Detecting sleep and physical activity changes across the perinatal period using wearable technology (2024)

Background

Pregnant women may not experience disruptions in sleep duration throughout the course of pregnancy, however, their sleep quality is dramatically impaired. Sleep quality deteriorates throughout pregnancy, reaching its lowest in the third trimester. The purpose of this study was to understand the change in sleep patterns across the perinatal period, as well as the impact of physical activity on sleep.

Methods

A total of 18 physically active women trying to conceive wore a WHOOP strap (a fitness monitor) across the perinatal period. Daily behavior changes were tracked including time awake, hours in deep sleep, physical activity, and time in moderate to vigorous physical activity.

Results

Women maintained overall physical activity levels during and after pregnancy and averaged 20.70min of physical activity and 6.97h of sleep per day. Total time in awake hours increased postpartum. Moderate-vigorous physical activity minutes improved deep sleep hours overall (Est. = 0.003h, p < 0.0001) and during pregnancy (Est. = 0.00001h, p = 0.0004). Similar effects were found for all activity minutes, although in post pregnancy the moderating impact of activity minutes no longer maintained significance (p = 0.09).\.

Conclusions

Wearable technology, including fitness monitors such as WHOOP straps offer a convenient and less invasive way to track sleep and physical activity during the perinatal period. The findings of this study indicate a positive connection between sleep and engaging in moderate to vigorous activity and any activity throughout the perinatal period. These results may help inform clinical and practical recommendations for physical activity to improve sleep outcomes for pregnant women.

AbstractSection Current Knowledge/Study rational

There is existing evidence to show that sleep quality declines throughout pregnancy. However, a study continuously gathering sleep data during pregnancy and after delivery has not been conducted. Additionally, less is known about how physical activity during this time may affect sleep quality.

AbstractSection Study impact

This study followed women throughout and after their pregnancy, using wearable technology to track sleep and physical activity during this time. The findings show that physical activity regardless of intensity is associated with a positive impact on sleep. Clinical recommendations regarding physical activity could potentially help improve the lives and sleep of women during this time.

In a society with hundreds of remedies to address poor sleep, an area of the lifespan has been often ignored by these, even though lack of quality sleep at this time may have the most lasting implications. Pregnancy and the post-partum period are times of significant physiological change, which can drastically impact sleep. Although pregnant women may be sleeping a similar or greater amount to non-pregnant women, the quality of that sleep may be substantially impacted, especially as they near their delivery date [1,2,3]. Common self-reported reasons for poor sleep include frequent urination [4], gastroesophageal reflux [5], insomnia, body aches [6], anatomical changes including a growing uterus and increased body weight [7], and the addition of a newborn with inconsistent sleep-wake cycles of their own [8]. Some reasons for sleep alterations during this period that are more subtle include hormonal influences [4, 9], sleep disordered breathing [10], restless leg syndrome [11], and changes to sleep architecture itself, which has been shown to change between trimesters [6, 10] and in the post-partum period [1].

Outcomes related to sleep disturbances and changes have been measured in this population primarily using subjective measures, like the Pittsburgh Sleep Quality Index (PSQI) [12]. A recent meta-analysis estimated that 45.7% of pregnant women had poor sleep quality [1] (as defined by PSQI scores > 5) [12]. While commonly used in the research and clinical setting to screen for sleep disorders in pregnant patients, subjective sleep measures must be interpreted with caution as they are generally less accurate than objective measures [13] by overestimating sleep latency [14] and both over and underestimating sleep duration [14, 15].

There are fewer studies assessing objective sleep; typically measured by polysomnography (PSG), which is the “gold standard” [16]. A review of PSG features during pregnancy showed that there are decreases in slow wave sleep (SWS) and rapid eye movement (REM) sleep stages during the third trimester [17]. Although it provides the most accurate measure of sleep, the use of PSG for research or as a screening tool is limited because it is labor intensive and has a high cost [18]. Additionally, a recent review on sleep in pregnancy highlighted the need for more objective measures to assess sleep since it is a current gap in the literature [3]. This review also recommended the need for longitudinal study of these objective measures to understand the full scope of these changes.

Other less invasive and more cost-effective wearable sleep monitors are becoming widely available for the general population [19]. Wearable devices record physiologic data similar to PSG in a less invasive way and can be used outside of the clinical environment for multiple nights. However, limited validation of commercial wearable devices is a barrier to their widespread use in research or clinical settings [20]. These devices have also not yet been substantially studied in the pregnant or post-partum population with regards to sleep data acquisition outside of exploratory studies [21]. One such wearable device from WHOOP (Strap 2.0; WHOOP, Inc., Boston, MA, USA) has been shown to have high concordance with PSG, with bias error ranging from 0.01% on REM sleep duration to 5.5% on non-REM (NREM) sleep duration and precision error ranging 6.69–7.8% respectively [22].

Due to existing evidence that objective and subjective sleep quality may diminish during pregnancy, there is a need to determine what interventions can help to improve sleep quality. There is some existing evidence that physical activity can help improve sleep during pregnancy. A randomized control trial of aquatic physical activity and sleep during the second and third trimester showed significant improvements in self-reported sleep quality among the intervention group [23]. Another indicated that reported moderate levels of physical activity during the first and third trimester were associated with better sleep quality [24]. Other studies specifically suggest that insufficient physical activity is associated with poor sleep quality during pregnancy and that increased exercise in early pregnancy can improve this [25, 26]. A systematic review and meta-analysis on physical activity and sleep during pregnancy indicated that pregnant women who regularly exercised had improved sleep quality [27]. However, to our knowledge, none of these studies explore the association of sleep and physical activity from pre-pregnancy to postpartum while using more objective measures such as wearable devices to measure sleep.

The aim of the present study is to investigate objective changes in sleep pre-pregnancy and throughout the postpartum period as measured by the wearable WHOOP strap. Specifically, our first aim was to use descriptive analyses to examine sleep patterns during the pre-conception, pregnancy, and post-partum time periods in a cohort of healthy women who were physically active at least three days a week prior to conception. Our second aim was to understand the relationship between physical activity during pregnancy and variations in sleep patterns, including awake hours, light sleep hours, deep sleep hours (also called slow wave sleep), and REM sleep hours. We hypothesized that increased activity would be associated with positive changes in sleep patterns during pregnancy and the immediate post-partum period after controlling for pertinent sleep covariates, such as seasonality and weekend sleep patterns.

Recruitment

Following Institutional Review Board approval (#1912819563) at West Virginia University (WVU), a total of 38 women were enrolled following informed written consent from March 2019 until August 2020. Inclusion criteria included women ages 18–35 years old who were currently physically active at least three or more times per week. Currently, ACOG recommends that women who regularly exercised before pregnancy could continue with no adverse effects, which is why individuals had to currently have a moderate level of physical activity to participate [28]. Women who were not currently pregnant, but hoping to conceive within the next six months, were also asked to participate. Prior research on this sample illustrating heart rate variability (HRV) and resting heart rate (RHR) findings during the perinatal period have already been published [29]. Recruitment occurred through the Reproductive Endocrinology and Infertility clinic at WVU and data was collected between March 2019 and July 2021.

Wearable device

Participants were given a WHOOP^® Strap 2.0 (WHOOP, Inc., Boston, MA, USA) and asked to wear it continuously from the time of enrollment and throughout pregnancy, delivery, and postpartum. The WHOOP^® strap has reasonable validation with PSG and has been suggested as a practical alternative for that more invasive measure [30]. The WHOOP^® devices were purchased with internal research funds through the WVU Center for Reproductive Medicine and provided to participants to keep as an incentive. Participants wore the strap on their nondominant arm allowing it to transmit continuous data to the participant’s smartphone and to a WHOOP cloud platform. Participants were able to see their daily physical activity information on their phones. The comprehensive data from all participants was then downloaded from the WHOOP cloud platform for analysis.

Measures

Data were imported from WHOOP^® in two different tables: (1) daily sleep measures and (2) recorded activity per participant. These tables were then combined with pertinent pregnancy information (i.e., date of conception, date of birth event). More details about variables utilized within this study from these three tables are presented below.

Daily sleep measures were assessed using reflectance photoplethysmography and motion with a three-axis accelerometer. That data was then put into algorithms to generate statistics on the participant’s sleep. Daily sleep included four measures: awake hours, light sleep hours, deep sleep hours, and REM sleep hours. Some measures also included a percentage of time asleep in each of those four assessments. These variables have been independently validated against sleep logs, polysomnography, and patient-reported sleep outcomes [22].

Recorded Daily Activity was measured by the WHOOP^® strap by a three-axis accelerometer and processed using a proprietary algorithm to create daily activity records. Recorded Activities were merged into a day-by-day measure of total time by day spent in each activity, and the three tables were merged by user ID and date. Daily minutes spent in a recorded activity in any zone were included as daily activity minutes, and daily minutes of zone 3 heart rate and higher were converted into a daily moderate/vigorous minutes variable.

Pregnancy, Seasonality, and Weekend Timing Information. These sleep and activity tables were connected to the women’s pertinent pregnancy dates, including conception and delivery date. Season was calculated by stripping the month from the date variable, and recoded as Spring (March-May), Summer (June-August), Fall (September-November), and Winter (December-February). The dates were also used to inform a Weekend/Weekday variable, which was then coded as 1 for Weekday, and 0 for Weekend. After calculating time and date variables, these dates were stripped from the data and not otherwise utilized.

Data analysis

The longitudinal data was assessed using a linear mixed models approach. After assumptions were checked and found satisfactory, several models were tested, including random intercept, random slope, both random intercept and slope, along with continuous (e.g., day to delivery) and categorical (e.g., trimester) time effects. To constrain time of pregnancy, analyses were restricted to 43 weeks prior to delivery and 8 weeks postpartum; this also ensured that most participants had data for the majority of the time period. Splines were fitted for non-linear patterns including continuous time fixed effects. Models were selected for best fit using lowest Akaike information criteria (AIC). Kuder-Richardson degree of freedom correction was used for all models. Restricted Maximum Likelihood (REML) method was used for all available data, which is appropriate for use with missing at random (MAR) or missing completely at random (MCAR) data. The best fitting model included a random intercept and random continuous time slope (days to delivery), with a variance components covariance matrix and splines set at birth. Two models are presented for each of the four outcomes (awake hours, light sleep hours, deep sleep hours, and REM sleep hours): Model 1: main effects only, including activity minutes per day, days until delivery, days after delivery, season with winter as the referent group, and weekday with weekend as the referent group; Model 2: all main effects in model one, but additionally includes the moderating effects of activity minutes per day on days until delivery and days after delivery. Fixed effects estimates along with standard errors, df, t-value and p-value are presented for each model. Sensitivity analyses included replacing activity minutes with moderate/vigorous activity minutes per day; all effects were similar although slightly attenuated so only the activity minutes models are included here. Additional sensitivity analyses included percentage of time spent in REM and deep sleep. In general, model fit was improved with hours spent in each sleep stage included in the model compared to using the percentage of each sleep stages’ duration. Thus, only models in hours are presented here. All data were analyzed using SAS 9.4 [31]. All descriptive statistics are reported either as frequencies and valid percentages of categorical variables, or mean, standard deviation, minimum and maximum values for continuous variables. For some descriptive analyses, data were summarized by both participant and week of pregnancy.

Participants

An initial 38 participants were enrolled in the study, although eight withdrew due to no longer planning to conceive, stated discomfort of the wearable device, or infertility stress. Additionally, a total of 12 participants did not conceive during the study period, leaving birth data for 18 participants. Women were followed for an average of 405.83 days (SD = 153.71; minimum = 142, maximum = 754) with a total of 7,305 days logged. Participants on average took 5.5 months to conceive. Few medical conditions were reported in monthly surveys that the women took, and all participants reported being moderately physically active throughout the perinatal period. All women gave birth between 37 and 41 weeks of pregnancy (39.16 ± 1.26), and their average age at time of delivery was 31.67 years (SD = 2.66). All births in this study were singleton, and the majority were delivered vaginally (61.11%). Further demographics can be found in Table1.

Changes during pregnancy

Based on WHOOP^® data, activity levels decreased over the course of the pregnancy from almost 28min of logged daily activity pre-pregnancy to 14min by the third trimester (Table2), with a slight increase during post-partum (almost 16min). There was a similar pattern for time spent in deep sleep (in hours) by trimester (decreased from 1.4h to 1.2h, and then a slight improvement to 1.3h during post-partum period). Inversely, the number of awake hours (during sleep periods) steadily increased throughout the entire study (from 0.65h to 1.17h). REM sleep hours steadily increased until birth (from 1.6 to 2.0) with a decrease post-partum that was still higher than pre-pregnancy amounts (1.8). Light sleep hours were more variable, increasing from pre-pregnancy (4.0) to first trimester (4.2) and then decreasing steadily to post-partum rates (3.4).

Model results

The linear mixed model results are presented for awake hours (Table3), light sleep hours (Table4), deep sleep hours (Table5), and REM sleep hours (Table6). Both main effects models (Model 1) and the adjusted model controlling for the moderating effect of activity minutes (Model 2) are included in all tables. The presented results focus on interpretation of the moderating effects models given improved model fit.

Awake hours. Consistent with descriptive results, we found an increase in awake hours both during pregnancy (Est. = 0.002, p = 0.0004) and post-partum (Est. = 0.01, p < 0.0001). This means that per day of pregnancy, there was a 0.002 increase in the number of awake hours during sleep. Both Spring (Est.=0.07, p = 0.02) and Summer (Est.=0.011, p = 0.0007) were associated with an increased time in awake hours relative to winter. Additionally, during the weekday, women spent less time awake relative to the weekend (Est. = -0.05, p = 0.02). No moderating effect of physical activity was found for awake hours during pregnancy; however, increased physical activity in minutes post-partum demonstrated a significant decrease of awake time (Est. = -0.001, p = 0.006).

Light sleep hours. More physical activity minutes were associated with an increase in light sleep (Est. = 0.004, p = 0.0002). Light sleep hours generally decreased across pregnancy (Est. = -0.004, p < 0.0001). There was an increase in light sleep hours during the Spring relative to Winter (Est. = 0.16, p = 0.0004) and a decrease on weekdays relative to weekends (Est. = -0.07, p = 0.02). More physical activity minutes during pregnancy increased light sleep hours over the course of the pregnancy (Est. = 0.00003, p < 0.0001).

Deep sleep hours. More physical activity minutes were associated with an increase in deep sleep (Est. = 0.002, p = 0.0006). Deep sleep hours generally decreased across pregnancy (Est. = -0.001, p < 0.0001) but improved post-partum (Est.=0.004, p < 0.0001). There was a decrease in deep sleep hours on weekdays relative to weekends (Est. = -0.05, p = 0.0001). More physical activity minutes during pregnancy were associated with increased deep sleep hours over the course of the pregnancy (Est. = 0.000001, p = 0.008).

REM sleep hours. More physical activity minutes were also associated with an increase in REM sleep (Est. = 0.005, p < 0.0001). REM sleep hours rapidly decreased post-partum (Est. = -0.005, p = 0.0002). A decrease in REM sleep was experienced during the Fall (Est.=-0.18, p < 0.0001) and Summer (Est. = -0.09, p = 0.004) relative to Winter; and a decrease in REM sleep hours on Weekdays relative to Weekends (Est. = -0.13, p < 0.0001). More physical activity minutes during pregnancy increased REM sleep hours over the course of the pregnancy (Est. = 0.000001, p < 0.0001); however, more physical activity minutes during post-partum appeared to decrease REM sleep hours (Est. = -0.0001, p = 0.00).

The cohort of 18 women demonstrated continued physical activity throughout pregnancy and following their delivery. Activity minutes (of any intensity) and vigorous activity minutes decreased in the third trimester and then increased again postpartum. Sleep duration was just below 7h per night (M = 6.97, SD = 1.59) and decreased in the third trimester and during postpartum. These results show some interruptions to both sleep and exercise in the final trimester and following birth, which is expected based on physiological changes to prepare for birth, as well as lifestyle changes following birth.

REM sleep when measured with the PSG typically has been found to decrease during pregnancy [32]. Instead, REM sleep typically transitions to more NREM sleep closer to birth. Our findings suggest that REM sleep slightly increases closer to birth, in contradiction with previous findings. More research is needed to understand these differences, some of which could be due to measurement type (PSG vs. WHOOP).

The overall findings indicate that moderate to vigorous physical activity minutes were associated with overall deep sleep hours during pregnancy, but not in postpartum. Similar effects were found for all activity minutes, regardless of intensity, although moderation by physical activity minutes was no longer significant postpartum.

However, it is also of note that these women were not meeting the guidelines of 150min of moderate to vigorous physical activity (MVPA) per week. The average minutes of MVPA per day was about 13min per day, totalling on average, 91min per week. In this area of Appalachia, however, only 46.8% of women meet the physical activity guidelines of 150min of physical activity per week [33]. Other research has suggested only 17.7% of pregnant women meet these guidelines [34]. So, although these rates of MVPA physical activity are not at the recommended levels, they are consistent with what is found in this population.

Additionally, although the number of physical activity minutes, especially later in pregnancy, was quite low, this could still have a meaningful effect on sleep and health. A recent study indicated that it was the number of steps per day (or overall physical activity) rather than MVPA that was associated with cardiovascular health [35]. Additionally, women’s overall sleep percentage increased by.09 extra points when they did light exercise in the afternoon. Another study comparing moderate aerobic exercise to usual prenatal care found improvements in all areas of sleep quality, although these were not statistically significant [36]. This also indicates that even when the effect of physical activity may be lower, there is still clinical and meaningful significance to the individual.

These results corroborate previous studies and meta-analyses indicating that physical activity may be associated with sleep quality [23,24,25, 27]. Additionally, this study extends the research regarding this topic by providing evidence of how physical activity affects sleep pre-conception as well as during postpartum.

Results suggest that any type of activity as well as moderate to vigorous activity is associated with improved sleep quality across the perinatal period. The implications of these findings provide greater understanding to both clinicians as well as their patients about the multi-faceted importance of remaining physically active throughout the perinatal period. It also helps empower pregnant women by understanding that any level of physical activity might help, which assists with motivation if individuals are less willing or able to undertake higher intensity exercise.

Additionally, because this sample includes some women pursuing and conceiving through assisted reproductive technology (ART), the conclusions from this study may be clinically useful in improving fertility outcomes. Research has shown that sleep duration and quality can be a modifiable target to increase the number of oocytes retrieved during a fertility cycle [37]. Other research indicates that poor sleep quality contributes to psychological distress for couples pursuing ART and that it can affect pregnancy outcomes [38]. Therefore, understanding the ways in which physical activity is associated with improved sleep may indirectly contribute to better fertility outcomes for this population throughout the perinatal period. The use of wearable technology for these issues among individuals pursuing ART (and non-ART samples) can provide individuals and their clinicians with more immediate data with which to modify physical activity and sleep quality.

Strengths, limitations, & future directions

This study has several strengths. First, although there are only 18 participants included in this study, a substantial amount of data was obtained for each participant due to continuous monitoring by the WHOOP^® straps, which provided an adequate data set for analysis. Second, the WHOOP^® strap resulted in daily readings showing patterns for sleep and exercise for each day of pregnancy. Third, we followed women throughout the perinatal period, allowing for a more in-depth look into sleep and exercise changes throughout that time. Fourth, this wearable device is a more objective measure of sleep than other studies, which have used the PSQI.

Some women included in the analysis were seeking fertility treatment, which limits the generalizability of the study to other pregnant women. Additionally, the WHOOP^® strap is not quite as accurate as clinical measures, but still provides meaningful data regarding sleep and other indicators. Another limitation is that we cannot determine directionality (causality) for the changes in the variables. Although physical activity was found to be associated with improved sleep, it could be that individuals with improved sleep are more able to engage in physical activity. Additionally, since the WHOOP^® strap is worn on the wrist, it cannot measure body position. The WHOOP^® strap, although validated against the PSG [30], has not currently been validated during pregnancy, which is needed to fully understand how it performs in this population. Finally, the act of tracking behaviors and seeing results in real time may contribute to behavior change; therefore, some changes to physical activity and/or sleep may have been related to tracking behaviors [39].

Future research will focus on larger samples, as well as using additional measures to assess body position and posture. There is burgeoning research to suggest that sleeping posture during the perinatal period can be associated with placental location [40]. More work also needs to be done to validate the WHOOP strap against the PSG in pregnancy. Understanding sleep duration, quality, and position could provide more insight into maternal and child health and outcomes.

AIC:

Akaike information criteria

ART:

Assisted reproductive technology

HRV:

Heart rate variability

MAR:

Missing at random

MCAR:

Missing completely at random

MVPA:

Moderate to vigorous physical activity

PSQI:

Pittsburgh Sleep Quality Index

PSG:

Polysomnography

REM:

Rapid eye movement

RHR:

Resting heart rate

REML:

Restricted Maximum Likelihood

SWS:

Slow wave sleep

WVU:

West Virginia University

Authors and Affiliations

1. Department of Social & Behavioral Sciences, West Virginia University School of Public Health, 64 Medical Center Drive, P.O. Box 9190, Morgantown, WV, 26505, USA

   Elizabeth A. Claydon

2. Department of Epidemiology & Biostatistics, West Virginia University School of Public Health, Morgantown, WV, USA

   Christa L. Lilly&Erin D. Caswell

3. Department of Obstetrics & Gynecology, Baptist Memorial Hospital, Memphis, TN, USA

   Dawna C. Quinn

4. Department of Obstetrics & Gynecology, West Virginia University School of Medicine, Morgantown, WV, USA

   Shon P. Rowan

Contributions

E.A.C. conceptualized the paper format and led manuscript writing, editing. S.D. conceptualized the research study, helped to recruit participants, and assisted with study oversight. D.C helped with patient recruitment and assisted with study oversight. C.L. ran and oversaw all statistical analyses. E.D.C assisted with statistical analyses and tables. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Corresponding author

Correspondence to Elizabeth A. Claydon.

Ethical approval and consent to participate

This study was approved by West Virginia University’s Institutional Review Board (#1912819563). The study complies with all requirements of the Declaration of Helsinki. Informed consent was obtained by all participants prior to enrolling in this study.

Consent for publication

Not applicable.

Clinical trial number

Not applicable.

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions