Study finds Apple Watch able to be used as diagnostic tool for sleep apnea, hypertension

Posted:
in Apple Watch
Researchers from the University of California San Fransisco in conjunction with the developers of the Cardiogram app have proven that using existing sensors, the Apple Watch can detect sleep apnea and hypertension with a high degree of accuracy, potentially opening up the ability for the device to be used as a long-term research tool for the conditions.




The study sponsored by the University of California surveyed 6115 participants with an Apple Watch through the Cardiogram app. A machine learning algorithm called "DeepHeart" was used to sift through the data for 70 percent of the study participants to diagnose the remaining 30 percent.

"The idea here is that by screening continuously you would identify people with hypertension who might not know they have it," said Cardiogram co-founder and study lead Johnson Hsieh told TechCrunch in an interview. "Thenm you'd guide them through the appropriate final diagnosis, which would be through a blood pressure cuff and then treatment."

Sleep apnea was detected with a 90 percent accuracy. Hypertension was diagnosed with an 82 percent accuracy.

Sleep apnea is a condition where the afflicted can stop breathing in their sleep. An estimated 22 million adults are affected in the U.S., with misdiagnoses common. About 75 million Americans are afflicted with hypertension -- a major risk factor for stroke and heart disease.

Comments

  • Reply 1 of 17
    bb-15bb-15 Posts: 283member
    Because of this article, I downloaded the Cardiogram app on my iPhone and have it running on my Apple Watch Series 2.
    Thanks. 
    doozydozenjSnivelywatto_cobraracerhomierusswjony0
  • Reply 2 of 17
    This will be where the watch ultimately is going. Collecting data about our health. 
    watto_cobrarusswJWSCjony0
  • Reply 3 of 17
    My Apple Watch charges and night and gets worn during the day.

    I suppose I could buy a second Apple Watch to charge during the day and wear during the night. :neutral: 
    edited November 2017 king editor the graterandominternetperson
  • Reply 4 of 17
    When I had my Series 1 Apple Watch, I put it on the charger for an hour before bed then for half an hour in the morning top up the charge to get thru the day. So I was wearing it 22 hours plus a day.
    edred
  • Reply 5 of 17

    Am I the only one petty enough to notice and be bothered by the passive voice in the article title?  "... able to be used as ..." seems so awkward.  Why not "... can be used as ..."?

    Yeah, I'm that guy.

    spheric
  • Reply 6 of 17
    I’ve worn my Apple Watch at night sinc the original (now on series 2). Like @Just_Iain, I would charge alternating evenings/mornings as my battery would last around 30 hours. With my current, I get around 30 when exercising, 48 when not doing anything.

    by altering charging habits, it’s easy to sleep with the watch. Just be sure to turn on theater mode at night so you don’t annoy your wife with a watch face flash in bed.
  • Reply 7 of 17
    I'm still a bit confused, so does this app read the existing data collection or would I need to have it constantly running?
  • Reply 8 of 17
    GeorgeBMacGeorgeBMac Posts: 11,421member
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
  • Reply 9 of 17
    zimmiezimmie Posts: 651member
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    davenrandominternetpersonJWSCsphericjony0
  • Reply 10 of 17
    Rayz2016Rayz2016 Posts: 6,957member

    Am I the only one petty enough to notice and be bothered by the passive voice in the article title?  "... able to be used as ..." seems so awkward.  Why not "... can be used as ..."?

    Yeah, I'm that guy.

    Well, I wasn’t until you mentioned it…
    russwJWSCbeowulfschmidt
  • Reply 11 of 17
    dewmedewme Posts: 5,684member
    stpat said:
    I'm still a bit confused, so does this app read the existing data collection or would I need to have it constantly running?
    I agree that the article is a little confusing and even a bit misleading, at least if you are expecting to install an app that will identify the wearer as being at-risk for hypertension and/or sleep apnea. The potentially important implication of this study, and one that still needs to be further qualified with additional studies, is that a statistically meaningful diagnosis for hypertension & sleep apnea is possible using only the sensors currently installed in the Apple Watch (and some other brand wearables I presume). This is significant because the Apple Watch only provides a fraction of the sensor data that is currently collected using existing diagnostic techniques such as formal sleep studies which use a much wider array of sensors and data collection techniques. So it's really more a proof-of-concept study to identify the absolute minimum amount of sensor data and analysis necessary to perform a useful diagnosis for two specific conditions. The proof-of-concept is backed with statistical rationalization using big-data analysis to reinforce the claims.

    This study is important to Apple because it provides them with positive feedback that even in its current form & function the Apple Watch has the potential to effectively contribute to the health, wellness, and comfort of its wearers. No doubt that Apple will continue to expand both the range of sensors and sophistication of health applications on the Apple Watch to make it useful for a much wider range of health diagnoses. This provides a focus for the Apple Watch that goes so far beyond traditional watch functions, much in the same manner that the iPhone is so much more than just a phone. 

    One last word of advice: if you have a concern about sleep apnea do not install the Cardiogram app expecting it to help you diagnose this potentially deadly condition. Talk to your doctor and follow his/her advice about undergoing a sleep study. One critical thing the Apple Watch does not currently measure is blood oxygen levels. When you stop breathing due to sleep apnea your blood oxygen levels plummet - which is extremely damaging to your heart, brain, and other internal organs.  
    larryjwJWSCjony0
  • Reply 12 of 17
    GeorgeBMacGeorgeBMac Posts: 11,421member
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
  • Reply 13 of 17
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
    Don’t forget that the Apple Watch also has a microphone and therefore can be used in conjunction with the oxygen monitor to pick up apnea which often results in snoring.
  • Reply 14 of 17
    zimmiezimmie Posts: 651member
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
    There absolutely are other things which could cause these same symptoms. The other things are vastly less common, though. I think this kind of tool could be enormously helpful right now as a rule-in test to direct further investigation. “Your watch noticed this unusual situation which may be affecting your sleep. You should talk to a doctor and get checked for apnea.”

    It is definitely possible to detect some medical conditions with extremely high confidence based only on indirect evidence. Microsoft published a study (https://blogs.microsoft.com/ai/2016/06/07/how-web-search-data-might-help-diagnose-serious-illness-earlier/) in which they detected pancreatic cancer (5-15% detection rate, 0.001% estimated false positive rate) based solely on a user’s Bing searches. The evidence from this trial (apnea/hypertension detection) needs to be confirmed, of course. That said, right now, it strongly suggests that apnea and hypertension can both be diagnosed with reasonably high accuracy from much less information than previously thought.

    The quantified self movement has already yielded an enormous corpus of data from a lot of individuals. This is much, much higher-resolution than the point-in-time tests doctors typically use to diagnose a condition. I would be shocked if we didn’t find things easier to detect than previously thought.
    jony0
  • Reply 15 of 17
    zimmie said:
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
    There absolutely are other things which could cause these same symptoms. The other things are vastly less common, though. I think this kind of tool could be enormously helpful right now as a rule-in test to direct further investigation. “Your watch noticed this unusual situation which may be affecting your sleep. You should talk to a doctor and get checked for apnea.”

    It is definitely possible to detect some medical conditions with extremely high confidence based only on indirect evidence. Microsoft published a study (https://blogs.microsoft.com/ai/2016/06/07/how-web-search-data-might-help-diagnose-serious-illness-earlier/) in which they detected pancreatic cancer (5-15% detection rate, 0.001% estimated false positive rate) based solely on a user’s Bing searches. The evidence from this trial (apnea/hypertension detection) needs to be confirmed, of course. That said, right now, it strongly suggests that apnea and hypertension can both be diagnosed with reasonably high accuracy from much less information than previously thought.

    The quantified self movement has already yielded an enormous corpus of data from a lot of individuals. This is much, much higher-resolution than the point-in-time tests doctors typically use to diagnose a condition. I would be shocked if we didn’t find things easier to detect than previously thought.
    100 some years ago we had snake oil salesmen selling ineffective potions promising to cure what ails ya.   We drew a line in the sand that said health claims need to be backed by solid evidence showing their reliability.  It "might work in some situations" does not meet that test.  It doesn't cross the line. 
    ...  I would hesitate before erasing that line just because it is being done with an Apple product.

    That said:  I am enthusiastically supportive of things like Health where real, actual data such as heart rate, steps, distance, time, etc...  are stored and can serve as a foundation for promoting health.
  • Reply 16 of 17
    zimmiezimmie Posts: 651member
    zimmie said:
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
    There absolutely are other things which could cause these same symptoms. The other things are vastly less common, though. I think this kind of tool could be enormously helpful right now as a rule-in test to direct further investigation. “Your watch noticed this unusual situation which may be affecting your sleep. You should talk to a doctor and get checked for apnea.”

    It is definitely possible to detect some medical conditions with extremely high confidence based only on indirect evidence. Microsoft published a study (https://blogs.microsoft.com/ai/2016/06/07/how-web-search-data-might-help-diagnose-serious-illness-earlier/) in which they detected pancreatic cancer (5-15% detection rate, 0.001% estimated false positive rate) based solely on a user’s Bing searches. The evidence from this trial (apnea/hypertension detection) needs to be confirmed, of course. That said, right now, it strongly suggests that apnea and hypertension can both be diagnosed with reasonably high accuracy from much less information than previously thought.

    The quantified self movement has already yielded an enormous corpus of data from a lot of individuals. This is much, much higher-resolution than the point-in-time tests doctors typically use to diagnose a condition. I would be shocked if we didn’t find things easier to detect than previously thought.
    100 some years ago we had snake oil salesmen selling ineffective potions promising to cure what ails ya.   We drew a line in the sand that said health claims need to be backed by solid evidence showing their reliability.  It "might work in some situations" does not meet that test.  It doesn't cross the line. 
    ...  I would hesitate before erasing that line just because it is being done with an Apple product.

    That said:  I am enthusiastically supportive of things like Health where real, actual data such as heart rate, steps, distance, time, etc...  are stored and can serve as a foundation for promoting health.
    Treatments and diagnostics are two radically different things. I can't believe I have to point this out to someone in the medical profession.
  • Reply 17 of 17
    zimmie said:
    zimmie said:
    zimmie said:
    As the pharmaceutical industry has demonstrated, you can prove pretty much anything you want with "a study".

    Serious health care professionals have learned to ask for the details and that the results be repeated independently.   It's why they only recognize "peer reviewed" reports.

    For myself, I am question how well a heart rate monitor could diagnose either apnea or (and especially) hypertension.   (However, there is a possibility that the sensors are picking up more than just heart rate).  In the meantime, I look forward to Apple adding additional sensors to the watch to measure things such as these and others... 
    When you stop breathing in your sleep, your body dumps a little adrenaline into your blood to wake you up and get you breathing again. This would show up as a spike in heart rate and some physical activity while otherwise stationary.

    Hypertension is a little more complicated. It is characterized by reduced heart rate variability, and the beats tend to be harder. Strictly, the sensor on the Apple Watch isn't a heart rate sensor, it's an optical blood flow sensor. Heart rate can be derived from the data it records, but so can flow volume.
    That is perhaps true...
    But, as a medical professional, I shudder at the thought of making a diagnosis based on such indirect physiologic manifestations.   Essentially, there are a dozen other things that can cause the same effects.

    It is why the FDA is so conservative in issuing approvals for medical devices.   They need to have reasonable confidence that their results are what they say they are.  
    There absolutely are other things which could cause these same symptoms. The other things are vastly less common, though. I think this kind of tool could be enormously helpful right now as a rule-in test to direct further investigation. “Your watch noticed this unusual situation which may be affecting your sleep. You should talk to a doctor and get checked for apnea.”

    It is definitely possible to detect some medical conditions with extremely high confidence based only on indirect evidence. Microsoft published a study (https://blogs.microsoft.com/ai/2016/06/07/how-web-search-data-might-help-diagnose-serious-illness-earlier/) in which they detected pancreatic cancer (5-15% detection rate, 0.001% estimated false positive rate) based solely on a user’s Bing searches. The evidence from this trial (apnea/hypertension detection) needs to be confirmed, of course. That said, right now, it strongly suggests that apnea and hypertension can both be diagnosed with reasonably high accuracy from much less information than previously thought.

    The quantified self movement has already yielded an enormous corpus of data from a lot of individuals. This is much, much higher-resolution than the point-in-time tests doctors typically use to diagnose a condition. I would be shocked if we didn’t find things easier to detect than previously thought.
    100 some years ago we had snake oil salesmen selling ineffective potions promising to cure what ails ya.   We drew a line in the sand that said health claims need to be backed by solid evidence showing their reliability.  It "might work in some situations" does not meet that test.  It doesn't cross the line. 
    ...  I would hesitate before erasing that line just because it is being done with an Apple product.

    That said:  I am enthusiastically supportive of things like Health where real, actual data such as heart rate, steps, distance, time, etc...  are stored and can serve as a foundation for promoting health.
    Treatments and diagnostics are two radically different things. I can't believe I have to point this out to someone in the medical profession.
    Brilliant!   I'm glad you figured that out!
    ... Now what was your point?
Sign In or Register to comment.