Community outreach and support for COPD patients enhanced through natural language processing and machine learning

Text analytics and sentiment analysis from SAS help the COPD Foundation provide health resources and emotional support for patients and caregivers

Breathing in and exhaling. It’s something the average adult does about 22,000 times a day – often without thinking about it. But it’s not that simple for more than 320 million people around the world who have chronic obstructive pulmonary disease (COPD). For many, breathing isn’t an involuntary action – it’s something they actively think about as everyday activities can result in shortness of breath, wheezing, coughing and tightening of the chest. Even when symptoms are at bay, the ability to breathe is still top-of-mind for those affected by the disease as they wonder what might trigger their next flare up.

With the onset of the COVID-19 pandemic, the stress of living with COPD increased even further. Patients and those who care for them needed an organization like the COPD Foundation that connected with researchers, health care professionals and patients, actively reaching out to listen, learn and take collective action.

Since its founding in 2004, the nonprofit foundation works to speed innovations that make treatment of COPD more effective and affordable. Through a unique 360-degree approach, the foundation mobilizes partnerships among individuals with COPD, caregivers, health care professionals, researchers, academic institutions, government agencies and industry leaders. Their mission is to prevent COPD, improve the lives of those affected by it and stop the progression of the disease.

Online community provides wealth of knowledge

Because people living with chronic lung conditions may also have mobility challenges, the COPD Foundation wanted to use technology to better reach its demographic in a more efficient way. In 2014, the foundation established its online community, COPD360social. Since then, it has seen significant engagement and growth in its user base and substantial gains in knowledge and information sharing.

With nearly 50,000 community members representing 150 countries, the foundation can gain first-hand insights from those who are affected by COPD, enabling it to better serve this community by learning what they need and want. But sometimes an abundance of information can be problematic. While the foundation’s staff could search on individual topics to gauge community sentiment, as the community grew, these individual searches were neither scalable nor efficient.

“You could collect data all day long, seven days a week, but unless you can properly analyze the data set, you can't get much value out of it,” says Vincent Malanga, Chief Information Officer of COPD Foundation. “While we have a fantastic platform to collect information, we needed to partner with someone who could help us expertly gather and analyze everything coming in so that we could understand what the data was saying and how to best respond.”

Finding the right solution

The COPD Foundation did not want to use a third-party analytics solution where another company would own its data, as this could limit its access to the data.

“Our goal was to make sure we had total ownership of the data we collected in COPD360social. We recognized early on that we wanted to use the data for research and analysis to ensure that we support our community’s needs, and we knew that artificial intelligence – particularly natural language processing and machine learning – would become more prevalent,” says Malanga. “There’s tremendous value in being able to analyze the data collected through members in the community discussing their real-world experiences living with this disease.”

And it’s not just the patients the COPD Foundation wants to hear from. It seeks insights from family members, friends and care providers. “Everyone who is impacted by the disease in one way or another can communicate on COPD360social,” Malanga says. “This diverse community gives us a lot of information on their challenges, accomplishments, goals and concerns, along with the everyday tips and tricks that people share with each other.”

The foundation chose SAS for in-house natural language processing of text data generated by survey responses on COPD360social.

“SAS has world-class analytics, and it met the needs of our enterprise architecture approach, where it’s scalable, secure, flexible and extendable,” Malanga says. We want software to be able to conform to what we need to do in the future – like possibly integrate with open source solutions and pre-trained language models – and SAS gives us that freedom.

“Once we decided SAS was the way to go, we knew we wanted our SAS solution hosted in a cloud environment, which is why we partnered with Pinnacle and went with an Amazon Web Services infrastructure to support it,” Malanga says. “While there are many moving parts, everything runs seamlessly, and we’ve had a very positive experience.”

Insights from patient surveys

At the onset of the pandemic, there were many unknowns. The foundation launched a series of global surveys to get first-hand perspectives and answers. Responses came from around the world, including the US, Canada, Europe, Australia, East Timor, Nepal, Pakistan and Israel, representative of both urban and rural areas.

“Through our direct interactions with the COPD community, we were in a position to understand what’s really important to this unique population,” says Ruth Tal-Singer, President and Chief Scientific Officer of COPD Foundation. “We had reports from China and Italy indicating that COPD patients and people with lung diseases were susceptible to poor outcomes, so we wanted to respond very rapidly to the growing pandemic.”

The foundation soon launched a dedicated COVID-19 webpage.

“In a patient advocacy setting, it’s essential that you don’t try to force an answer and predict what people are thinking,” says Danielle Boyce, Senior Research Data Analyst at the COPD Foundation. “It’s vital that the surveys the foundation conducts have free-text fields where participants can answer open-ended questions, rather than closed surveys with predetermined answers they can choose from.”

Natural language processing uncovers hidden insights

Using text analytics and sentiment analysis, Boyce generated word clouds, enabling the foundation to explore volumes of unstructured data, identify patterns and create reports to help with community outreach and support.

“The speed at which we can analyze data with natural language processing software is unheard of in traditional research settings,” Boyce says. “What would have taken me three months to do in another setting, line by line analysis, I can do very quickly with SAS.”

The foundation analyzed open-ended text to rapidly gain insights from the COPD community.

“We could quickly identify top-of-mind issues for this population, as words began jumping out at us – like medications, mask use, gloves, grocery shopping, curbside pickup – and we could see how these words were connected to other words, like doctor, fear, isolation, social distancing and depression,” Boyce says. “SAS allows us to pinpoint topics, enabling us to understand our community members’ concerns and needs, both now and as they evolve throughout the duration of the pandemic.”

Unexpected discoveries lead to coordinated action

There were a number of surprising findings for the COPD Foundation – and these discoveries directly influenced the foundation’s response to its community members. Here are four of the most noteworthy concerns:

Access to oxygen: Some respondents were worried about maintaining home oxygen delivery and pickup services. They were physically unable to carry empty oxygen tanks outside their houses, but the delivery service would no longer come inside to exchange tanks because of social distancing mandates.

Supply chain: Numerous community members voiced distress over supply chain issues that could prevent them from receiving their essential medications. This could result from bottlenecks in the manufacturing and distribution process, as well as increased demand for inhaler medications such as albuterol for COVID-19 patients.

Employment: Essential workers with COPD were worried about wearing face masks – especially for long periods of time – due to their breathing difficulties. Some were concerned they would no longer be able to keep their jobs during the pandemic.

Exercise and pulmonary rehabilitation: The ability to continue physical fitness routines during COVID-19 was top-of-mind for many. Respondents were concerned about their access to gyms, supervised exercise programs and pulmonary rehabilitation services – particularly after hospitalization.

Based on concerns raised on COPD360social and in the surveys, the foundation conducted Q&A sessions, webinars and other advocacy activities, as well as created helpful fact sheets and reference guides. For example, in response to the concerns about exercise, the foundation generated specific webpages and videos that demonstrate at-home exercises people could do with their doctors’ approval.

“Using SAS to sift through volumes of text data and analyze sentiment, we were able to quickly pivot our focus and concentrate on topics relevant to our community,” Tal-Singer says. “It’s extremely valuable for us to obtain feedback so quickly because we’re able to provide answers and support in real time. The concerns brought to light helped form some of our public policy and advocacy efforts on behalf of the COPD community.

Operational improvements and preparing for what’s to come

Another benefit the COPD Foundation sees as a result of using SAS is the ability to achieve more with limited resources – both funds and headcount.

“We have hard data to support what we’re doing at the foundation, which is so important when you’re a nonprofit,” Tal-Singer says. “Nonprofits only have so much funding, and we must be judicious with how we spend money. The more we can do with analytics, the better we can serve all our partners, our patients and our caregivers. We can allocate resources more effectively through the ability to be nimble and successfully integrate information from different sources.”

The work that the foundation is doing in response to COVID-19 is enabling recovery and helping reimagine a brighter future for the COPD community. Lessons learned will help shape the foundation’s practices and responses to upcoming threats.

“We have this great connected infrastructure,” Boyce says. “We can ask questions, quickly analyze data, publish the findings and share everything on COPD360social again and again – a loop of information that helps a vast number of people. It’s truly amazing.”

“Our response to COVID-19 has demonstrated our ability to be agile and explore new things,” Tal-Singer adds. “We have a major effort to make sure we address our community’s needs because our world is not going to go back to baseline. Much of what changed during the pandemic is likely here to stay for a long time, so we’re adapting the way we work and serve the community now and in the years to come.”