Wearable technology promises to revolutionise health monitoring. But for millions of people (including those with darker skin, larger bodies, and chronic conditions) the data it produces can be dangerously inaccurate.
Wearable health technology (the category of devices that now encompasses smartwatches, fitness trackers, continuous glucose monitors, and pulse oximeters) has been one of the most celebrated technological developments of the past decade. The promise is compelling: real-time, personalised health data, available to anyone, that empowers individuals to understand their bodies and make better decisions about their health. The Apple Watch can detect atrial fibrillation. Continuous glucose monitors are transforming diabetes management. Wearable ECG devices are catching cardiac abnormalities that might otherwise have gone undetected for years.
The promise is real. But so are the problems, and they cluster, with troubling consistency, around the same groups of people: those whose bodies, health histories, or physiological characteristics were not centred in the design and testing of these devices.
The Pulse Oximeter Problem: A Case Study in Invisible Bias
Pulse oximeters, devices that measure blood oxygen saturation, became a household name during the COVID-19 pandemic, as they offered a way for people to monitor a key indicator of respiratory deterioration at home. Hospitals had relied on them for decades. What many users did not know, and what the medical literature had documented for years without any sense of urgency, was that pulse oximeters are significantly less accurate on darker skin tones.
A landmark study published in the New England Journal of Medicine in 2020 found that pulse oximeters were nearly three times as likely to miss low oxygen levels in Black patients compared to white patients. This had real consequences during the pandemic: patients with dark skin tones who appeared to have normal oxygen readings may in fact have been dangerously hypoxic. The FDA acknowledged the problem and issued guidance, but the devices have remained on shelves and in clinical use with only marginal improvements in calibration for diverse skin tones.
The reason for the discrepancy is physiological: pulse oximeters work by shining infrared light through the skin and measuring how much is absorbed by oxygenated versus deoxygenated haemoglobin. Melanin, the pigment that gives skin its colour, absorbs light differently, and the algorithms underpinning most commercial devices were calibrated primarily on lighter-skinned subjects. This was a known variable. It was simply not treated as a priority.
Heart Rate Monitors, Fitness Trackers, and Larger Bodies
The same optical principles that affect pulse oximetry also affect the photoplethysmography (PPG) sensors used by most consumer smartwatches and fitness trackers to measure heart rate. Research has found that these sensors are measurably less accurate on darker skin tones, particularly during exercise, when blood flow patterns change. For someone using a fitness tracker to monitor cardiovascular health, this is not a trivial error.
Body size and composition also affect accuracy in ways that the wearable industry has been slow to address. Devices designed primarily for average-bodied users may fit poorly on larger wrists, sit too loosely to maintain consistent sensor contact, or be calibrated on datasets that underrepresent people with higher body fat percentages. Research has found that calorie counting on fitness trackers, one of the most popular features, can be off by as much as 40% for some users, with the errors concentrated in those whose bodies differ most from the ‘average’ user on which algorithms were trained.
A pulse oximeter was nearly three times as likely to miss low oxygen levels in Black patients compared to white patients. This had real consequences during a pandemic. The error was known. It was simply not treated as a priority.
Women’s Health Tech: Progress and Persistent Gaps
The women’s health technology market (period trackers, fertility monitors, menopause apps, and hormone cycle tools) has grown substantially in recent years, fuelled by significant unmet need and, more recently, by concerns about data privacy following the overturning of Roe v. Wade in the United States. The quality of these tools, however, varies enormously.
Many period tracking apps were originally calibrated around a model of a 28-day menstrual cycle; a figure that is statistically average but individually unrepresentative for a large proportion of users, including those with polycystic ovary syndrome (PCOS), endometriosis, or irregular cycles related to other health conditions. Research published in NPJ Digital Medicine found that the accuracy of fertility prediction in popular period tracking apps was modest at best for the average user, and significantly lower for those with irregular cycles.
Black women, who face significantly higher rates of conditions like fibroids, PCOS, and preeclampsia, and who have historically been underrepresented in gynaecological research, have the least adequately designed health technology to support them. This is not a small gap: fibroids affect an estimated 70-80% of Black women by age 50, but the leading period tracking and women’s health apps have rarely been tested on populations with these conditions.
The Path to More Equitable Health Technology
The solutions are not mysterious. Clinical validation studies for health monitoring devices should be required to include sufficiently diverse populations as a condition of regulatory approval. Algorithms trained on non-representative data should be required to disclose those limitations clearly, including in consumer-facing materials. The FDA, MHRA, and equivalent regulators have begun to move in this direction, but the pace has been inadequate relative to the rate at which these devices are proliferating into clinical use.
At the industry level, diversifying the teams that design and test health technology would help. Research consistently shows that diverse development teams are better at anticipating the needs and vulnerabilities of diverse users. This is not a comfortable truth for an industry with well-documented diversity problems, but it is an important one.
And at the individual level, it is worth approaching wearable health technology with a degree of critical awareness. The data it provides can be enormously valuable. It can also be wrong, and knowing the difference may, for some people in some circumstances, matter more than the industry has acknowledged.