Improving Patient Care with Machine Learning At Beth Israel Deaconess Medical Center
Improving patient care with machine learning
Inefficiencies in hospital management and operations are not only extremely costly to providers, insurers, patients, and taxpayers, but they can result in precious resources being diverted away from patient care. These inefficiencies drive healthcare costs up and can contribute to life-threatening medical errors.
The work now underway at BIDMC strives to identify new methods that can be shared across the healthcare industry, with the goals of advancing better patient outcomes, decreasing hospitalizations and readmissions, and lowering health care costs for all Americans. BIDMC’s machine learning research seeks to create data-based solutions and processes to address these challenges, be scalable across the healthcare industry, and further enhance patient care.
An initial BIDMC research project used machine learning to optimize the schedules of its 41 operating rooms and align those schedules to improve patient flow in the inpatient setting. Another project has leveraged machine learning to improve operational flow within operating rooms. Now, incoming pre-surgical document packages will be scanned as images and processed with TensorFlow on Amazon SageMaker, hosted in BIDMC’s secure AWS cloud. This machine learning driven process automatically recognizes and inserts consent forms into corresponding electronic health records (EHR), saving hospital staff hours of manual work. BIDMC built a model that scans EHRs to look for key elements like a completed consent form. If a consent form isn’t found, a signal appears on the EHR and nurses follow up with those patients.
Similarly, BIDMC has more than 490 inpatient medical/surgical beds that are highly occupied, and its team strives to successfully perform surgical procedures in order that patients can be treated and recover in a timely manner. However, procedures were sometimes delayed or rescheduled because a completed History and Physical (H&P) form, which is required before surgery can start, could be difficult to locate in the documentation that is sometimes faxed to the hospital. To solve this, BIDMC now uses Amazon Comprehend Medical to extract key medical terms and insights that are used in a machine learning model to identify H&Ps. As a result, valuable time can be saved and delays and rescheduling can be prevented.
“Advancements in technology and deep learning have the power to advance care and make a meaningful difference in the lives of thousands of patients and providers,” said Manu Tandon, Chief Information Officer at Beth Israel Deaconess Medical Center.
“Every minute spent on cumbersome clerical tasks and management adds up to millions in lost productivity and directly impacts patient care,” said John Halamka, MD, Executive Director, Health Technology Exploration Center at Beth Israel Deaconess Medical Center and International Healthcare Innovation Professor at Harvard Medical School. “This machine learning research sponsorship will support our commitment to using new and emerging technologies in health care to drive projects that will transform care for patients at BIDMC and around the world.”
Supporting patient adherence and operating room efficiency at BIDMC
Additional projects underway at BIDMC involve predicting which patients are likely to keep their scheduled office appointments and which are not. This project is being built using the Apache MXNet deep learning API and Amazon SageMaker. It will help BIDMC reach out to patients who might miss appointments so that care can be delivered in a timely manner, improving the patient experience and outcomes.
Similarly, BIDMC has developed another machine learning model built on AWS that can detect where simple operating room schedule modifications would improve efficiency, save costs, and balance the load of the hospital during busier times. At the same time, the model can predict the outcome of changes to the schedule and identify what mitigations will minimize negative impacts on patient care.
Making it easier to plan ahead in the Emergency Department
Future projects include assessing the overall level of risk in intensive care units and predicting when the hospital will experience an unexpectedly high volume of incoming patients. For example, BIDMC’s emergency department (ED) typically sees a surge in patient visits during the middle of the week, which can strain hospital resources. BIDMC and academic research partners will analyze datasets, including ED admissions, transfers between healthcare institutions, referrals, pre-scheduled surgeries, patient discharges, and other variables using services such as Amazon QuickSight and Amazon Forecast.
Because of the high volume of data collected, BIDMC will use the AWS Cloud to load and process the necessary data quickly and significantly speed up model training. And using machine learning services like Amazon SageMaker, researchers at BIDMC will build deep learning models that are capable of making highly accurate predictions of where and when space will free up in the hospital for unexpected patients. These projects will help build effective models with the long-term vision of deploying them across the healthcare industry and beyond.
“We are proud to be a part of the innovation happening in healthcare right now and are keen to support organizations like BIDMC who are leading the way in using machine learning technologies to deliver enhanced, personalized care and improved patient experiences,” said Swami Sivasubramanian, Vice President of Machine Learning at AWS. “Supporting BIDMC’s efforts with machine learning services and expertise is a natural extension of the long relationship between our organizations, and we’re excited to help enable their researchers to accelerate the development of models that can advance patient care. BIDMC’s innovations using AWS machine learning services like Amazon SageMaker will ultimately pave the way for other healthcare providers to save lives and reduce costs for patients nationwide.”
AWS is proud to sponsor BIDMC’s research which is a continuation of the company’s mission to put machine learning in the hands of all developers, across the public sector, education, healthcare, and beyond.