Your HRV Recovers at Altitude Before Your Oxygen Does

Athletes travel to high altitude for performance benefits; others have to cope with it to climb mountains, and some of us just have to deal with it because we like to holiday in beautiful places. I used to go on altitude training camps and have climbed mountains. I always struggled with the “thin air”, and it took days to feel better. One holiday took me to the top of Kilimanjaro, and the low air pressure made me so sick I can barely remember standing at the summit!

Am I alone in my terrible ability to cope with low air pressure? This week's analysis gave me the chance to find out. Can we see the physiological signs of altitude show up on a population level in wearable data?

The physiology is straightforward. As elevation increases, the partial pressure of oxygen decreases, so less O₂ crosses from the alveoli into the blood, resulting in lower arterial oxygen saturation. Acutely, that hypoxemia is compensated for by an increase in breathing rate. That ventilatory drive is itself a form of physiological stress, and we would also expect a more sympathetic overnight compensation: higher heart rate and lower HRV. Over a period of a few days, as the body adapts (in a few different ways), those markers should drift back toward baseline. There may also be an increased sleep drive as part of that adaptation. More time asleep to support recovery under hypoxia.

I used wearable data to follow travellers who spent time at low elevation (activity median under 500 m), then moved into altitude inferred from the same day’s GPS activity (median elevation at least 1,000 m). I required a clean within-person baseline: at least seven lowland sleep nights in the 21 days before arrival, plus a 14-day washout after any prior high stay. The result is a large natural experiment: 14,040 users, 20,648 trips, and 50,572 altitude nights. Median trip elevation was about 1,560 m, ski-town and mountain-city altitudes, not Everest Base Camp, which makes the signals more relevant to ordinary travel than extreme expedition physiology alone.

Methodology

A bit more detail on the method. I filtered for sustained altitude stays rather than all trips. A trip must have same-day high-elevation GPS activity (≥1,000 m) on day 0 and high activity on 1 in 3 days through day 5. That hopefully filters out users who drove up the mountain to exercise and then returned to lower elevations to sleep.

I was concerned about whether I should use day 0 or day 1 as the first night at altitude. Many people may travel to a location, sleep and exercise for the first time the next day. As a sensitivity check, I re-indexed the analysis so that day 1 = the first high-elevation activity (with the prior night presumed already at altitude), and it yielded similar SpO₂ and autonomic curves. Many of the recorded activities in these datasets are short work that devices automatically pick up and record. So an activity doesn’t have to be a “proper” exercise session. I decided to go with requiring high activity on day 0, which has the benfit that we can be fairly confident users are sleeping at high locations.

On the first altitude night, the average person in this cohort differs from their home baseline in several ways. Overnight SpO₂ and autonomic markers move in the expected hypoxic direction (HR up, HRV down), while sleep duration slightly increases. That small increase in sleep drive could be partly physiological (adaptation under hypoxia), but it could also be environmental: places at altitude are often quieter, closer to nature, and darker at night, all of which help sleep.

Recovery by days since arrival

While travelers remained at elevation, several markers moved partially back towards their low-elevation baselines. Mean SpO₂ improved from 2.67 percentage points below baseline on night 0 to 1.78 points below baseline on day 5.

Sleep minimum heart rate and HRV also partially rebounded, although device-derived resting heart rate did not show the same pattern. I think the interesting point here is not that oxygen saturation failed to “normalize”; a lower value is expected while the hypoxic exposure continues, but that mean SpO₂ remained clearly below the low-elevation baseline after five nights.

Table 1. Arrival-night change vs home baseline (sustained 5-day cohort)

Dose response by elevation band

The changes are not one-size-fits-all. When I stratified trips by elevation, a clear dose response appeared: higher bands show larger SpO₂ drops and stronger autonomic impacts. That gradient is hard to explain by other factors, such as jet lag or hotel beds alone. Travel disruption should not present such a clear relationship with meters gained. Hypoxia should. Guidance tells us that ascent rate and absolute elevation as core risk modifiers for acute altitude illness; these wearable curves show the same dose logic in nightly SpO₂ and autonomic markers.

Table 2. Day-0 Δ by trip elevation band (sustained cohort)

The increase in resting HR across days 0-5 is an interesting and unexpected finding. The simplest explanation is that even while SpO₂ and HRV are starting to partially recover, people in this cohort are still under a higher sympathetic “recovery demand” because the inclusion rule keeps them doing high-elevation activity through days 3–5 (so they are exercising a lot!), and altitude itself can keep sleep physiology disrupted in a way that shifts the wearable’s resting HR estimate upward.

The pattern is consistent with incomplete acclimatization plus residual load from ongoing activity at elevation. It may also be a quirk in how resting HR is calculated by different providers, and how it is impacted by (relatively) high values. Disrupted sleep would drive more and higher HR episodes during the night, pulling resting HR depending on how it’s derived.

5. SpO₂ Δ distributions by day (≥2,000 m)

The distribution view reinforces the mean curves. From day 1 onward among trips ≥2,000 m, the full SpO₂ distribution sits below home baseline, and the mean is not only driven by a few extreme responders.

Means drift only slightly across days 1–5 while travelers remain at altitude, suggesting sustained rather than rapidly resolving desaturation. Richalet’s work on physiological risk factors for severe high-altitude illness also underscores that individual ventilatory and oxygen responses vary widely; these ridge-lines make that heterogeneity visible.