Slashdot Mirror
Fitness Trackers Out of Step When Measuring Calories, Research Shows (theguardian.com)
Fitness devices can help monitor heart rate but are unreliable at keeping tabs on calories burned, research has revealed. From a report on The Guardian: Scientists put seven consumer devices through their paces, comparing their data with gold-standard laboratory measurements. "We were pleasantly surprised at how well the heart rate did -- under many circumstances for most of the devices, they actually did really quite well," said Euan Ashley, professor of cardiovascular medicine at Stanford University and co-author of the research. "At the same time we were unpleasantly surprised at how poor the calorie estimates were for the devices -- they were really all over the map." The team tested seven wrist-worn wearable devices -- the Apple Watch, Basis Peak, Fitbit Surge, Microsoft Band, Mio Alpha 2, PulseOn, and Samsung Gear S2 -- with 31 women and 29 men each wearing multiple devices at a time while using treadmills to walk or run, cycling on exercise bikes or simply sitting.
There are too many variables involved in determining calories burned by any biological entity, and these 'fitness bands' are not the only device that has this problem, either. The closest you can come are devices that measure power generated by your muscles (PowerTap hubs, SRM or Quarq cranksets and other bicycle-mounted direct measuring instrumentation like them, rowing machines with an ergometer, etc), and even then there is a variable 'biological efficiency' term in the equation that means you can only call it an estimate. Some of the worst accuracy devices are things like the treadmills and stationary bikes in a public gym, which just use statistical averages of a range for calories burned given a level of exertion, and even then they tend towards the high end of the range to keep people motivated to continue using the treadmill or stationary bike. Something like these fitness bands work in a similar way, and I'd fully expect that they too tend to estimate on the high end of the range of 'calories burned' to keep you motivated. The fact that they track heart rate means it's a little more accurate, given one more term in the equation, and if they have a way to enter your bodyfat percentage, that would improve the estimation also. There are other factors you can plug into such an equation to make it more accurate, but in the end it's still just an estimate because of the efficiency factor. Therefore: none of what is being claimed here about it's lack of accuracy in 'calories burned' is terribly surprising. Of course for the average person, moving their body for significant amounts of time (not just getting off the couch, going to the 'fridge for another soda/beer, then back to the couch) is good regardless of trying to track 'calories burned', and to be quite honest, your dietary intake is more important when trying to lose excess bodyfat than exercise is. Of course you don't need any 'fitness bands' for that, just a decent pair of running shoes, so that doesn't make the company any money now does it?
I also feel like it should be fairly obvious, if you take the time to think about it. What can these devices measure? Heart rate and motion, generally. Some will ask you what kind of exercise you're engaged in, but it rarely gets more specific than "running" or "cycling". There might be a GPS, in which case, the device can know roughly how much distance you cover.
So take each individually:
* Heart rate: Different things can effect your heart rate to different degrees, and there's not a direct/exact correlation between hear rate and calories burned.
* Motion: One of these things just knows how much your wrist is moving, which often won't tell you much. Sit still and shake your arm a little, and then run around for an equal amount of time. The watch can't tell the difference.
* Kind of exercise: You can't know how many calories you've burned simply from the information, "You were riding a bike for 10 minutes." Were you going up-hill or down-hill? How heavy is your bike? How well maintained is your bike? What gear were you in? How bumpy is the road?
* GPS: It knows you went from point "A" to point "B", which gives a distance and theoretically an altitude and speed. Were you in a car or on a bike? Were you carrying anything? Were you doing jumping jacks while your travelled? When you went up 10 meters in altitude, did you climb the stairs or ride and escalator?
So on their own, none of these things would get you to an accurate reading. Admittedly, you could try to combine these measurements to make the reading more accurate. For example, if the device knows you're riding a bike, it can use the GPS to determine how fast you're going, whether you're going up-hill or down. It can measure your wrist movements and heart rate to guess about how hard you might be pedaling. However, doing this kind of calculation would probably take some machine learning to figure out how to combine these things for each different kind of exercise, and even then it would probably change for different people in different circumstances, requiring some kind of calibration. And even then, it might be fooled or confused somehow.
So I think, reasonably, all we can expect with the current technology is a vague estimate. However, that's not useless. If I use my fitbit every day, and yesterday it says I burned 200 calories, and then today it says I burned 2,000 calories, that is likely a good indication that I was far more active today than yesterday. Did I actually burn precisely 2,000 calories today? Probably not. Did I actually burn 10 times the number of calories as yesterday? That might be a decent estimate (I think that'd be worth studying). Did I do a significant amount of additional exercise today, probably improving my health? I think that's a safe bet, and mostly that's what people really need to know. After all, the numbers of calories listed on food packaging is also an estimate. If you're trying to line up your calories burned to calories eaten, you're fooling yourself.