Predicting asthma attacks in kids

Millions of children have trouble controlling asthma attacks, leading to missed school days and scary trips to the emergency room. What if you could predict when a child will have an asthma attack and take steps to prevent it? That is what a multidisciplinary group of researchers at the University of California, Los Angeles, and the University of Southern California is hoping to do with the help of new wearable environmental sensors, smart devices, and mobile health technologies.

Pediatric asthma is the most prevalent chronic childhood disease in the US, says Alex Bui, director of medical imaging informatics and director of the Los Angeles Pediatric Research Using Integrated Sensor Monitoring Systems (PRISMS) Center at UCLA. Many things can play a role in asthma attacks. “It’s not just air pollution,” Bui says. Other contributors include common allergens, individual susceptibility factors, and behavioral factors. “Even though we treat asthma as a single disease, we often really need to get to the point of having more individually tailored care for it to be effective,” Bui says.

The Southern California team is building an informatics platform that integrates commercially available air pollution sensors as well as wearable environmental sensors developed by academic researchers. The project is part of the PRISMS initiative established in 2015 by the US National Institutes of Health. Information from the sensors, along with a person’s geolocation, physical activity, and health data, is wirelessly transmitted to the person’s smart watch and smartphone in real time. Participants use the smartphone to self-report symptoms and information related to daily activities. The informatics platform also uses the individual’s location to integrate weather, traffic, and air-quality data into the data stream.

The idea is to provide researchers with more detailed information about individual children with asthma to support epidemiological studies, says Bui, who is leading the effort to develop the informatics platform. Researchers can analyze the data to build models for predicting exposure scenarios that lead to asthma exacerbation for a particular individual. Participants can use such information to avoid certain situations or to know when to take control medication and carry an emergency inhaler.

A team led by Rima Habre, an assistant professor of clinical preventive medicine at the University of Southern California’s Keck School of Medicine, tested the platform last year in an epidemiological pilot study. The researchers recruited 20 children aged 8–16 with moderate to severe asthma. The children carried and wore various devices and answered questions on a smartphone for 1 week.

During the day, the children wore a smart watch that used low-energy Bluetooth to connect to the sensors and that served as a hub for the various devices, Bui says. The smart watch also provided information about the child’s activity level and physiological measures, such as heart rate.

The children also wore a palm-sized commercial air-quality monitor, called an AirBeam, for measuring airborne particulate matter with a diameter less than 2.5 µm (PM_2.5). Such particles are the most dangerous to health because they can penetrate deep into the lungs. They clipped the sensor to their backpack or belt during the day and could place it nearby in the same room while at home. They carried an inhaler that doubled as a medication sensor; it notified the informatics platform when the inhaler delivered control or rescue medication. In addition, the children measured their lung function twice a day, once in the morning and once in the evening, with a Bluetooth-enabled spirometer that determined the volume of air exhaled as a function of time.

The children’s gear also included a smartphone, which provided a larger screen for answering survey questions, extra computational power, and encryption to make the data framework more secure. The smartphone–smart watch combination provided researchers with GPS data to track an individual’s location.

To engage the children, the researchers created an animated dragon that showed up on the smartphone when the children needed to answer questions related to certain events. The questions provided researchers with more information about what an individual was doing. For example, “If a personal sensor on a child finds PM levels have gone up in the environment, the smartphone may issue a quick questionnaire asking the child: What is going on in the environment? Are you near a stove? Are you near a freeway?” Bui says.

When a child answered questions on the smartphone, the dragon was happy. When questions were unanswered, the dragon got hungry and wanted to be fed answers. To encourage children to answer questions, the researchers gave them a gift card at the end of the study in an amount tied to their compliance in answering the questions.

“We end up with a very highly correlated set of spatial and temporal information about where the individual is at any given point in time and what is happening with the individual,” Bui says. “We can put all of this together to try and build more individually tailored predictive models of what is happening.”

The researchers used feedback and exit surveys from participants in the 2018 study to improve the platform in terms of color and usability of the sensors. They are now testing the system in a formal epidemiological study launched in February, with a goal of recruiting 40 kids with moderate to severe asthma and monitoring them for 2 weeks. So far, at least 26 participants have enrolled in the study, Habre says.

They are also testing three environmental sensors which were developed as part of the PRISMS program, in a real-life setting with kids with asthma. One of the new sensors is a wrist-worn device developed by researchers at Arizona State University for monitoring ozone. Another sensor, developed by researchers at Columbia University and AethLabs, measures black carbon and brown carbon, which are markers of air pollution generated by different types of combustion. A third sensor, developed by researchers at the University of Washington, detects particulate matter. The UW sensor can also collect small particles from sources such as diesel, wood smoke, and cigarettes for later analysis.

The researchers are continually tweaking the platform, making improvements as they learn “what is working and what is not working,” Habre says. For example, the team learned that low-cost sensors need to be calibrated more frequently than sensors the researchers were used to.

People also experience a lot of PM_2.5 peaks during the day, Habre says. “Do we want to show people these peaks that might trigger some kind of false alarm when we don’t know what it means for their health?” she asks. The researchers show participants hourly aggregate levels on a map to make the data “visually interesting,” Habre says.

The researchers are still struggling with privacy issues related to GPS data. “You can really derive a lot about an individual’s behavior just from looking at their GPS data,” Habre says. Because of the potential benefits, many subjects are not worried about being tracked, she adds.

“Mobile health approaches have amazing potential” to help researchers understand individual exposures, Habre says. These approaches are making it possible “to understand how multiple exposures, behaviors, stressors, and ecological factors all work together to affect asthma.”