Improving Patient Safety Outcomes With Precision Medicine
A collaboration between Johns Hopkins APL and Johns Hopkins Medicine in Baltimore is applying data science expertise to combat preventable harms in the health care system, using the Precision Medicine Analytics Platform.
Tue, 08/01/2023 - 09:40
Preventable harms are a well-known problem across the health care industry. Among the most common harms is venous thromboembolism, more commonly known as VTE — a term that includes life-threatening conditions caused by blood clots, such as pulmonary embolism and deep vein thrombosis.
VTE affects nearly one million people each year in the United States, claiming up to 100,000 lives and leaving many more saddled with long-term complications. Now, a collaboration between the Johns Hopkins Applied Physics Laboratory (APL) in Laurel, Maryland, and Johns Hopkins Medicine (JHM) in Baltimore is applying data science expertise to help make preventable VTE a thing of the past.
The effort leverages the Precision Medicine Analytics Platform (PMAP), a multiyear project across Johns Hopkins focused on accelerating precision medicine for clinicians and researchers by building a tool to facilitate discovery from data collected by JHM. It draws heavily on expertise within JHM’s Armstrong Institute for Patient Safety and Quality.
Data-Driven Tools for Targeting the Continuum of Care
“To prevent a hospital-associated VTE, critical steps are taken as soon as a patient is admitted,“ explained Luke Mullany, a biostatistician, epidemiologist and computer scientist in APL’s Research and Exploratory Development Department who is leading the work. “Initially, all patients are assessed upon admission for their risk of a VTE event during their stay in the hospital, and appropriate measures should be taken to try and prevent those events from occurring.”
But it doesn’t end there. Once a patient is identified as being at risk for VTE, physicians have to prescribe preventative measures drawn from evidence-based clinical guidelines, ensure that the patient actually follows through with those preventative treatments and, finally, monitor them after discharge to confirm that VTE is no longer a concern.
In collaboration with JHM leaders, Mullany and Vince Pulido, a data scientist in APL’s Asymmetric Operations Sector, are developing tools that enable their JHM counterparts to easily track their success rate in meeting these objectives. One of the first tools they developed was an automated report summarizing how closely the physicians’ orders aligned with recommendations from the clinical practice guidelines. By looking at a one-page summary, physicians can see a percentage that tells them how well they did, alongside their colleagues, over a 30-day period. The report is sent monthly to more than 300 providers in the medical and surgery departments.
Mullany and Pulido are also working on a web application that can visualize this data and make it available for hospital leadership to view at any time.
“We want to ensure that providers can answer some fundamental questions,” Mullany said. “Are we improving over time? Can we identify high-risk versus low-risk patients? Are we seeing more VTE events in patients admitted to certain service lines relative to other areas? By leveraging data that’s already being collected, we can enable leaders to answer questions at a lot of different levels about how providers are doing.”
Automated Information Gathering
Mullany and Pulido are also working on more advanced and technically challenging tools, applying machine-learning techniques to create algorithms that can automatically mine information from reports and other medical records. Among these is a tool for extracting patient outcomes from radiological imaging reports, including ultrasounds and X-rays. “By developing automated tools for extracting and organizing information, we’re hoping to make physicians’ lives easier by reducing their workload,” Pulido said.
These tools could also present new possibilities for collaboration between APL and JHM, he added. “Having access to these automated reports opens up further avenues for research about different ways to stratify patients in terms of their VTE risk,” he said. “We think we can offer long-term support with machine learning and data analytics on that front.”
Improving Outcomes Across the Health Care System
Taking a wider view, Suma Subbarao, a computer scientist and project manager at APL who helped oversee this effort, said that the approach taken to develop these tools — using data analytics to help physicians and hospital leadership monitor the continuum of care and intervene to prevent avoidable harms — will eventually benefit not only JHM but also hospitals around the United States and even the world.
“VTE is a common problem across many medical systems, and with PMAP we look to develop tools with the idea of scaling and sharing what we’ve learned,” she said. “We’re always asking the question of how we can generalize our successes and share them with other health care systems, nationally and globally.”
Mullany and Pulido collaborated closely with JHM physicians Michael Streiff, Elliott Haut, Margaret Krasne and Brandyn Lau, as well as Richard Day, center director of the Armstrong Institute.
The Applied Physics Laboratory, a not-for-profit division of The Johns Hopkins University, meets critical national challenges through the innovative application of science and technology. For more information, visit www.jhuapl.edu.