Fitbit's 150B hours of heart data • Hacker News Board

Re: Fitbit's 150B hours of heart data

#21

Post by wjnc » Wed, Sep 05, 2018, 8:41 AM UTC

They missed the sampling effect across ages. I've seen it in action quite a few times already. Prior: Conditional on Having a fitbit, the 70 yr old is a larger deviant from average health, than the 30 yr old. That effect could easily explain away the falling RHR with older age. The population in that age group is getting more healthy by self-sampling. The proposed solution in the article (the use of beta-blockers) seems inadequate. What proportion of the population and the Fitbit-users is on those medicines? (Article does mention some stats in the multicountry comparison.)

As always the bayesian/multilevel approach could help: buying a Fitbit is a decision with information on your health-status in itself, model it likewise. Add information on weigth, BMI, activities per week etc. as indicators of in which class of users you fall (the uberfit, healthy, the aspiring, ...). Then use the predicted RHR per age instead of the data itself.

Re: Fitbit's 150B hours of heart data

#22

Post by lostlogin » Wed, Sep 05, 2018, 8:42 AM UTC

com2kid wrote: post #6

Bishonen88 wrote: post #5

Earlier quoted context omitted.

Am a Garmin user for years. Was in a hospital ~10 months ago and was connected to a heart-rate-monitoring-thingy via a finger-clip. At the same time I had my Garmin Fenix 3 on my other arm. Was looking at the screen and comparing it to what my watch was showing and it was +-2 points at all times. Tried holding my breath to see what happens and both were almost identical.

Optical heart rate when not moving is pretty easy to do. Measurements are much harder when the user is moving around erratically, such as is typical with exercising. Source: I worked on a fitness product using optical HR for 3 years.

My experience of hospital grade equipment for monitoring the heart is that it isn’t great with movement either. This was in an MR scanner (magnetic field and RF) so I ones milage may vary.

Re: Fitbit's 150B hours of heart data

#23

Post by kendallpark » Wed, Sep 05, 2018, 8:51 AM UTC

Yes, there are cases when a sub-60 resting heart rate is bad. For the vast majority of people, lower is better. The athlete's heart undergoes remodeling that is similar in outcome to pathological processes (eg, cardiomegaly and cardiac hypertrophy), but is in fact healthy, not harmful.

https://en.wikipedia.org/wiki/Athletic_heart_syndrome

I have personally found that average RHR per week tends to correlate strongly with my own physical activity. With consistent workouts, I can get down to 47 bpm. When my RHR gets above 55, I know I've been slacking off for too long (or it's an exam week and I'm stressed).

Re: Fitbit's 150B hours of heart data

#24

Post by kendallpark » Wed, Sep 05, 2018, 8:56 AM UTC

saagarjha wrote: post #4

> modern smartwatches and fitness bands can track your pulse continuously, day and night, for months. Imagine what you could learn if you collected all that data from tens of millions of people! Please no. Surely nobody thought collecting health data from all their users was a good idea, right? > That’s exactly what Fitbit (FIT) has done. It has now logged 150 billion hours’ worth of heart-rate data. From tens of mil…

> "Anonymized". Just with the data posted above, it seems like it'd be pretty easy to correlate the data to a person. And auto-enrolling everyone who buys a health product is not an excuse to mine all their data.

Hospitals are allowed to anonymize protected health information in their medical records to mine the data (they can distribute it too) without the patient's consent. PHI is way more protected than consumer health information, which is what Fitbit has.

Re: Fitbit's 150B hours of heart data

#25

Post by stef25 » Wed, Sep 05, 2018, 8:57 AM UTC

kendallpark wrote: post #23

Yes, there are cases when a sub-60 resting heart rate is bad. For the vast majority of people, lower is better. The athlete's heart undergoes remodeling that is similar in outcome to pathological processes (eg, cardiomegaly and cardiac hypertrophy), but is in fact healthy, not harmful. https://en.wikipedia.org/wiki/Athletic_heart_syndrome I have personally found that average RHR per week tends to correlate strongly w…

How old are you?

How many work outs or what frequency thereof does it take to notice lowering of your RHR ?

Re: Fitbit's 150B hours of heart data

#26

Post by daniel_iversen » Wed, Sep 05, 2018, 9:04 AM UTC

Zarath wrote: post #20

daniel_iversen wrote: post #11

Earlier quoted context omitted.

This large scale data analytics is probable the best (or at least usually fastest) way to solve the worlds big health challenges I would guess. Nobody forced anyone to wear a smart watch or health tracker. I for one thinks the world is slightly better off because of companies like Fitbit!

The world may be better off, but certainly not because they massively collect biometric data, but because their product encourages exercise.

Both right? You may not like it but which research wouldn’t benefit from a large useful dataset? And to the post above re: assuming consent, sure maybe they should make it crystal clear up front how they’ll use your data.

Re: Fitbit's 150B hours of heart data

#27

Post by scotty79 » Wed, Sep 05, 2018, 9:11 AM UTC

saagarjha wrote: post #4

> modern smartwatches and fitness bands can track your pulse continuously, day and night, for months. Imagine what you could learn if you collected all that data from tens of millions of people! Please no. Surely nobody thought collecting health data from all their users was a good idea, right? > That’s exactly what Fitbit (FIT) has done. It has now logged 150 billion hours’ worth of heart-rate data. From tens of mil…

Do you really think than having less data about human health is better ... because ... privacy?

That goes against any progress that was ever achieved in medicine. Or do you have some way of reconciling that man sticking a finger up another man's rectum is good and private but keeping records of something like a heartrate throughout the day is ... im-private? ... and bad?

Re: Fitbit's 150B hours of heart data

#28

Post by kendallpark » Wed, Sep 05, 2018, 9:13 AM UTC

stef25 wrote: post #25

kendallpark wrote: post #23

Yes, there are cases when a sub-60 resting heart rate is bad. For the vast majority of people, lower is better. The athlete's heart undergoes remodeling that is similar in outcome to pathological processes (eg, cardiomegaly and cardiac hypertrophy), but is in fact healthy, not harmful. https://en.wikipedia.org/wiki/Athletic_heart_syndrome I have personally found that average RHR per week tends to correlate strongly w…

How old are you? How many work outs or what frequency thereof does it take to notice lowering of your RHR ?

Young adult. It takes about 1-2 weeks to notice a trend. I'm just looking over my summer RHR data and it pretty much matches my activity level. Steady decline in May and June when I was working out consistently. Steady climb after I injured my knee and had to lay off a bit. Decline when I got back on the exercise horse. Incline towards the end of the summer when workouts became rare due to wrapping up projects and traveling.

On a good week, I'll do 1-2 hours of cardio 5x a week, maybe with some HIIT thrown in. On a bad week I'll do nothing.

Re: Fitbit's 150B hours of heart data

#29

Post by ack0 » Wed, Sep 05, 2018, 9:18 AM UTC

rawoke083600 wrote: post #3

Now if only it were accurate...

I think this is the real problem.

My wife was worried her heart rate according to fitbit was high. Her doctor measured it at 15-20 bpm lower.

After she started doing up the fitbit band one notch tighter the average heart rate it recorded dropped 10 bpm.

Re: Fitbit's 150B hours of heart data

#30

Post by rando444 » Wed, Sep 05, 2018, 9:42 AM UTC

ack0 wrote: post #29

rawoke083600 wrote: post #3

Now if only it were accurate...

I think this is the real problem. My wife was worried her heart rate according to fitbit was high. Her doctor measured it at 15-20 bpm lower. After she started doing up the fitbit band one notch tighter the average heart rate it recorded dropped 10 bpm.

Is that really a problem with the equipment though?

They explain exactly this in the booklet it comes with.

If a sphygmomanometer gives an inaccurate reading because the doctor doesn't put the cuff on properly is the machine to blame?

I only say this because I've tested both a fitbit, and a garmin on each arm while wearing a chest strap and found both results to be within a satisfactory deviation of the chest strap.