Loads of people love their FitBits and smartphone step counters. One of my FB friends posts her step counts every day, and I think it’s great that she’s working on health-according-to-her and asking for (and getting) support. Getting in your obligatory 10,000 steps has become the new “apple a day” folk wisdom, although there is not general agreement about this among fitness and medical experts. But we’ll save that for later.
In a recently published article, researchers present results suggesting that smartphone (Apple and Android) activity-tracker apps are inaccurate for some parts of the population in some activities. For example, in cases of:
- slower walking speeds;
- variable gaits or walking patterns;
- location and walking surface;
- larger-sized walkers;
- short-distance walks;
there was significant undercounting of steps. Note: they did two very small studies (16 and 32 participants, respectively). The second study included people with BMI >30. They found that the smartphone apps had massive errors ranges (0–200% of steps taken). However, a big concern was underreporting of walks that fall under the list I gave. Why worry about this? Here are a couple of reasons:
First, what app you are using and which OS update you’re running WILL affect your step counts. Not only that– here’s what they say below:
Compared to the generally positive performances reported for laboratory-based treadmill walking, the extraordinarily large error ranges imply that the measured gait kinematics (which depend on walking speed, phone position, gender, body mass index, age, and race) may strongly influence the percentage of steps counted by consumer-grade smartphone apps. Furthermore, the differences in percentage of steps counted between the 2017 and 2018 studies suggests that phone up- dates and different app versions could also potentially change the outcomes of any smartphone-based assessments.
In short, treadmill walking is not like walking in the world. We use lab results (and rightly so) to develop ways to solve real-life problems and create new ways to help us in our life and health goals. But real-life mileage varies– sometimes a lot. This is important information, and if it’s inaccurate, that doesn’t help us with our own goals.
Laboratory-assessed and daily-life gaits are different [18]. The larger error ranges for slow or short walks observed in both studies are problematic because in daily-life people walk at slow and fast cadences (according to bimodal distributions) [20] and complete far more short walks than long walks.
Second, the norm that smartphone fitness tracker technology uses to count steps seems to undercount the kinds of steps taken by:
- females;
- larger people;
- non-white people;
- people with different gaits, including those with various dis/abilities
- slower walkers;
- short walks.
With respect to large-scale physical activity monitoring, we need to be mindful that unconscious bias embedded into black box smartphone apps may favour stereotypical (tread-mill like) and faster walking. These issues present a risk of systematically undercounting steps for slower [and shorter walks; walks by larger people, females, people from different racial or ethnic groups, people with non-stereotypical gait, and different preferred phone positions.
That’s not good news. Most of the world is not male, white, 18–25 years old, and with a BMI of 18–22, walking at least 5k/hour in a regular gait on a treadmill. This is a problem with most exercise/physical activity research, and we’ve written about this often.
But here’s a third problem that really worries me. All this data from smartphone apps is used to make health claims and recommendations for the whole population, and to make clinical judgments about individuals’ health status. Here’s what they say:
We must stay mindful that the black box models sometimes used to analyse big data may reinforce society’s existing biases. This may include the sometimes inflated link between physical inactivity and obesity.
We don’t want inaccurate data to drive health policy and clinical practice. There is overwhelming evidence that physical activity improves both health and longevity, but doesn’t significantly affect body weight. We do not want underreported data that disproportionately underreports the physical activity of already vulnerable groups (like those with disabilities) and populations already affected by health disparities (like racial minorities). This misinformation could be used to blame groups for their health conditions.
Big caveat here: this is only one study. As they say, further research is needed. So I’m not saying to throw away that very nice expensive Apple watch yet… But as we pursue health-to-us in all the ways we do, maybe paying attention to the more qualitative (in addition to the quantitative) aspects of activity is in order. That is, noting that when you ride, walk, chase your dog, swim, dance, stretch, lift, paddle, swing, you’re integrating activity in your way in your life.
And all that counts, no matter what the fitness apps say.
You’re right. We tend to think about accuracy and individuals. I say I don’t care if my phone undercounts because it’s just a matter of judging one day against the next. But when we share and aggregate data, accuracy matters. Like if we think on the basis of tracker data that women and larger people take fewer steps when that isn’t true. Great post. Good food for thought.
Thanks– this was a first cut at some bigger issues, which we are all interested in and affected by. The aggregation and use of possibly inaccurate data for health policy and maybe insurance (in the US) is a chilling thought.
We proved this yesterday. Phone step counters undercount! My partner’s Apple phone app indicated 2.9 miles, 480 feet elevation gain; my cell phone Strava (GPS based) indicated 3.1 miles, 739 ft. elevation gain. And we were walking together the entire time! This was on the popular Tumamoc Hill here in Tucson.
Hi– How interesting that there was that much variation! It doesn’t matter much in the individual case, but adding it up it presents distorted accounts of different folks’ activity.