Insights into maternal sleep: a large-scale longitudinal analysis of real-world wearable device data before, during, and after pregnancy

Abstract

Background: Current understanding of pregnancy and postpartum sleep is driven by limited lab or self-reported data. Our goal is to use consumer wearable devices through an observational study to reveal longitudinal, real-world sleep patterns in this population.

Methods: We analysed retrospective, de-identified Fitbit device data from 2540 users in the United States and Canada who met strict wear-time requirements (≥80% daily usage for ≥80% of the time periods of interest [12 weeks prepregnancy, throughout pregnancy, and 20 weeks immediately postpartum]). We tracked sleep time and stages using Fitbit devices.

Findings: Pregnant participants experienced a peak in total sleep time (TST) at 10 weeks (447.6 ± 47.6 min), exceeding their prepregnancy average (425.3 ± 43.5 min) before declining throughout pregnancy. This initial TST increase, mirrored by time in bed (TIB), was driven by more light sleep. Deep and rapid-eye movement sleep decreased significantly throughout pregnancy, with maximum reductions of 19.2 ± 13.8 min and 9.0 ± 19.2 min respectively by pregnancy end (two-sided t-test, p < 0.001 for both). Sleep efficiency also slightly declined during pregnancy (median drop: 88.3%-86.8%). Postpartum, TIB remained below prepregnancy levels by 14.7 ± 45.7 min one year after birth and 15.2 ± 47.7 min at 1.5 years after birth.

Interpretation: This study revealed a previously unquantified initial increase in sleep followed by decreases in both quantity and quality as pregnancy progresses. Sleep deficits persist for at least 1.5 years postpartum. These quantified trends can assist clinicians and patients in understanding what to expect through their pregnancy and postpartum journey.

Funding: Google, LLC.

Keywords: Postpartum-related sleep; Pregnancy-related sleep; Sleep architecture; Wearables.

Fig. 1

Study Flow. From the participants that met the inclusion criteria of being the primary caregiver to a child under 1 year old. Those who gestated the child and had pregnancy lengths between 23 and 42 weeks were included in the analysis. Participants who wore their device for at least 80% of the day (more than 19 out of 24 h) were divided into two cohorts. The pregnancy cohort consisted of 2540 participants who wore their wearable device for at least 80% of the baseline period (12 weeks immediately prior to pregnancy start) as well as throughout pregnancy. The postpartum cohort included 2314 participants who wore their device for at least 80% of the same baseline period (12 weeks before pregnancy) as well as the first 20 weeks after giving birth. 1876 participants were in both cohorts.

Fig. 2

Number of participants analysis. a. Density distribution of pregnancy length for participants in the pregnancy cohort. The dotted line marks 38 weeks of pregnancy. Over 70% of participants had pregnancies exceeding this length. b. Mean percentage of participants in the pregnancy cohort included in each week of analysis. The number of participants per day was averaged across each week. The percentage of participants is relatively stable until it starts decaying in week 33, reaching a minimum of 72.5% participants in week 38. c. Percentage of participants in the postpartum cohort included in each week of analysis. The negative time (grey dots) represents the baseline at 12 weeks prepregnancy. The break in the x-axis (//) indicates that the time is not continuous. The dashed vertical line represents time of delivery and positive time values refer to the number of weeks postpartum. Week 1 is the first week after giving birth. The percentage of participants with sleep data is above 70% the second week after giving birth.

Fig. 3

Resting Heart Rate (RHR) changes before pregnancy. Analysis done on the pregnancy cohort. RHR was normalized by subtracting the average during prepregnancy for each participant. Pregnancy start was determined using the participant’s reported gestational length and delivery date. RHR was aligned to the pregnancy start and averaged per day for all participants (N = 2540 participants). Mean RHR is depicted in black and standard deviation in grey areas. The dashed line indicates the estimated first week of pregnancy which coincides with a decrease in RHR, expected during menstruation. The cyclical changes before pregnancy start occurs at a frequency characteristic of the menstrual cycle (∼4 weeks) and it decays in amplitude the furthest away from the alignment point (pregnancy start/dashed line) because of variation in menstrual cycle lengths for different participants.

Fig. 4

Sleep changes during pregnancy. For all the panels, the negative time values represent the 12 weeks before pregnancy (week −11 to week 0, inclusive). Results are shown for up to 38 weeks of gestation, which includes >70% of the participants (N = 2540). Zero marks the week prior to the start of pregnancy (dashed line). Week 1 refers to the first week of pregnancy. a. Time in bed during pregnancy. TIB was calculated as the time between start and end of the nighttime sleep recording and normalised using prepregnancy values per participant (nTIB). The mean was obtained by averaging the nTIB changes per participant per week and then taking the average across participants (black dots), the standard deviation across participants is depicted by the grey shaded area. b. Time in bed (TIB) chart. Time in bed was calculated as the time between start and end of the nighttime sleep recording. Each line indicates the time in bed for different percentiles. The percentiles are indicated by text above each line. c. Daytime sleeping. The number of participants who had at least one sleep recording during day time (sleep events that occur between 10 a.m. and 8 p.m.) is averaged per week (black dots) and divided over the total number of participants in the postpartum cohort. The standard deviation per week is shown in grey. d. Total sleep time (TST). TST is defined as the duration of time spent in deep, light, and REM stages. The mean of TST during prepregnancy is removed per participant, referred to as normalisation. Averaged per participant per week first and then across users. The solid black line is the mean, and the shaded grey area represents the standard deviation. e. Sleep stage changes. Fitbit’s sleep stages algorithm quantifies sleep into the following stages: deep (corresponds to sleep stage N3), light (corresponds to sleep stages N1 and N2), wake, and rapid eye movement (corresponds to sleep stage REM). The mean duration in each sleep stage during prepregnancy was subtracted per participant, referred to as normalisation. The normalised values are averaged per week for each participant and then average across participants per week. f. Sleep efficiency. Sleep efficiency is the percentage of the time in bed spent asleep (TST/TIB). Each line indicates the sleep efficiency for different percentiles. The percentiles are indicated by text below each line.

Fig. 5

Sleep and behaviour changes postpartum (weeks 1–20) relative to prepregnancy baseline (weeks −11 to 0, inclusive). For all the panels, the negative time (grey dots) represents the baseline at 12 weeks prepregnancy. The break in the x-axis (//) indicates that the time is not continuous. The dashed vertical line represents time of delivery and positive time values refer to the number of weeks postpartum. Week 1 includes the 7 days after delivery, the time series is discontinuous between 0 and 1 to indicate that these periods (prepregnancy versus after delivery) are not contiguous in time. a. Time in bed after delivery. After birth, the number of participants that get more than one sleep recording during nighttime is increased from 3.4% to 26.1% due to multiple awakening episodes (Supplemental Table S1). To account for this, TIB postpartum is the sum of TIB for all nighttime sleep events (events that happen between 7 p.m. and 11 a.m.) and the mean TIB during prepregnancy is subtracted, referred to as normalisation (nTIB). The mean was obtained by averaging the nTIB per participant per week and then taking the average across participants (prepregnancy: grey dots, postpartum: black dots). Only weeks with at least 3 nights of data were included. The standard deviation across participants is depicted by the grey shaded area (N = 2314). b. Time in bed chart. Time in bed was calculated as described in A. Each line indicates the time in bed for different percentiles. The percentiles are indicated by text below each line (N = 2314). c. Steps taken during the night. The normalised sum of steps per minute from midnight to 5 a.m. was averaged per participant per week and then across participants. Only step values below 99th percentile were included (N = 1652). The mean (prepregnancy: grey dots, post-partum: black dots) and standard deviation (grey area) are plotted. d. Daytime sleeping. The number of participants who had at least one sleep recording during daytime (sleep events that happen between 10 a.m. and 8 p.m.) is averaged per week (prepregnancy: grey dots, post-partum: black dots) and divided over the total number of participants in the postpartum cohort (N = 2314).