Result
DST did not affect sleep metrics at the population level
Initially, we analysed how the median values of sleep metrics changed across the population during the three periods. Our analysis revealed minimal changes in these measures before, during, and after the DST period (see Figure 1 and Table 1). Specifically, we found no significant difference in sleep efficiency (as defined as the percentage of time awake within a designated sleep session) across the three periods. The sleep scores exhibited only a marginal difference of 0.54 units between the pre-and post-DST phases. Notably, there was a slight increase in total sleep duration during the DST transition days, returning to pre-switch levels thereafter. While there were no discernible differences in Sleep HR, we observed a slight rise in heart rate variability (HRV) during sleep in the week following DST. Lastly, we found no discernible variation in sleep onset time across the three periods.
![](https://cdn.prod.website-files.com/622a0bff6151d22d7cab707a/6680197e746a0e04c5ba8d4f_fig1.png)
Figure 1: Violin plot displaying the median distribution of sleep measures before, during, and after DST in the combined user group (N=5145).
![](https://cdn.prod.website-files.com/622a0bff6151d22d7cab707a/667fd97969cd32468c3faf92_tab1.png)
Table 1: Median and median absolute deviation of the sleep metrics across the three periods (N=5145).
Since we measured the metrics at the participant level across the three periods and our data had a non-parametric distribution, we employed mixed-effects models to assess any statistical dependencies. In these models, the three time periods were used as fixed effects, and each user was treated as a random effect to account for repeated measures. Although we found statistically significant differences in various sleep metrics across the different periods, the magnitude of these differences, as indicated by the coefficient values, was very small (Table 2). Additionally, we calculated the intraclass correlation coefficient (ICC) to determine the proportion of total variability in the sleep metrics contributed by differences between users (individual variability) rather than changes due to DST. We found that for nearly all sleep metrics, the ICC was close to 0.60, indicating that a significant portion of the variability is due to individual differences. This suggests that individual-specific factors have a greater impact on sleep than the time shift caused by DST.
![](https://cdn.prod.website-files.com/622a0bff6151d22d7cab707a/667fd979432df1a9f09e5a76_tab2.png)
Table 2: Comparison of coefficients, p-values, and intraclass correlation coefficients (ICC) obtained from mixed-effects model for sleep metrics across pre-switch, transition, and post-switch periods.
Subgroup Analysis Reveals Differential Effects of DST on Sleep
Based on the results from the mixed-effects models, we investigated whether the sub-groups within the user population might be differently affected by the DST switch. We stratified users into three groups based on their pre-switch sleep scores: below 60, between 60 and 85, and above 85 (Figure 2; Table 3). Interestingly, users with pre-switch sleep scores above 85 experienced a decrease of 4.75 points in their sleep scores after the DST switch, with most of this effect attributed to DST (indicated by a low ICC value of 0.18; see Figure 2a). Conversely, users with pre-switch sleep scores below 60 showed an increase of 6.87 points in their sleep scores, also reflected by a low ICC of 0.16, and an additional 30 minutes in sleep duration (Figure 2b). For users with pre-switch sleep scores between 60 and 85, we observed only negligible differences, with very high inter-individual variance (Figure 2c), similar to the overall population.
![](https://cdn.prod.website-files.com/622a0bff6151d22d7cab707a/6680197f2f09f3e46b73734f_fig2.png)
Figure 2: Violin plot illustrating the median distribution of sleep measures before, during, and after DST, categorised into three groups based on pre-switch sleep scores: a) <60 (N=392), b) 60-85 (N=3803), and c) >85 (899).
![](https://cdn.prod.website-files.com/622a0bff6151d22d7cab707a/667fd97900741aee88ad5194_tab3.png)
Table 3: Comparison of coefficients, p-values, and intraclass correlation coefficients (ICC) obtained from a mixed-effects model for sleep metrics across pre-switch, transition, and post-switch periods, stratified by pre-switch sleep score groups: <60 (N=392), 60-85 (N=3803), and >85 (N=899)d on pre-switch sleep scores: a) <60 (N=392), b) 60-85 (N=3803), and c) >85 (899).
Conclusions, Limitations and Future Directions
In our analysis, we aimed to investigate the impact of DST on sleep in Ultrahuman Ring AIR users. Although we found significant differences in sleep metrics before and after the onset of DST at the population level, but given the small effect size we conclude that DST had minimal impact on the user's sleep. Despite this, it is important to note that when comparing pre-and post-switch periods, we observed a longer tail for sleep efficiency and a shorter tail for total sleep duration in the post-switch period. This indicates that while most users had minor changes, a subset of users were influenced by the DST switch. Studies using bedtime data indicate that adjustments to DST on weeknights are nearly immediate, while adjustments towards weekends typically take about two weeks 2. Since we averaged sleep onset times over the week, our results align with this observation as our data predominantly reflects weekday sleep patterns.
When we stratified the data based on pre-switch sleep scores, we observed that users with a pre-switch sleep score below 60 (which would be considered less optimal) experienced an increase in their sleep scores after the DST switch. This result was unexpected. However, it is possible that these users had irregular sleep patterns during the pre-switch period, which normalised in the post-switch period. Therefore, the observed improvement in sleep scores may not be due to the DST switch itself, but rather a return to a more typical sleep pattern. Furthermore, since these measures are the average values of the week before the DST shift, they may represent either chronically non-ideal sleepers or those with large night-to-night sleep score variations. Thereby, the observed changes might be influenced by fluctuations in sleep patterns unrelated to the DST transition, underscoring the need for a more robust baseline measurement to draw definitive conclusions.
While numerous studies indicate that DST negatively impacts sleep, many of these studies do not account for individual factors that could influence sleep patterns 2,5. A smaller body of research suggests that the effects of DST may be more pronounced in individuals with pre-existing sleep debt 2,6. It is important to note that most users of Ultrahuman Ring AIR are proactive about their health. Therefore, it is plausible that our study examines a subpopulation characterised by relatively good sleep and other health profiles and our findings may not apply to the broader population.
In summary, our observations reveal significant individual variability in the responses to DST on sleep. Studies show that morning-type people experience less sleep disruption from DST compared to evening-types, emphasising the role of individual differences 7. The effects of DST on sleep are complex and nuanced, potentially influenced by factors such as sleep hygiene 8, lifestyle choices, and individual stress levels 9. Our findings suggest that while some users experience notable changes in sleep patterns, others are minimally affected. To better understand these variations and identify the various sub-groups, further research is required to stratify the user base with greater granularity and account for the aforementioned factors.
Reach out to support@ultrahuman.com for commercial queries and science@ultrahuman.com for scientific queries.