The Shortcoming of Remote Monitoriing

Health tech keeps making the same mistake: it measures the right things but ignores the context that makes them meaningful.

Wearables collect time-stamped data from your body all day and all night. Heart rate variability, skin temperature, respiratory rate, sleep architecture. Every one of these signals follows a 24-hour circadian rhythm. A "low" HRV at 2pm means something completely different from the same reading at 2am. Your resting heart rate in the early evening isn't the same biological event as your resting heart rate during deep sleep.

But most wearables present these as flat scores against population averages, as if your biology were static. No wonder users report feeling anxious, confused, or pressured by numbers that fluctuate "without explanation." The fluctuations aren't random. They're circadian. The device just doesn't tell you that.

Now a new wave of wearables (Clair, Peri, Level Zero) is trying to estimate hormones, estrogen, progesterone, LH, cortisol, from these same proxy signals. The ambition is real. But here's what nobody in that conversation is saying: every proxy signal they're using is a downstream circadian biomarker, and the hormones they're estimating are themselves circadian outputs. Cortisol peaks about 30 minutes after waking. Progesterone rises are nocturnal. The LH surge has specific timing architecture.

Validating these estimates against a single morning blood draw is the wrong gold standard for a continuous monitor. You'd want concordance across the full 24-hour cycle.

A cortisol reading without circadian context is noise, not signal.

This connects to something I keep seeing in health AI research too. Most studies test whether AI can match a doctor's diagnostic accuracy using clean clinical vignettes. But that's answering a question nobody is actually asking. The real question is: when a scared, confused person types symptoms into ChatGPT at 2am, does that interaction lead to better decisions than what they would have done otherwise?

Same problem. Different domain. The technology works in controlled conditions. But health doesn't happen in controlled conditions. It happens at 2am, mid-cycle, post-travel, under stress, in bodies that run on clocks researchers keep forgetting to account for.

The next generation of health tech, whether wearables or AI, won't win by adding more sensors or more parameters. It will win by understanding context: biological timing, individual variation, and the messy reality of how humans actually live with their data.

Your body has a clock. Your health tools should know how to read it.

https://www.linkedin.com/posts/gillesfrydman_circadianhealth-wearables-digitalhealth-activity-7429141255684513792-QfAl?utm_source=social_share_send&utm_medium=member_desktop_web&rcm=ACoAAAAqTbwBl7WHwBdULQ1iB1ThcCkr32EMVjE