COPD Foundation Uses AI to Focus Patient Needs, Find the Undiagnosed
Artificial intelligence (AI) is becoming ubiquitous in all areas of life — used in the houses people live in and the cars they drive, and in the smartphones used every day to post updates on social media.
But now AI is making inroads into the nonprofit world, with the COPD Foundation using the emerging technology to determine the services the community needs — and to find people who aren’t yet diagnosed with the chronic inflammatory disease.
Over the years, the COPD Foundation has developed an interactive online network of more than 54,000 people on COPD360Social for nontuberculous mycobacterial lung disease, bronchiectasis, and chronic obstructive pulmonary disease (COPD).
Additionally, the COPD Foundation released three COVID-19 surveys to its community, which yielded a total of 1,800 respondents. Between historical surveys and COPD360, the foundation has gathered reams of information with a high level of insight into the patient community.
“When I joined [the COPD Foundation], I realized that we have a gold mine in terms of understanding the heartbeat of our community and understanding what people are concerned about,” said Ruth Tal-Singer, PhD, president and chief scientific officer of the foundation, in an interview with COPD News Today.
Analyzing all that data, especially open-ended narrative responses, might take a year for a human to sort through. But for natural language processing, which is a branch of AI that can analyze language and extract keywords, the COPD Foundation was able to turn the survey results into meaningful action within months.
“[AI] really has been helpful in rapidly synthesizing and processing these conversations that otherwise would really be difficult to classify,” said Danielle Boyce, senior research data analyst for the COPD Foundation.
The raw data from the open-ended questions of the COVID-19 surveys and COPD360Social conversations were run through SAS Viya’s artificial intelligence system, and Boyce saw a pattern emerge.
Many topics, conversations, and open-ended survey responses were revolving around long-term oxygen therapy, a widely used treatment that has been shown to improve survival in COPD and other respiratory disorders.
Boyce dove deeper into the data and discovered the story of one woman who could not access her oxygen tank because of COVID protocols. The elderly woman was not able to lift the tanks, and the delivery driver could not come into her home and do the lifting for her.
“We never could have predicted that ever. We never would have asked that in a closed-ended question. It totally would have been missed,” Boyce said. “And that allowed us to start looking at oxygen.”
That’s what led to one of the COPD Foundation’s newest initiatives, Oxygen360. Its goal is to offer solutions to some of the most difficult problems surrounding supplemental oxygen, such as heavy liquid oxygen options, the lack of automated systems that respond to a person’s oxygen saturation or pulse, and uncomfortable nasal cannulas.
To surmount these issues, the COPD Foundation is seeking patient feedback, by way of its monthly OxyTalk meeting, surveys, and analyses of discussions on COPD360Social, which will, in turn, be analyzed using natural language processing.
“We all operate with extremely limited resources as it is, and COVID-19 was moving so quickly, but we knew what we needed to do,” Tal-Singer said. “I knew where to prioritize the resources of the foundation, and that’s the beauty of artificial intelligence.”
The COPD Foundation also is using AI to further one of its priorities to find the “missing millions,” or people who have COPD but have not yet been diagnosed.
The nonprofit is in the final stage of developing a natural language processing system in conjunction with Savana, a company based in Spain that specializes in applying AI to electronic medical records. Savana’s AI is expected to be deployed in a primary care network with millions of patients to determine if it can be used to identify people who may have undiagnosed COPD or be at risk of developing COPD in the future.
“The objectives of this work are twofold — to capture the people with undiagnosed disease, so they can be managed better, and to identify people with early disease, where we think there’s much better opportunities for intervention to actually slow the progression of their disease,” said Bruce Miller, senior scientific director for COPD360Net, which is the COPD Foundation’s accelerator for therapeutics and digital health tools.
“Unfortunately, most people with COPD are not seen until they’ve lost about half or more of their lung function,” Miller said.
Privacy for health data also is top of mind for the COPD Foundation, both for its work understanding patient priorities and helping find the undiagnosed. It uses data that already has broad consent from the patient, and if the organization doesn’t already have that, it must work with an institutional review board to get permission to use it.
Savana and SAS Viya electronically de-identify patient information; however, sometimes that data slip through. For example, a person may enter his or her name in a survey response meant for listing medications. But the two companies and the COPD Foundation are prepared, as everyone who works with data must go through ethics training to properly handle any sensitive information that may arise.
Miller is expecting to have the natural language processing startup in the care network in the next weeks and will be able to determine its success by the end of the year, and Boyce has already demonstrated the power of AI to determine the issues patients care about.
“[Natural language processing] and AI really unlock data that’s otherwise inaccessible, or without a tremendous amount of manual work, is impossible to look at,” Miller said.