Electronic health records have been widely adopted in the hope that they will save time and improve the quality of patient care. But due to fragmented interfaces and cumbersome data entry procedures, physicians often spend more time navigating these systems than interacting with patients.
Researchers at MIT and Beth Israel Deaconess Medical Center are combining machine learning and human-machine interaction to create a better electronic health record (EHR). They developed MedKnowts, a system that unifies the processes of searching for medical records and documenting patient information in a single, interactive interface.
Driven by artificial intelligence, this “smart” EHR automatically displays personalized and patient-specific medical records when a clinician needs them. MedKnowts also provides autocomplete for clinical terms and automatically fills in fields with patient information to help physicians work more efficiently.
“At the origin of the EHRs, there was this immense enthusiasm that organizing all of this information would be useful in order to be able to track billing records, report statistics to government, and provide data for scientific research. But few have stopped to ask the profound questions about their usefulness to the clinician. I think a lot of clinicians feel that this EHR burden has been placed on them for the benefit of bureaucracies, scientists and accountants. We entered this project wondering how EHRs could actually benefit clinicians, ”says David Karger, professor of computer science at the Laboratory for Computing and Artificial Intelligence (CSAIL) and lead author of the article.
The research was co-authored by CSAIL graduate students Luke Murray, who is lead author, Divya Gopinath, and Monica Agrawal. Other authors include Steven Horng, attending physician in emergency medicine and clinical manager of machine learning at the Center for Healthcare Delivery Science at Beth Israel Deaconess Medical Center, and David Sontag, associate professor of electrical engineering and computer science at MIT and member of CSAIL. and the Institute for Medical Engineering and Science, and Principal Investigator at the Abdul Latif Jameel Clinic for Machine Learning in Health. It will be presented at the Association for Computing Machinery Symposium on Software and User Interface Technology next month.
A problem-oriented tool
To design an EHR that would benefit physicians, researchers had to think like physicians.
They created a note-taking editor with a side panel that displays relevant information about the patient’s medical history. This historical information appears in the form of maps focused on particular issues or concepts.
For example, if MedKnowts identifies the clinical term “diabetes” in the text as a clinician type, the system automatically displays a “diabetes card” containing medications, lab values, and extracts from previous records relevant to the patient. treatment of diabetes.
Most EHRs store historical information on separate pages and list medications or lab values alphabetically or chronologically, forcing the clinician to search the data to find the information they need, Murray explains. MedKnowts only displays information related to the particular concept the clinician is writing about.
“It’s more in line with how doctors view information. Often times, doctors do this unconsciously. They will go through a medication page and focus only on medications that are relevant to current conditions. We help do this process automatically and hopefully get some things out of the doctor’s head so that they have more time to think about the complex part, which is figuring out what is wrong with the patient and what is wrong with the patient. develop a treatment plan, ”Murray said. .
Pieces of interactive text called bullets serve as links to related maps. When a doctor types a note, the autocomplete system recognizes clinical terms, such as drugs, lab values, or conditions, and turns them into bullets. Each bullet is displayed as a word or phrase that has been highlighted in a certain color based on its category (red for a medical condition, green for a drug, yellow for a procedure, etc. )
Through the use of autocomplete, structured data on the patient’s conditions, symptoms and medication usage is collected without any additional effort on the part of the physician.
Sontag says he hopes this breakthrough “will change the paradigm of creating large-scale health data sets to study disease progression and assess the true effectiveness of treatments.”
After a year-long iterative design process, the researchers tested MedKnowts by deploying the software in the emergency department at Beth Israel Deaconess Medical Center in Boston. They worked with an emergency doctor and four hospital scribes who enter notes into the electronic health record.
Deploying the software in an emergency department, where doctors operate in a very stressful environment, involved a delicate balance, says Agrawal.
“One of the biggest challenges we faced was trying to get people to change what they’re doing now. Doctors who have used the same system, and done the same click dance so many times, are forming a kind of muscle memory. Whenever you’re going to make a change, is the question whether it’s worth it? And we definitely found that some features were used more than others, ”she says.
The Covid-19 pandemic has also complicated the deployment. Researchers had visited the emergency department to get a feel for the workflow, but were forced to end those visits due to Covid-19 and were unable to be in the hospital while the system was deployed.
Despite these initial challenges, MedKnowts became popular with scribes during the month-long rollout. They gave the system an average rating of 83.75 (out of 100) for usability.
Scribes found the autocomplete feature particularly useful in speeding up their work, according to the survey results. Plus, the color-coded chips helped them quickly scan notes for relevant information.
These early results are promising, but as the researchers review the comments and work on future iterations of MedKnowts, they plan to proceed with caution.
“What we’re trying to do here is make it easier for the doctors and let them speed up. There is a certain risk there. Part of the purpose of bureaucracy is to slow things down and make sure all i’s are pointed out and all t’s are crossed out. And if we have a computer pointing i’s and crossing t’s for doctors, it can actually defeat the goals of bureaucracy, which is to force doctors to think twice before making a decision. We need to think about how to protect physicians and patients from the consequences of increased physician efficiency, ”says Karger.
A longer term vision
The researchers plan to improve the machine learning algorithms that drive MedKnowts so that the system can more effectively highlight the most relevant parts of the medical record, says Agrawal.
They also want to take into account the needs of different medical users. Researchers designed MedKnowts with an emergency department in mind, an environment where doctors typically see patients for the first time. A primary care physician who knows his patients much better would likely have different needs.
Longer term, researchers plan to create an adaptive system that clinicians can contribute to. For example, maybe a doctor realizes that a certain cardiology term is missing from MedKnowts and adds that information to a map, which would update the system for all users.
The team is exploring commercialization as a route for further deployment.
“We want to create tools that allow doctors to create their own tools. We don’t expect doctors to learn to be programmers, but with the right support, they might be able to radically customize the medical applications they use to truly match their own needs and preferences, ” explains Karger.
This research was funded by the MIT Abdul Latif Jameel Clinic for Machine Learning in Health.