Many data scientists are excited at the prospect of using health trackers and wearable technology to perform health research. The hope is that this data will shed light on risk factors and predictors of disease in a manner that was never possible before. However, there are valid concerns that this sample size isn’t representative of the entire population.
Socioeconomics has a profound impact on the health of the population, whether through access to healthcare services, education, work, and familiar environment, etc. The list of things that negatively impact low socioeconomic populations is painstakingly long and unfair. The health disparity between high and low socioeconomic populations is large. Are wearable technologies adding to that disparity?
First and foremost, wearable technologies are expensive. You need a smartphone and need to spend well over $100 just to get a device. Then, many wearable techs come with “premium” subscriptions if you truly want the full value of your device. They are very much a luxury item that a person only purchases if they are passionate enough about tracking their health and if they have the means to purchase one.
A person who is struggling financially is not going to rush out to buy the newest FitBit or Garmin watch. The cost barrier (at this point in time) makes it so that people who are using wearable technologies are already at an advantaged place both in life and, consequently, are in a better health position. In fact, a 2016 study showed that the largest income group of people buying Fitbit trackers were those who had an annual household income of $100,000. This means that the average population of wearable users are representative of the most privileged class. The inaccessibility of these devices skews the data toward the most privileged and away from those who need health information the most.
Another issue is diversity. In North America, the low socioeconomic populations are disproportionately minority populations that statistically have worse health outcomes. If these groups of people then don’t have access to wearable technology, then the data is not representative of their population. This heavily skews the wearable tech data to be biased towards the privileged. Those who truly need this data are then not included in the analytics. Their healthcare would then not be accurately tailored to their unique health situations.
What about age? Wearable tech is exactly that - tech. This creates an age and knowledge barrier that further influences the skew of user data. If a person doesn’t understand technology, then they will be less likely to want to purchase a wearable device. In 2016, the average age of the FitBit consumer was 35 to 44 years old. If the average user falls within one age cohort then the data is not representative of many other age cohorts such as older adults, children, and adolescents.
Then there is the issue of gender. In 2016, 72% of Fitbit consumers identified as female. As such, the data is skewed toward one gender. Gender bias is already present in many areas of medical research. So, we need to work hard to push against this type of bias if we are to use this type of Big Data in medical research.
To truly have representative health data, we need to alleviate these barriers to access to wearable tech and health tracking. We need to increase the diversity of the health data we are collecting by increasing access to populations such as low socioeconomic status, older adults, rural populations, and so on. Only then will we truly have a big picture of health that will be most beneficial for our whole population.