Pregnancy according to My Apple Watch
Pregnancy according to My Apple Watch
In 2023, my partner and my mother gave me an Apple Watch. My initial mental model of a smartwatch was “notification buzzer I have to charge every night,” which was not entirely wrong. But I wore it, and after a few days the neuroscientist in me started poking around in the Apple Health data. A wrist accelerometer estimating VO₂ max from gait is not, in principle, that different from what I do with flies. The hardware is sleeker; the subject is larger and less amenable to anesthesia.
So I started treating myself as a longitudinal dataset: VO₂ max as a rough fitness proxy, resting heart rate as a recovery signal, walking metrics as things I did not yet have a clean hypothesis for but was happy to let accumulate. What I did not anticipate was that the watch would record my second pregnancy almost from the beginning: gestation, birth, postpartum, and the slow, still-incomplete process of getting my body back under me. I also did not anticipate that I would wait so long to look at the full time series. For someone whose job is looking at longitudinal data, I was impressively late to my own dataset. Postpartum haze remains undefeated.
When I finally pulled the data together — VO₂ max, resting heart rate, gait symmetry, aligned from 2023 to 2026 — I was surprised by how plainly pregnant I was in the data. Long before I sat down to analyze anything, the watch had just been recording.
The broad story is recognizable. Apple Watch estimates VO₂ max from heart rate and motion during outdoor walks. Mine peaked at 38.6 ml/kg/min in June 2023, declined through pregnancy, and bottomed out at 21.2 in April 2025, about eight weeks postpartum. Resting heart rate rose across the same window. Both are recovering now. Neither is back. If this were the only layer of the plot, it would be a fairly clean pregnancy-and-postpartum recovery story: cardiovascular load goes up, estimated fitness goes down, then the traces begin to move back toward baseline.
The gait metrics are where it gets more interesting, because they do not tell the same story. Walk asymmetry measures the difference in timing between left and right foot contact. Mine increased roughly fourfold during the second and third trimesters, which makes sense: changing center of mass, changing load, changing mechanics. After pregnancy ended, the asymmetry mostly resolved. The cause was mechanical, and when the cause went away, the signal did too.
Double support time is different. It measures how much of each stride is spent with both feet on the ground; higher values usually mean a slower, more cautious gait. Mine increased during pregnancy and has not really come back down. The watch did not know I had plantar fasciitis. It did not know about foot pain, physio referrals, or the fact that I now think about footwear like an old person with a spreadsheet. It only captured the behavioral consequence: I spend more of each stride with two feet on the ground than I used to. That is often what measurement is. The sensor does not know the story. It records the shape of the compensation.
To make this more explicit, I put three metrics into a state space: VO₂ max, resting heart rate, and one locomotor metric. Both plots use the same first two axes. The only thing I change is the third. With walk asymmetry as the locomotor axis, the trajectory almost closes. The body leaves the pre-pregnancy region, arcs through pregnancy and postpartum, and comes most of the way back. If this were the only gait metric I looked at, I might conclude that recovery is mostly complete.
With double support time as the locomotor axis, the loop stays open. Same VO₂ max, same resting heart rate, different locomotor metric, different conclusion. That is the part I like, and also the part that makes me twitchy. Recovery is not a single number. It depends on which measurement you trust to represent the thing you care about. Asymmetry says one perturbation resolved. Double support says another one is still present.
The dashed line in each plot connects “now” to my 2023 peak. I have not closed that gap. VO₂ max is still about eight points below where it was. Double support time still has not returned. The watch does not know about kettlebell sessions, pull-up progress, sleep debt, toddlers, weaning, or the quiet indignity of plantar fascia. It just keeps recording. I will write a follow-up if the ring closes.
Technical note: Data are from Apple Health export. I used daily aggregated CSVs only, not raw health records. Smoothed trends use a 30-day Hamming window on a linearly interpolated daily grid. VO₂ max coverage is sparse, about 11% of days and roughly three readings per month, so the smoothed line interpolates between observations. Interactive notebook: Observable ↗.