In a detailed article from 2024, we outlined the pitfalls of testing wearables and their health features in particular. We pointed out that such studies are not secret but publicly accessible. While these studies can be more or less comprehensible for regular people, readers should be familiar with some basic terms for proper interpretation, which also applies to a newly published study from Apple. This study relates to the hypertension notification feature on the Apple Watch.
To be more exact, Apple now offers a high blood pressure warning and will soon offer it on other smartwatches as well. However, this is not a measurement that, after calibration or through a classic measurement with a cuff, displays specific blood pressure values. Instead, it is a screening intended to alert users that they possibly have high blood pressure.
According to Apple, the algorithm was trained on 3,216 subjects, validated on 3,878 people, and then tested on 2,236 people. Apple also lists information about the proportion of overweight participants, as is customary, and subjects are therefore categorized into different groups. In the results, Apple reveals the sensitivity and specificity.
In the validation of medical studies, sensitivity refers to how many people who are actually sick are correctly identified as sick by a (properly performed) diagnostic procedure. For Apple’s high blood pressure screening, the sensitivity is 41.2%, which means that less than half of people with hypertension have been alerted. Specificity came in at 92.3%*, which is not necessarily surprising. Specificity indicates how many healthy people are correctly identified as healthy. These two values are related. By shifting the decision threshold, the two values can be modified.
We don’t want to dive too deep into these results, especially not in comparison to competing products. However, we do want to point out that a higher sensitivity in such screening procedures generally seems more beneficial, even if it comes with a significant reduction in specificity. You have to consider the context. A false positive in cases of high blood pressure warnings does not lead to invasive, risky and possibly expensive procedures. It only leads to a completely harmless conventional blood pressure measurement, which can perhaps even be performed by users themselves with inexpensive blood pressure cuffs. We cannot determine if the high blood pressure detection on the Apple Watch fails due to a lack of sensitivity or simply technical inadequacies.
*The individual values each represent the center of the 95% confidence interval (37.2-45.3%, 90.6-93.7%) across all subject groups.
I have been active as a journalist for over 10 years, most of it in the field of technology. I worked for Tom’s Hardware and ComputerBase, among others, and have been working for Notebookcheck since 2017. My current focus is particularly on mini PCs and single-board computers such as the Raspberry Pi – so in other words, compact systems with a lot of potential. In addition, I have a soft spot for all kinds of wearables, especially smartwatches. My main profession is as a laboratory engineer, which is why neither scientific contexts nor the interpretation of complex measurements are foreign to me.
My fascination for technology goes back a long way to the Pentium II era. Modding, overclocking and treasuring computer hardware has since become an integral part of my life. As a student, I further developed a keen interest in mobile technologies that can make the stressful college life so much easier. After I fell in love with the creation of digital content while working in a marketing position, I now scour the web to bring you the most exciting topics in the world of tech. Outside the office, I’m particularly passionate about motorsports and mountain biking.
Source link
