Sleep apnoea screening methods compared: Smart devices wake us up to the issue

Sleep is having a moment. We’ve recently seen prominent documentaries released on sleep, the emergence of sleep influencers on social media platforms, and even sleep tourism! Yes, it’s a thing.

A strong driver of this increased interest is the emergence of smart devices that let people track their sleep. These devices are giving many of us new perspectives on our sleep, and at a granular level – breaking down our night’s sleep into easy to visualise categories and scores.

Armed with this information, people can use these devices to tune their sleep cycles just as they would use them to tweak their diet, their exercise routine, or their minute-to-minute activity levels. Far more than just helping people count how many hours of REM sleep they’re getting, these technologies are also shining a light on one of the world’s most prevalent sleep disorders: sleep apnoea.

A night-time companion for millions of Australians

Sleep apnoea involves interruptions in breathing during sleep. These interruptions can last from a few seconds to minutes and can occur multiple times throughout the night. The most common type is obstructive sleep apnoea (OSA), caused by a blockage of the airway, usually when the soft tissue in the throat closes during sleep. Affecting up to 38% of the population by some counts[1], OSA impacts millions of Australians and over a billion people worldwide, with a commensurate impact on our healthcare system.

Common symptoms during sleep include snoring, gasping for air, and low blood oxygen levels. Upon waking, these interruptions in sleep leave affected people under-rested, resulting in excessive daytime sleepiness and difficulty concentrating. It can also lead to health problems including cardiovascular issues, high blood pressure, and diabetes. In patients with these comorbidities, health professionals often fail to detect the underlying sleep apnoea – just one reason why approximately 80% of people with OSA remain undiagnosed and untreated.

A new era in OSA detection

These same devices that are slicing and dicing your night’s sleep into REM, light, and deep sleep and spitting out a score are now offering to help detect signs of sleep apnoea.

In the past, simple phone-based apps would use the onboard microphone to detect a person snoring, correlating that to a risk of apnoea. Newly launched features on Apple and Samsung smartwatches are taking a different approach, measuring wearer movements during the night using onboard accelerometers to detect breathing disruptions. In addition, Samsung Galaxy Watches use their onboard pulse oximeters to measure blood oxygenation levels.

To form a recommendation, the watch must be worn for a set period of time – two consecutive nights for Samsung Galaxy Watches, and at least 10 nights within a 30-day period for Apple Watches. This data is then processed by a proprietary machine learning algorithm, that calculates whether the wearer is at an elevated risk of sleep apnoea – both manufacturers are careful to note that any recommendations from the watch require confirmation by a doctor.

Sleep apnoea by the numbers

Both algorithms were assessed against the gold standard of polysomnography (PSG) in a sleep laboratory and both achieved clearance from the U.S. Food and Drug Administration (FDA) as “safe and effective” to “assess risk of sleep [apnoea]”.

But how accurate are these systems? And how many people will they help? To answer these questions, we sifted through both companies’ published data as well as a range of third-party sources.

Sleep apnoea
detection algorithm
False
negatives
False
positives
Accuracy*
Apple*[2]
33.7%
1.5%
94.2%
Samsung[3]
8.9%
12.3%
88.1%

*Weighted by prevalence of population with mild or no sleep apnoea.

Apple algorithm performance

Apple has optimised its algorithm to minimise false warnings, and it has done a good job in this regard. In its clinical validation, it mistakenly alerted just 1.5% of people with mild or no sleep apnoea (adjusted for the prevalence of each). However, this risk averse approach means it is less effective at alerting people who do have sleep apnoea. Of those with moderate or severe sleep apnoea, 33.7% did not receive an alert! Consequently, a lot of people may mistakenly think that the lack of a notification from their smartwatch means they do not have the condition, when this is far from certain.

Samsung algorithm performance

Samsung’s algorithm has a more balanced approach, falsely alerting 12.3% of people with mild or no sleep apnoea and neglecting to warn 8.9% of people with moderate or severe sleep apnoea in the study in its submission to the FDA.

Barriers to adoption

There is an often-overlooked limitation of these smartwatch offerings: They don’t have significant reach. Only 17.3% of Australians own an Apple or Samsung smartwatch and many do not wear them while sleeping. Conversely, 50.4% of Australians own an iPhone, which makes it the ideal device for democratising access to OSA testing.

We are launching a simple iPhone test for issues with sleep that provides a result within minutes.

Frequently asked questions

Can a smartwatch detect sleep apnoea?

Apple and Samsung algorithms hold FDA clearance to assess the risk of sleep apnoea, and both are reasonably accurate: 94.2% and 88.1% against polysomnography. Neither diagnoses the condition; both manufacturers state that a recommendation requires confirmation by a doctor.

Which is more accurate, Apple or Samsung?

Apple is more accurate overall (94.2% against 88.1%), but misses far more cases: a 33.7% false negative rate against Samsung’s 8.9%. Samsung over-refers instead, with a 12.3% false positive rate against Apple’s 1.5%. They fail in opposite directions.

Why are smartwatches not suitable for workforce screening?

Two reasons, neither about accuracy. Only a minority of any workforce owns a compatible device, and ownership is not randomly distributed. And a programme requiring two to ten nights of consistent wear will not complete for a large share of those who begin it: least of all shift workers, whose risk is highest.

How many nights of wear do they need?

Two consecutive nights for Samsung Galaxy Watches; at least ten nights within a 30-day period for Apple Watches.

Does any screening method replace a sleep study?

No. Polysomnography remains the reference standard and the route to diagnosis. Screening identifies who should be referred; it does not establish what they have.

Optalert breakthrough

We discovered a biomarker that will transform the lives of millions of people with issues with their sleep.

References:

[1] Senaratna CV, Perret JL, Lodge CJ, et al. Prevalence of obstructive sleep apnea in the general population: A systematic review. Sleep Med Rev 2017;34:70-81. doi: 10.1016/j.smrv.2016.07.002. Search PubMedobs

[2] Apple Inc., Estimating Breathing Disturbances and Sleep Apnea Risk from Apple Watch, September 9, 2024. Retrieved October 21, 2024, from: https://www.apple.com/health/pdf/sleep-apnea/Sleep_Apnea_Notifications_on_Apple_Watch_September_2024.pdf

[3] U.S. Food and Drug Administration, Center for Drug Evaluation and Research. Samsung De Novo Classification Request for Sleep Apnea Feature, De Novo Summary (DEN230041), May 31, 2023. Retrieved October 21, 2024, from: https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230041.pdf

Facebook
Twitter
LinkedIn