Smartphone-recorded breathing sounds are ‘quite good’ for predicting OSA

Key takeaways:

• Both iOS and Android phones had high levels of accuracy in predicting obstructive sleep apnea.
• Smartphone recordings could allow for more follow-ups with patients compared with polysomnography.

Sleep breathing sounds recorded via smartphone yielded a “quite good” prediction of obstructive sleep apnea, according to study results published in JAMA Otolaryngology-Head & Neck Surgery.

Jeong-Whun Kim

“Our technology can work as a kind of a triage for hospital level 1 (type 1) full night polysomnography (PSG),” Jeong-Whun Kim, MD, PhD, of the department of otorhinolaryngology–head and neck surgery at Seoul National University Bundang Hospital, told Healio. “OSA is an underdiagnosed and undertreated disease; we hope our easy-to-use technology can be helpful in addressing these problems.

“[Further,] in a follow-up period during PAP therapy/oral device or before and after surgery and even during weight reduction/exercise/lifestyle modification, clinicians have difficulty repeating PSG because of high cost and insurance coverage,” Kim added. “Our technology can be powerful for follow-up periods to monitor the patients.”

In a prospective, diagnostic study, Kim and colleagues assessed breathing sounds from 101 Korean adults (mean age, 48.3 years; 50.5% women) while they to slept see if sound recorded by an iOS smartphone and Android smartphone could be used to predict OSA.

Alongside the two smartphones, researchers also used level 2 home PSG to record participants’ breathing during sleep and act as a reference.

Researchers conducted binary classifications for different threshold criteria based on an apnea hypopnea index (AHI) threshold of five, 15 or 30 events per hour.

Of the total cohort, 50 adults did not have OSA, whereas 21 had mild OSA, 16 had moderate OSA and 14 had severe OSA.

Both smartphones had a similar AHI range (iOS, 0-75.6 events per hour; Android, 0-74 events per hour). The range of AHI was slightly smaller when derived from home PSG (0-69.7), and the mean AHI (12.4 events per hour) was closer to the mean found with the Android phone vs. the iOS phone (12.6 vs. 13.7 events per hour).

Also, researchers found that AHI from home PSG was strongly correlated with AHI from each of the smartphones (iOS, r = 0.958; Android, r = 0.953).

The iOS and Android devices had comparable sensitivity values at an AHI level of five events per hour (92.6% vs. 92.2%), 15 events per hour (90.9% vs. 90%) and 30 events per hour (93.3% vs. 92.9%).

Both devices also had almost identical specificity values at five events per hour (iOS, 84.3% vs. Android, 84%), 15 events per hour (94.4% vs. 94.4%) and 30 events per hour (94.4% vs. 94.3%).

Lastly, the accuracy of both devices improved as the AHI level increased. iOS-recorded audio yielded an accuracy of 88.6% for an AHI threshold of five events per hour, which went up to 93.3% at the AHI level of 15 events per hour and 94.3% at the level of 30 events per hour.

A similar pattern was observed when assessing the accuracy of the Android phone at the AHI levels of five events per hour (88.1%), 15 events per hour (93.1%) and 30 events per hour (94.1%).

“Before this trial, we already tested [this] several times in hospital settings and found that using the sound information will work to diagnose sleep apnea, so we were expecting a good result,” Kim told Healio. “Nevertheless, the prediction performance was quite good, and we were happy.”

Kim has several plans for future research on the use of smartphones as they relate to OSA.

“In the future, I have to validate that our easy-to-use technology is helpful for treatment outcomes and early detection of OSA patients,” Kim told Healio. “I hope this technology will be eligible for prescription of PAP therapy and for monitoring during PAP use. I also hope this can be used to monitor patients during oral device use and after any kind of sleep apnea surgery. Further, I aim to monitor the patients who want to modify their lifestyle and lose weight.”

For more information:

Jeong-Whun Kim, MD, PhD, can be reached at kimemails7@gmail.com.