Analysis of sleep records from 100,000 people
The Coronavirus Pandemic has Changed Our Sleep
By Jeff Huang, Liyaan Maskati, Nediyana Daskalova on 2020-06-16
The coronavirus pandemic has significantly affected both our work and our leisure in unprecedented ways. But a third pillar of our everyday lives has been less studied: how has the pandemic affected our sleep?
This analysis will present aggregated sleep records from over 100,000 anonymous people from all over the world, who have shared their data for analysis with a cloud upload add-on in a sleep tracking app Sleep as Android. The tracking is toggled by each person as they go to bed and wake up, and many people have years of sleep records, up to the last date of data on May 20, 2020. Technical notes are in italics.
Comparing Two Tuesdays in April 2019 and April 2020
Let's start with a simple comparison. On the first two Tuesdays in April 2019 (uneventful days a year ago), about half of the people woke up before 7:00am to start their day. It was most common to wake up between 6:30am and 7:30am. But one year later, on the first two Tuesdays in April 2020, many countries were in the initial stages of lockdown with most people waking up after 7:00am local time. There's a shift to a later wake up time. The data here controls for day of week, part of month, tries to remove naps, and only includes people who tracked on all four days.
People Go To Bed Late on Weekends, Wake Up Even Later
If we want to dig further, we have to account for the day of week. Because sleep is shifted during normal times too, notably on weekends. Here's several months of sleep with bars marking the time in bed, using the median time across everyone (about 100,000 people). Weekends are colored orange while weekdays are colored blue, so you can see that on average people go to bed half an hour later, but wake up nearly 1.5 hours later on weekends.
Let's try to quantify what this means for different people using histograms of people, a different way to look at the data. We can compare two situations based on each person's sleep records: weekdays pre-pandemic versus weekends pre-pandemic. We'll get to looking at sleep during the pandemic in the next section.
For the following set of histograms, we consider data that is tracked continuously up to March 15, 2020 (with no tracking gap of over a week) and only include people who have at least 30 days in each of the two situations. The median bed time or wake time is calculated for each person, and then each person is bucketed based on how much their schedule has changed on average. This makes the data about changes robust to various forms of noise in each person's data.
We see that on weekends, bed time is shifted later, up to an hour for most people. What's interesting is the wake up time is spread out much more. Waking up several hours later is not uncommon. The y-axis in these charts are the number of people represented in each bar.
When we tally up the time slept, including naps during the day, most people get more sleep on the weekends. This could be a hint that many people accrue a sleep deficit during the weekday.
The Pandemic has Affected Peoples' Sleep Differently
Now instead of comparing weekends versus weekdays, let's compare two different situations: weekdays before and after the pandemic. The pandemic has had an impact on sleep across the world, but sometimes differently depending on the country. Note that the following chart shows the effect from Italy encountering the pandemic earlier, and Sweden not imposing a lockdown during the pandemic.
Based on when people started staying home, we can use March 16, 2020 as a rough marker for larger sleep changes in most of the world. Other location-based analyses are available. For the remaining analyses, we're ignoring weekends during the pandemic because we already know that weekends are different and there's only a small sample of pandemic weekends for each person.
So by comparing the weekdays before March 16, 2020 with the weekdays after that date, we can see how people have slept before and during the pandemic. Again, we can use histograms to summarize each person's change into one number and group everyone by the amount of change. During the pandemic, the total hours slept remains relatively unchanged for 51% of people (within 20 minutes compared to pre-pandemic). That is, comparing the dates before and during the pandemic for each person, the median hours slept in both cases are relatively the same for about half of people.
For remaining half, there is a different story. In fact, 32% of people overall sleep longer than they did before COVID, while 17% of people sleep less on average. Nearly twice as many people sleep more hours than the number of people sleeping fewer hours.
Overall, for about half the people who we analyzed, the pandemic has resulted in different sleep behavior, which does not mimic pre-pandemic weekend sleep behavior either. There's more to investigate, and we plan to continue this analysis but wanted to share some data while the pandemic is ongoing.
This analysis is not peer reviewed or published in an academic venue, and may contain errors despite our best efforts. While the data reports time in bed, it does not reflect the quality of sleep or awakenings. I'd discourage generalizing from this data because while the sample size is large, it is a self-selecting group of people who chose to use a sleep tracking app. If you'd like to find out what affects your own sleep, try our research group's SleepCoacher App.