Insights about travel-related sleep disruption from 1.5 million nights of data

Abstract

Study Objectives

The growing adoption of reliable wearable sleep trackers has made it possible to analyze multi-night, objective sleep data from travelers on a scale hitherto not possible. This study was designed to complement previous studies of jet lag and circadian re-alignment conducted under highly controlled laboratory conditions and small-scale field studies typically involving specialist samples such as professional athletes.

Methods

De-identified sleep data from 15 days before and after 64 847 trips from 57 240 Oura Ring users were analyzed. Trips were at least 1000 km and originated from North America and Europe. We characterized the evolution of sleep timing, duration, and macro-architecture, considering the influence of number of time zones crossed, direction of travel, and habitual sleep patterns.

Results

Sleep disruption began with curtailed sleep on the night before travel because of early awakening. Sleep duration was shorter during and immediately following travel but returned to within ~12 minutes of baseline after ~2 days. In contrast, changes in sleep timing and sleep architecture were considerably slower to recover, with sleep timing not returning to baseline after 15 days. Sleep disruption was more severe with eastward travel and across more time zones. Interindividual differences in both sleep duration and timing equilibrated with travel.

Conclusions

Both structural and intrinsic circadian factors influence sleep during travel. Sleep homeostatic mechanisms drive recovery of sleep duration quickly although architecture is still compromised. Re-alignment of sleep timing to the new time zone takes significantly longer.

Graphical Abstract

Graphical Abstract.

Statement of Significance.

In today’s globalized world, travel-related sleep disruption affects more people than ever before, with consequent human and economic costs. The increasing use of reliable wearable sleep trackers now enables us to characterize how travel affects sleep on a much greater scale than previously possible. By analyzing sleep data before and after thousands of flights crossing different numbers of time zones, we found that non-circadian factors (e.g. flight departure times, and pre-travel sleep patterns) influence how we sleep during travel in addition to jet lag and provide a comprehensive characterization of sleep of everyday travelers. These data provide a new insight into how sleep is affected by everyday travel under real-world conditions, which have been difficult to replicate in previous studies.

Increase in international travel [1] has resulted in millions of people experiencing travel-related sleep disruption, which has both human and economic costs [2]. This disruption arises from circadian misalignment (jet lag), the stress associated with travel and changes to normal daily routines [3, 4], and an uncomfortable travel environment [5–7].

Laboratory-based isolation studies use shifts in the light–dark cycle to simulate changing time zones. These shifts result in a reduction in circadian rhythm amplitude and misalignment in circadian phase that gradually re-entrains to the new light–dark timing [8–11]. This disruption shows directional asymmetry, with recovery following a phase advance, analogous to eastward travel, taking longer than that following a phase delay [9–12]. Results from field studies involving actual travel typically reflect this circadian misalignment with sleep onset time (SOT) and wake time (WT) occurring earlier following westward travel and later following eastward travel [13–17]. However, results from field studies are not always in accordance with in-lab studies [8, 18–20], suggesting that real-world environmental exposure and social factors influence recovery from travel.

While field studies reflect the experience of everyday travelers more closely than lab-based studies, participants in such studies have often been military personnel [21, 22], professional athletes [18, 23–25], or air crew [26–28], who may face behavioral constraints that everyday travelers do not encounter. Furthermore, because most travel is for leisure and not business [29, 30], many travelers have the discretion to alter their behavior to mitigate the effects of travel-related sleep disruption in ways not available to participants in more controlled studies (e.g. changing their sleep patterns, consuming alcohol or caffeine) [31].

We investigated travel-related sleep patterns from 15 days before to 15 days after an outbound trip of over 1000 km. We analyzed objective sleep measures from 64 847 trips made by 57 240 Oura Ring users, examining the evolution of pre- and post-flight effects on sleep timing and duration as well as the influence of direction of travel, number of time zones crossed, age, sex, and pre-travel sleep. The data collection method was passive and imposed no additional burden on participants; therefore, user behavior around the travel period was likely more representative of a general population of travelers than prior studies. Together, these features enabled the characterization of sleep during the peri-travel period more comprehensively than ever before.

Methods

Data collection

Data were collected by Oura Health Oy (Oulu, Finland) in accordance with their Terms and Conditions (https://ouraring.com/terms-and-conditions) and Privacy Policy (https://ouraring.com/privacy-policy-oura-health). All user data were de-identified, and no user location data beyond their origin country was shared outside Oura. The National University of Singapore Institutional Review Board deemed this study exempt from formal review because the data contained no personally identifiable information. The Oura Ring and App [32] have been extensively validated against polysomnography (PSG) [33–38].

Sample characteristics

Data from 57 240 Oura Ring Gen3 users (32 166 female, 25 074 male) were included in the analysis. Users were 18–80 years old (mean [SD] male: 46.04 [13.03]; female: 42.88 [13.08]). A total of 50 555 users contributed a single trip during the analysis period and 6685 users contributed multiple trips (mean of 2.14 trips for those contributing multiple trips). See Section 1.1 of the Supplementary Material for additional sample details.

Trip detection

Travel detection.

Duration and distance of travel were identified using mobile phone geolocation data. City-level user location recorded by the Oura App (reported from the phone IP address) at the first login of the day was taken as the location of the previous sleep episode. Location data was recorded on every day on which there was an app login.

Trips were only included in the analysis if users logged into the app on the day following the night prior to travel (Day −1), the night of travel (Day 0), and the night following travel (Day 1). In addition, the first login on the day of travel was required to be within 6 hours of the recorded WT. This reduced the chances of travel prior to data syncing, which would make it appear as though travel had occurred one day earlier. The median first login time was 07:48 am and the median duration between reported WT and login time was 32 minutes, indicating that most users logged in shortly after they woke up in the morning. This suggests that the location reported at the first login provided a valid estimate of where users slept the previous night.

In addition to location data provided by the phone IP address, time zone data is also contained in the time data broadcast to the phone from its cellular network. This time zone is associated with Oura’s sleep data. Trips were excluded if the IP-based longitude change was more than ±30° from the expected change based on time zone information from the cellular signal. This enabled us to exclude cases where VPN use made it appear as though the user had changed location.

Trip selection and characteristics.

A total of 64 847 trips were included in the analysis (Table 1). Trips took place between June 16, 2023, and October 11, 2023. Including 15 days on either side of travel, the full data analysis period ran from June 1 to October 27, 2023. The start of the analysis period coincided with the release of Oura’s then-newest sleep staging algorithm on June 1, 2023; data were only included if the user had upgraded to that algorithm (there were no further changes to the algorithm within the analysis period). The end of the analysis period was before the time change in Europe from Daylight Savings Time to Standard Time on October 29; therefore, the time difference between all locations was consistent during the analysis period. Trips were only included if the following criteria were met: (a) they contained data for at least 20 days out of the 31-day analysis period, (b) they were at least 1000 km in length (see Section 2.1 of the Supplementary Material for travel distance details), (c) they were outbound trips from the user’s home country and (d) they did not involve a fractional time zone change. Location data indicated that none of the users had travel in the 15-day baseline period.

Table 1.

Number of Trips According to Number of Time Zones Crossed, Direction of Travel, and Region of Origin

Time zone change	Total number of trips	Total West	West from North America	West from Europe	Total East	East from North America	East from Europe
0	37 232	—	—	—	—	—	—
1	10 492	5189	4592	597	5303	4353	950
2	6261	3340	3227	113	2921	2610	311
3	7064	3661	3653	8	3403	3360	43
4	224	97	90	7	127	97	30
5	679	225	63	162	454	386	68
6	1222	288	78	210	934	875	59
7	563	136	11	125	427	366	61
8	476	189	78	111	287	280	7
9	457	127	45	82	330	320	10
10	119	58	37	21	61	61	0
11	57	51	51	0	6	6	0
12	1	0	0	0	1	1	0

Available data for 15 days before and after travel were included in the analysis (31 days in total, termed Day −15 to Day 15; the baseline period comprised Day −15 to Day −2). Day −1 was the last day the user slept at their origin location. Westward travel is typically during the day and results in an extension of the nychthemeron (going “back” in time), whereas eastward travel results in a shortening (going “forward” in time). Therefore, sleep on Day 0 for westward travel is likely to be at the destination; however, for eastward travel, particularly for longer trips, sleep on Day 0 is likely to be during travel. It should be noted, however, that we do not have the time of travel for individual trips; these inferences are made from flight departure time patterns (Figure 2B) and the effects of transmeridian travel on solar time experienced by the traveler.

Figure 2.

(A) Change from baseline in SOT, WT, and TST on the night/morning before travel. For comparison with flight departure data, only transatlantic flights were included in the Long flight groups (i.e. westward trips from Europe and eastward trips from North America). * p < .05, *** p < .001. (B) Distribution of flight departure times for Short (1–3 time zones crossed; inner ring) and Long (5–9 time zones crossed; outer ring) westward and eastward flights. Data are for departures from New York JFK and London Heathrow airports during the week of August 1, 2023–August 7, 2023. Data retrieved from https://aerodatabox.com.

Trips crossing 1–3 time zones were categorized as Short trips and those crossing 5–9 time zones as Long trips (see Section 1.3 of the Supplementary Material for the number of nights contributing to the estimate of each group’s sleep measures). This grouping reflects the distribution of flights in our dataset; most transmeridian trips were Short, intra-continental trips; the majority of others crossed 5–9 time zones and were predominantly transatlantic trips. There were few trips crossing 4, or 10 or more time zones, resulting in unreliable estimates; therefore, these were excluded from analysis. Furthermore, inspection of sleep timing and duration characteristics for ungrouped trips showed shared features for trips crossing 4 or more time zones that were not present for Short trips (see Section 2.2 in the Supplementary Material).

Approximately 72% of the initial dataset involved trips originating in North America, and 19% originated in Europe; therefore, to simplify the interpretation of the results, we included only trips originating in North America and Europe. Of the 64 847 trips included in the analysis, 58 623 (90.4%) were from North America and 6224 (9.6%) were from Europe.

Sleep measures

Sleep measures were only included for days when users were either in their origin location or their first destination; that is, data were excluded if the user traveled to another time zone (either an onward trip or returning home).

Oura provided sleep measures for the longest sleep period of over 3 hours duration each day. Sleep measures comprised: total sleep time (TST), SOT, WT, sleep onset latency, and sleep efficiency. Sleep architecture was reported as time in deep sleep, light sleep, rapid eye movement (REM) sleep, and wake after sleep onset (WASO). Baseline sleep measures were calculated as the mean over the baseline period (Days −15 to −2). Baseline chronotype was calculated as mean mid-sleep time on weekend nights (MST_WE) during the baseline.

For each sleep period, outlier and physiologically implausible data were excluded from the analysis. For both the sleep and physiological measures reported by Oura, values outside the range indicated here were excluded: TIB and TST: 3–14 hours; sleep efficiency: 30%–98%; sleep onset latency: 0–120 minutes; WASO: 0–300 minutes; average heart rate: 30–110 bpm; lowest heart rate: 20–100 bpm; heart rate variability: 0–250; breathing rate: 10–22 breaths per minute; temperature: 34–38°C. If any single measure fell outside the acceptable range, this was an indication that data was not being collected accurately and all measures for that sleep period were removed. Baseline nights were subject to additional timing criteria, where sleep periods with SOT outside 08:00 pm–04:00 am or WT outside 04:00 am–12:00 pm were excluded from the analysis to remove periods of shift work or other non-typical sleep episodes.

Following the exclusion of outlier data (56 627 sleep periods) and sleep episodes where users were no longer in their original destination (288 119 sleep periods), a total of 1 469 123 sleep periods were included in the analysis (an average of 22.66 per trip).

Statistical analysis

Data were analyzed using multi-level linear models. Primary models predicted change from baseline with age (mean-centered), sex, trip day, travel direction, and number of time zones crossed as fixed effects and trip ID as a random effect. Up to third-order interactions were included in the model. Because of the limited number of repeated measures, adding user ID to the model explained no, or very little, additional variance and did not appreciably improve model fit; therefore, it was not included in the model (see Section 1.4 in the Supplementary Material for full model specification). While the baseline was calculated using Days −15 to −2, only Day −5 onwards were included in the model. For analysis of effects on the day before and the day of travel, linear models were fit to Day −1 and Day 0, respectively, without the trip ID random effect. For tests of effects after travel, models were fit using data from Days 2–4, Days 6–8, and Days 10–12 (Days 0 and 1 were excluded to minimize the short-term sleep duration effects; see Section 2.2 in the Supplementary Material for illustration of day groupings). ANOVAs (with Satterthwaite’s degrees of freedom) were conducted, and post hoc pairwise comparisons used either the multivariate t distribution or the Tukey correction for multiple comparisons. Effect size estimates (partial eta-squared, η_p²) are provided in addition to p-values because they are not influenced by sample size. Data processing was conducted in Matlab and statistical analysis in R using the lmertest [39] and emmeans [40] packages.

Results

Pre-travel sleep disruption

Sleep duration on the night before transmeridian travel was 30–50 minutes shorter than baseline (Figure 1C) because of earlier WT on the day of travel (Figure 1B; see Section 1.2 of the Supplementary Material for baseline sleep measures and Section 3 for example actograms of individual trips). Shortening of sleep was likely influenced by systematic biases in the distribution of flight departure times according to both the length and direction of travel (Figure 2). Longer eastward flights tend to be overnight, departing in the late afternoon or evening, while Short (crossing 1–3 time zones) and Long (crossing 5–9 time zones) westward flights depart throughout the day (Figure 2B). This allows for a later WT on days with a Long eastward flight. Consequently, the shortening of TST associated with earlier WT was less for Long eastward trips than for Short trips or Long westward trips (TST: ps < .001, WT: ps < .001; Figure 2A). Therefore, Long eastward travel was associated with less pre-travel sleep disruption than other trip types.

Figure 1.

Estimated marginal means and 95% confidence intervals for (A) sleep onset time (SOT), (B) wake time (WT) and (C) total sleep time (TST) for Long (dark) and Short (light) trips going East (red) and West (blue) and travel within the same time zone (gray). Change in SOT and WT is relative to local time; therefore, negative values represent timing earlier than baseline, and positive values represent timing later than baseline. Also, see Figure S6 for sleep onset latency (SOL) and sleep efficiency (SE).

Sleep during and after travel

As expected, eastward and westward travel had opposite effects on sleep timing. Traveling east was associated with later SOT and WT at the destination than at the origin (relative to local time). Conversely, traveling west was associated with earlier SOT and WT (Figure 1, A and B). Furthermore, sleep timing did not fully return to baseline within the 15-day post-travel period.

Travel direction modulated the effect that the number of time zones crossed had on sleep timing. After the first 2 days following eastward travel, the difference in sleep timing between Long and Short trips was small (SOT: 12.28 minutes; WT: 8.85 minutes). However, following westward travel, the difference was larger (SOT: 47.44 minutes; WT: 32.88 minutes). This was reflected in a significant interaction between direction and trip type for both SOT (F(1, 20729) = 402.42, p < .001, η_p² = 0.02) and WT (F(1, 20787) = 195.20, p < .001, η_p² = 0.009).

On the day of travel, TST was shortest for Long eastward trips (Figure 1C; mean TST reduction [95% CI] Long eastwards: 61.87 minutes [56.65, 67.08], Long westwards: 20.22 minutes [14.015, 26.44], Short eastwards: 17.88 minutes [16.13, 19.63], Short westwards: 13.57 minutes [11.88, 15.26], p < .001 for pairwise comparisons between Long eastward and other trip types). On the following night, there was a rebound in TST, resulting from an earlier SOT, likely as a result of increased sleep pressure arising from poor sleep on the preceding night. TST recovered to within ~12 minutes of baseline within ~2 days following travel, although following Long eastward travel it was significantly shorter than baseline (p < .01 for Days 2–9). Following westward travel, TST did not increase above baseline until Day 4 where it remained until Day 13. The increase above baseline was not consistently significantly above baseline (p < .05 on Days 5, 9, 10, and 12); however, this might reflect the smaller sample size for Long West trips (965 trips) compared with Long East ones (2432 trips).

Changes in sleep architecture

On the night prior to travel, particularly for westward travel, there was an increase in the proportion of deep sleep and a decrease in the proportion of REM sleep (Figure 3, B and C). On the day of travel, Long trips, particularly eastward, were associated with an increase in WASO and a decrease in REM sleep (Figure 3, C and D). Following travel, Long eastward travel was associated with a sustained decrease in the proportion of deep sleep and REM, and a concomitant increase in light sleep and WASO. Westward travel, however, showed a more limited increase in the proportion of deep sleep and a reduction in light sleep. Following any transmeridian travel, there was an increase in the proportion of WASO and a decrease in the proportion of REM which could last over a week (also see Figure S7 for time spent in each sleep stage).

Figure 3.

Changes in sleep macro-architecture for Long (dark) and Short (light) trips going East (red) and West (blue) and travel within the same time zone (gray). Note different y-axis (ordinate) scales; dotted lines are plotted at ±2% for comparison. REM, rapid eye movement sleep; WASO, wake after sleep onset. Bands represent 95% confidence intervals. Dots indicate days on which percent change for transmeridian trips is significantly different from 0 at p < .001.

Effects of number of time zone crossings, age, and sex

To characterize how sleep timing was influenced by the number of time zones crossed, we analyzed the second, third, and fourth days after travel (Days 2–4; Figure 4, solid lines). These days were chosen to minimize the impact of sleep deprivation on Day 0 and the subsequent increase in sleep pressure on Day 1. Additionally, we included Days 6–8 (Figure 4, dashed lines) and Days 10–12 (Figure 4, dotted lines) in the analysis to illustrate the temporal dynamics of recovery (also see Section 2.2 of the Supplementary Material). Trips crossing five and six time zones were combined into one group (plotted at 5.5 hours in Figure 4), as were trips crossing seven, eight, and nine time zones (plotted at 8 hours). This was done to increase the number of observations in each group while maintaining meaningful separation between the groups.

Figure 4.

The effect of crossing different numbers of time zones on (A) SOT, (B) WT, (C) TST. Note different y-axis (ordinate) scales for TST (Panel C). Westward trips are plotted in blue and eastward trips are plotted in red. Estimated marginal means averaged over Days 2–4 (solid lines), Days 6–8 (dashed lines) and Days 10–12 (dotted lines) are plotted. Bands represent 95% confidence intervals. See Figure S8 for effects on SOL and SE.

On Days 2–4, earlier SOT and WT of up to about an hour were observed for westward travel, increasing approximately linearly with more time zones crossed (Figure 4, A and B). For eastward travel, trips crossing 1–3 time zones showed increasingly later SOT and WT, of a larger magnitude than for westward travel, up to a maximum of about 1 hour (SOT: (F(1, 19341) = 797.45, p < .001, η_p² = 0.04); WT: (F(1, 19586) = 314.51, p < .001, η_p² = 0.02). Sleep timing returned toward baseline over the subsequent days. For westward travel, most of the recovery had taken place by Days 6–8, with little, if any, further change by Days 10–12. For eastward travel, sleep timing still showed recovery between Days 6–8 and Days 10–12 for longer trips.

TST for shorter westward trips was reduced by up to ~12 minutes (on Days 2–4); for longer trips, TST became slightly longer with increasing time after the start of travel. For eastward trips on Days 2–4, TST reduced in an approximately linear fashion with the number of time zones crossed reaching a maximum of less than 15 minutes (Figure 4C; also see Figure S8). As a result of the shifts in SOT and WT, the effect of travel on sleep duration was less than the effect on sleep timing.

To investigate how changes in sleep timing and duration with travel were influenced by age and sex, we tested their effects on Days 2–4 (when sleep disruption was maximally influenced by circadian factors, rather than short-term homeostatic pressures). There were main effects of both age (F(1,18847) = 48.81, p < .001, η_p² = 0.0026) and sex (F(1, 19149) = 4.74, p = .03, η_p² = 0.0002) on TST. The effect of age reflects the fact that, across both direction of travel and number of time zones crossed, older people showed a smaller decrease in TST following travel than younger people. With each year of age advancement, there was a 0.37-minute smaller decrease in TST; therefore, a 20-year-old would be expected to show about a 15-minute larger reduction in TST than a 60-year-old. The estimated marginal means for sex were −4.47 (95% CI: −6.03, −2.90) minutes for women and −7.15 (95% CI: −8.83, −5.48) minutes for men, indicating that the reduction in TST was only slightly greater for men than for women.

There was a small change in SOT with increasing age (F(1, 20488) = 22.84, p < .001, η_p² = 0.0011), but no effect of sex (F(1, 20720) = 1.79, p = .18, η_p² = 0.0001). There was also a significant interaction between age and direction of travel (F(1, 20486) = 5.56, p = .02, η_p² = 0.0003). These results indicate that the delay in SOT following eastward travel was reduced with increasing age. However, the advance in SOT was unaffected by age for westward travel.

Moderation of interindividual sleep differences following travel

Prior research has found that differences in TST across individuals equilibrated with travel [41]. People with less than 7.5 hours of habitual weekday sleep slept more during periods of travel; conversely, those with more than 7.5 hours slept less.

Our results are consistent with these findings. Users in the quartile with the shortest baseline TST showed the largest increase in TST following travel, while those with the longest baseline TST showed the largest decrease (Figure 5A, see Figure S9 for additional trip types).

Figure 5.

Moderation of (A) TST with baseline sleep duration, and WT with baseline chronotype following (B) no time zone crossings, (C) westward, and (D) eastward travel. Note the difference in y-axis (ordinate) scales. Users were grouped into quartiles based on habitual sleep duration (mean baseline TST; panel A) and chronotype (mean baseline weekend mid-sleep time, MST_WE; panels B, C, and D). Bands represent 95% confidence intervals. Also see Figures S6–S8.

Grouping users into chronotype quartiles based on pre-travel MST_WE (weekend mid-sleep time), we also found that for later chronotypes, WT following westward travel remained earlier than baseline, while for earlier chronotypes it returned to baseline (Figure 5C, see Figure S10 for additional trip types). There was a parallel effect for eastward travel; WT remained significantly later than the baseline for earlier chronotypes (Figure 5D, see Figure S11 for additional trip types). These effects were similar, but smaller, for travel within the same time zone (Figure 5B).

Discussion

Travel direction, number of time zone crossings, and pre-travel preparation all disrupt sleep. Their effects are influenced by individual differences in chronotype, habitual sleep duration, and, to a much lesser extent, age and sex.

Pre-travel sleep disruption, well-known to travelers who live a long way from the airport or who book early flights, is modulated by flight timing [42]—with later flights associated with less pre-travel sleep disruption. Going to sleep earlier on the night before travel could mitigate this sleep loss [43], but there is no evidence in our data that, on average, travelers adopt this strategy. Taking a later flight may also be beneficial, but more desirable flight times tend to be more expensive.

Unsurprisingly, sleep during a flight (likely to occur on Long eastward trips) was significantly shorter and of lower quality than habitual sleep, because of factors such as hypobaric hypoxia [5], an uncomfortable sleeping environment [6, 7], and a shortened night. While we do not have information about the cabin class distribution of our sample, future research could investigate whether being in a more comfortable class results in better in-flight sleep and how that affects subsequent sleep disruption. The following night, however, was characterized by earlier sleep onset and increased sleep duration [17, 25], presumably driven by increased sleep pressure resulting from the previous night’s poor sleep. This may result in travelers feeling as though they are not suffering from jet lag immediately following travel, only for the circadian misalignment to become evident as sleep pressure is normalized.

Sleep duration from Day 2 onwards was generally within 12 minutes of baseline, even for Long trips, indicating only a modest sustained impact. This contrasts with sleep timing which took much longer to recover and often did not return to baseline. The rate of re-alignment of the sleep–wake cycle is slower than the rate of circadian re-entrainment reported in previous research of 1–1.5 hours/day [8, 15]; however, those estimates were derived under significantly different experimental conditions, and from physiological (e.g. core body temperature, urine excretion) and performance measures, rather than sleep timing. This pattern of results might be explained by the characteristics of our sample. Because we analyzed only outbound flights, and most flights are for leisure, we presume most of our sample were vacationers. Therefore, they probably had flexibility in their sleep timing, allowing them to maintain their sleep duration with no pressure to realign quickly to local time.

Consistent with previous research, our results demonstrate a directional asymmetry in the change in sleep timing. For trips up to three time zone crossings, the magnitude of shift in sleep timing was greater for eastward trips than for westward ones. Thereafter, the change in timing for eastward trips remained constant, while for westward trips it continued to shift even earlier. This may reflect a limit to the amount eastward travelers can delay waking because they need to get up for the day’s activities. For westward travel, however, there may be fewer constraints on going to bed and getting up earlier than normal.

Directional asymmetry in sleep disruption was also evident in sleep architecture. Long eastward trips were characterized by a sustained increase in the proportion of WASO and light sleep with an accompanying reduction in the proportion of deep and REM sleep lasting for 5–6 days. The time change with eastward travel makes it challenging for travelers to fall asleep at their habitual bedtime; they also wake up at an earlier point in their circadian phase, resulting in a curtailment of REM sleep [44]. Westward trips showed a decrease in the proportion of REM and light sleep and an increase in the proportion of deep sleep and WASO, although these changes were more limited than those associated with eastward travel. The increase in day length caused by the time change and subsequent early awakening may result in greater accrual of sleep pressure, the dissipation of which is reflected by a greater proportion of time spent in deep sleep. Therefore, while sleep duration recovers substantially within a few days of travel, sleep macro-architecture may remain compromised for up to a week after Long eastwards travel.

Our results support previous research [41] showing that sleep duration normalized with travel. In the present work, we show this is also the case for sleep timing. In addition to vacation, other periods characterized by increased flexibility, such as the weekend and during the COVID-19 pandemic, typically show an increase in sleep duration and shift to later sleep times [45–47]. However, this was not the case here; rather, it appears that once the habits and pressures on sleep associated with the home environment and everyday life were released, disparities in sleep patterns were reduced. These results are consistent with those showing that the shift to an earlier circadian phase during a camping trip was greater in later chronotypes [48].

Interindividual differences in circadian and sleep systems [49] mean that there can be substantial variability in how individuals respond to transmeridian travel [9]. Lab-based studies have demonstrated that circadian amplitude reduces with age, although how this affects sleep following travel in real-world conditions is not clear [50]. Our results indicate that older people show a smaller reduction in sleep duration than younger people; however, this may simply result from older people having shorter habitual sleep. There were minimal effects of age on changes in sleep timing, consistent with findings that older people do not necessarily have impairment in circadian phase shifting compared with younger people [51].

A common strategy vacationers use to combat the acute sleepiness and fatigue experienced during travel is to take naps [31]. In this analysis, we included only the longest daily sleep episode over 3 hours because naps are difficult to detect unambiguously with current non-EEG-based wearables. Future research with reliable nap detection would help characterize their impact on post-travel sleep recovery. We included only outbound trips because they have the advantage of an extended baseline period free from travel, which is unlikely to be the case for return flights. Nevertheless, pressures on sleep around outbound and inbound trips will likely differ. Similarly, vacation and business travelers have different obligations and behavioral constraints during travel; therefore, additional research is needed to characterize how these affect sleep. Finally, most Oura Ring users live in North America and Europe, but with increasing adoption of wearable devices, wider geographical coverage will be available in the future.

This research could be profitably extended by collecting a wider variety of data. Information about the travel (e.g. flight times and locations), the traveler (e.g. demographic and phenotypic information), measures taken to counter sleep disruption (e.g. hypnotics, caffeine/alcohol intake), and peri-travel behavior (e.g. work or social obligations) would enable more precise characterization of travel-related sleep disruption and potential customization of advice to mitigate its effects. A further extension would be to measure the timing, duration, and intensity of travelers’ light exposure [52]. Light is the dominant zeitgeber and is critically important in synchronizing our circadian system to local solar time [52]. Therefore, collecting continuous light exposure data in parallel with sleep data would enable analysis of how different types of exposure influence re-alignment in unconstrained, real-world conditions. Meal [53] and exercise [54] timing also affect circadian synchronization, albeit to a lesser extent than light; therefore, this data would also be useful in determining their contributions to re-entrainment.

In sum, while waiting for the enhancements in the next generation of studies, we provide the most comprehensive characterization of travel-related sleep disruption to date, both in terms of the number of trips analyzed and the duration of the period surrounding travel. This should provide clear empirical evidence on which to base advice about the travel-related sleep disruption an average long-distance traveler can expect on their next flight.

Supplementary material

Supplementary material is available at SLEEP online.

Funding

This work was supported by funds from the Yong Loo Lin School of Medicine and the Lee Foundation awarded to Dr Michael W.L. Chee.

Disclosure Statements

Financial disclosure: MWLC is a member of the Oura Health Oy Medical Advisory Board. RV is a full-time employee of Oura Health Oy. ARW and JLO declare no competing interests. Nonfinancial disclosure: None.

Author Contribution

A.R.W.: Conceptualization, Methodology, Formal analysis, Writing – original draft, Writing – review and editing, Visualization. R.V.: Conceptualization, Data Curation, Writing – review and editing. J.L.O.: Conceptualization, Methodology, Writing – review and editing. M.W.L.C.: Conceptualization, Writing – review and editing, Supervision, Funding acquisition.

Data Availability

The data used in this study are the property of Oura Health Oy and are not shared publicly. Access to anonymized and privacy-protected data may be granted to a qualified academic investigator with agreement from Oura Health Oy and M.W.L.C.