. Data is strictly related to their clinical visit,” Bhimji added. People consent to have their data added to the EMR, including for research purposes.

Guerard pointed out that just moving away from paper records is a giant leap toward greater privacy protections. Furthermore, the de-identified data will be stored using elevated levels of data protection, including encryption. Unlike ChatGPT and other AI-based systems, the data is not open source, meaning no information is pulled from the Internet. A closed system offers greater privacy protections.

Guerard, Muhammed and Bhimji plan for the new EMR at Medical Mission Abroad clinics to go live starting in August.

The three collaborators acknowledged the experts at the Frost Institute for Data Science and Computing for brainstorming ideas to develop the technology. They also worked closely with many members of the Miller School faculty, who provided invaluable feedback regarding the EMR during development.

It’s not too late to get involved with the Agnodice project. “We’re always looking for help,” said Muhammed. Students, trainees or faculty with experience in coding, Python computer language and/or EMR development are especially encouraged to reach out. Email [email protected] for more information.

Tags: Agnodice, Medical Mission Abroad