The Effect of the COVID-19 Pandemic on Substance Use Patterns and Physiological Dysregulation in Pregnant and Postpartum Women

Abstract

Background:

The SARS-CoV-2/COVID-19 pandemic has been associated with increased stress levels and higher alcohol use, including in pregnant and postpartum women. In the general population, alcohol use is associated with dysregulation in the autonomic nervous system (ANS) which is indexed by heart rate variability (HRV). The objectives of this study were to: 1) characterize changes in substance use during the SARS-CoV-2/COVID-19 pandemic via baseline self-report survey followed by mobile ecological momentary assessment (mEMA) of substance use, and 2) examine the associations between momentary substance use and ambulatory HRV measures in pregnant and postpartum women.

Methods:

A subset of pregnant and postpartum women was identified from the ENRICH-2 prospective cohort study. Participants were administered a baseline structured phone interview which included the Coronavirus Perinatal Experiences (COPE) survey and ascertained prevalence of substance use. Over a 14-day period, momentary substance use was assessed three times daily and HRV measurements via wearable electronics were captured. Associations between momentary substance use and HRV measures (root mean square of successive differences [RMSSD] and low frequency/high frequency [LF/HF] ratio) were examined using a mixed effects model that included within-subject (WS) and between-subject (BS) effects and adjusted for pregnancy status and participant age.

Results:

The sample included 49 pregnant and 22 postpartum women. From a combination of a baseline and 14-day mEMA surveys, 21.2% reported use of alcohol, 16.9% reported marijuana use, and 8.5% reported nicotine use . WS effects for momentary alcohol use were associated with LnRMSSD (β=−0.14; P=.005) and LnLF/HF ratio (β=.14; P=.01) when controlling for pregnancy status and maternal age. No significant associations were observed between HRV measures and instances of marijuana or nicotine use.

Conclusions:

These findings highlight the negative effect of the SARS-CoV-2/COVID-19 pandemic on psychological health of pregnant and postpartum women associated with higher substance use, which, in turn, is associated with ANS dysregulation potentially putting some women at risk of developing substance use disorder.

Introduction

The World Health Organization (WHO) declared the novel coronavirus, SARS-CoV-2, and associated COVID-19 disorder a global pandemic in March 2020 (Rajkumar, 2020). As of September 2022, there were approximately 54,000 new cases reported daily, more than 95 million total cases reported, and more than 1 million deaths nationwide (CDC, 2022). Widespread social/physical distancing measures were critical to minimize the spread of SARS-CoV-2/COVID-19; however, these measures were associated with significant consequences, including increased psychological stress, depression, anxiety, and reduced quality of life (Restubog et al., 2020). While the ramifications of infection have lessened with the advent of vaccine availability, SARS-CoV-2 continues to change resulting in the emergence of new variants that spread rapidly requiring continued safety measures (CDC, 2022).

Importantly, women are disproportionately affected by the pandemic. Recent national survey results demonstrated that 13.6% of United States (U.S.) adults reported symptoms of serious psychological distress in 2020 compared to 3.9% in 2018, and this increase was greater in women compared to men (approximately 16% vs 12% respectively) (McGinty et al., 2020). Additionally, women were reported to have significant increases in low mood, anxiety, poor sleep, loneliness, and excess alcohol consumption compared to pre-pandemic (Phelan et al., 2021). Women also disproportionately reported significantly higher symptoms of posttraumatic stress disorder (PTSD) and emotion regulation difficulties related to the pandemic compared to men (Jiang et al., 2020). COVID-19 also shifted the paradigm of typical pregnancy and postpartum experiences. In a recent nationwide survey of pregnant women (n=2,740), 93% of participants reported increased stress related to possible COVID-19 infection, 64% reported increased stress related to possible loss of income, and 56% reported stress related to loss of childcare (Moyer et al., 2020). Pregnant women also reported high levels of depression and anxiety related to the COVID-19 pandemic (Ceulemans et al., 2021).

Prior studies have demonstrated that consumption of substances, such as alcohol, increased after the experience of traumatic events (Flory et al., 2009; Ma and Smith, 2017). Emerging data on substance use during the COVID-19 pandemic are consistent with these historical reports. A recent systematic review found a trend towards increased alcohol consumption in the general population during the pandemic (Roberts et al., 2021). Similarly, findings from a 2020 international survey of adults revealed that 36% of respondents reported an increase in alcohol use during the pandemic, and those with increased alcohol consumption were more likely to be older, have children, and report depression, anxiety, or impulsivity (Sallie et al., 2020).

A common thread among stress related disorders and alcohol use is dysregulation in the dynamic feedback loop between the body and the brain (Bates et al., 2019). Key structures in this loop are located in the central autonomic network (CAN) and the peripheral autonomic nervous system (ANS) (Bates et al., 2019). The peripheral ANS regulates involuntary processes, such as heart rate, and is composed of the sympathetic nervous system (SNS) and the parasympathetic nervous system (PNS) (Waxenbaum et al., 2022). Heart rate variability (HRV) reflects the complex interplay between the SNS, which speeds things up, and PNS, which slows things down via the vagus nerve, resulting in the complex beat-to-beat variability between consecutive heartbeats (Porges, 2007). Higher HRV is generally found in healthy individuals who exhibit a high degree of resilience to change (Eddie et al., 2015). Of note, HRV is reduced in individuals (including pregnant women) with stress disorders, such as anxiety and PTSD, and substance use disorders (Braeken et al., 2013; Chalmers et al., 2014; Chang et al., 2012; Minassian et al., 2015; Ralevski et al., 2019).

During the initial stages of the pandemic, social distancing placed extraordinary limitations on clinical research prompting development and adaptation of alternative research approaches, such as remote data collection via participants’ phone and wearable electronics. Mobile ecological momentary assessment (mEMA) involves repeated sampling of an individual’s behavior in real time thus capturing fluctuations over time. Similarly, a number of commercial wearable devices have been introduced for continuous monitoring of HRV in the real-world setting (Hinde et al., 2021). Sensors worn during varying levels of activity have been reported to perform consistently and accurately in estimation of real time second-by-second heart rate (Chow and Yang, 2020). In addition, when compared to gold-standard electrocardiogram (ECG)-derived HR signals, photoplethysmography (PPG)-derived signals from fitness trackers and smart devices have been shown to provide valid results (Nelson and Allen, 2019; Nissen et al., 2022).

The objectives of this study were to: 1) characterize changes in alcohol and other substance use during the COVID-related pandemic via baseline self-report surveys followed by mobile ecological momentary assessment (mEMA) of substance use and 2) examine the associations between momentary substance use and ANS imbalance measured by ambulatory HRV. We hypothesized that alcohol use among pregnant and postpartum women would increase during the COVID-19 pandemic and this would be associated with higher physiological arousal assessed by ambulatory HRV. We also explored congruence of substance use reports via traditional phone surveys and mEMA surveys.

Materials and Methods

Study design and population

Study participants were recruited from the established prospective cohort study - ENRICH-2, and participated in an additional study visit (funded by the NIAAA Administrative Supplement mechanism). Briefly, ENRICH-2 pre-birth cohort includes four prospective study visits (V): V1) baseline prenatal visit (12-27.9 gestational weeks); V2) early third trimester (28-32 weeks); V3) assessment during the hospital stay for labor and delivery; V4) infant’s assessment at approximately 6-months of age. Inclusion criteria for the parent study were as follows: 1) singleton pregnancy, 2) gestational age: 12-28 weeks, 3) planning to deliver at the University of New Mexico (UNM) hospital, 4) planning to reside in New Mexico for 6 months after delivery; and 5) ability to provide written consent in English. Exclusion criteria (either at enrollment or disqualifiers later in the study) were: 1) prenatal use of cocaine, methamphetamines, medications for opioid use disorder (methadone, buprenorphine), or MDMA (per self-report, positive urine drug tests, or medical records review); 2) delivery <35 gestational weeks; 4) fetal/neonatal diagnosis of a major structural abnormality or severe complication.

For the Administrative Supplement sub-study, any ENRICH-2 participant who was either pregnant or had a child ≤ 18 months of age and had access to a smart-phone was eligible. Enrollment was initiated in December, 2020 and data collection was completed in May, 2022. The study was approved by the UNM Human Research Review Committee; all participants provided written informed consent. After enrollment in the sub-study, participants completed a structured phone interview with a trained study coordinator. Following the phone interview, participants were guided through the installation of the mEMA phone application (Ilumivu, Inc., Cambridge, MA, USA, https://ilumivu.com). They received cues to complete three brief mEMA surveys per day (waking, afternoon, and bedtime) over a 14-day period (42 maximum surveys per participant). All surveys were timestamped at onset and completion by the mEMA application. Each participant was also delivered a Garmin Vivosmart^® 4 fitness tracker. Participants received detailed training for use of the wearable sensor and written instructions were provided for future reference. Participants were instructed to wear the device on their non-dominant hand to continuously track heart rate for 14 days to coincide with the administration of the mEMA surveys. Data were stored locally on the participant’s smartphone until the participant was within WiFi range, at which point the data automatically downloaded to a secure server.

Measures

Coronavirus Perinatal Experiences Impact Survey (COPE-IS)

In order to assess the experiences of new and expectant mothers related to the COVID-19 pandemic, the COPE–IS was administered. The COPE-IS is a novel 50-item questionnaire that was designed to assesses the experiences of pregnant and postpartum individuals in the time of the COVID-19 pandemic, including perinatal experiences, exposures and symptoms (self and family), financial concerns (current and expected), social support, social and activity restrictions, coping and adjustment, emotions and feelings, as well as physical and mental health (Thomason et al., 2020).

COVID-19 Community Response Survey (CRS) Module 10 – Substance Abuse

Changes in substance use, including alcohol, tobacco/nicotine, and marijuana, during the pandemic were assessed at baseline through administration of a modified version of the CRS Module 10 via a phone interview. Participants were asked to report: 1) for the past 30 days, the number of days they drank, used tobacco or other forms of nicotine, or used marijuana or other forms of THC, and quantities of substance use; and 2) during the pandemic, whether they used alcohol, tobacco/nicotine, or marijuana/THC more than usual or less than usual. These surveys are part of an NIH supported and verified toolkit of recommended protocols related to COVID-19.

mEMA Surveys

At each mEMA survey, questions were asked about use of three substances since the previous mEMA survey. Specifically, participants were queried about any use since the previous mEMA survey by asking for the number of drinks for alcohol, number of puffs from cigarettes, and units of THC for marijuana. Participant responses were converted to binary variables of use or no use.

Heart Rate Variability (HRV) Measures

Garmin Vivosmart4 device uses optical technology in which a series of lights emitted by photoplethysmography (PPG) sensors against the skin illuminate capillaries to detect changes in blood volume, thus providing heart rate data in real time (Chow and Yang, 2020). HRV indices were derived via Ilumivu software (Ilumivu, Inc., Cambridge, MA, USA, https://ilumivu.com). using the inter-beat-interval (IBI) -the time between successive heart beats (also known as the period between successive ‘R’ peaks or R-R interval). The Vivosmart^® 4 device has been shown to produce reliable heart rate data with the mean absolute percentage error below the designated acceptable 10% threshold (Chow and Yang, 2020).

HRV time domain measure: Time domain measures quantify the amount of variability between consecutive heartbeats, or IBI measurements (Shaffer and Ginsberg, 2017). The most common time domain measure is the square root of the mean of the squares of differences between consecutive R-R intervals (RMSSD) which reflects parasympathetic tone and vagally mediated HRV (Shaffer and Ginsberg, 2017).

HRV frequency domain measures estimate frequency spectrum (Malik et al., 1996). The low frequency (LF; 0.04–0.15 Hz) to high frequency (HF; 0.15–0.40 Hz) (LF/HF) ratio of HRV is reported here. Increase in the HF band indicates a more mature parasympathetic system (vagal tone), and lower LF/HF ratio indicates increased parasympathetic activity, thus is often used as an index of parasympathetic tone (Shaffer and Ginsberg, 2017). Heart rate epochs of five-minute intervals immediately following completion of each mEMA survey were selected for analysis – an approach previously used in other studies utilizing mEMA and HRV measures (Dennis et al., 2017).

Data reduction and analysis

Summaries of 5-minute intervals of the HRV data were provided by Ilumivu (Ilumivu, Inc., Cambridge, MA, USA, https://ilumivu.com). Briefly, preparatory to summarizing, data were analyzed offline by using the raw IBI streamed from the Garmin device. A fast Fourier transformation (FFT) was applied separately to IBI data to extract HF-HRV from a de-trended, end-tapered IBI time series. The spectrum for the selected R-R interval segment was calculated via Welch’s periodogram method, in which R-R interval data were reduced using a window width of five minutes (300 seconds) (Fagard et al. 1998). The HRV values reported here were calculated from 5-minute intervals following the completion of an mEMA survey.

Participant characteristics and baseline survey responses are reported using means and standard deviations (SD) for continuous variables and number (percentage) for categorical variables. Differences between pregnant and postpartum women for baseline reports of substance use, changes in substance use due to COVID, any mEMA report of substance use across the 14-day survey period, and any baseline and/or 14-day mEMA report of substance use were examined using the Fisher’s Exact test. The maximum time span for HRV measurements was 300 seconds; however, for some HRV measurements there was insufficient data captured for the full 5-minute assessment period. HRV observations with recorded time less than 270 seconds (4.5 minutes) were omitted from analysis. After review for observation length, HRV data were reviewed for outliers, and observations greater than 4 SDs were removed from analyses HRV data distributions were reviewed for normality assumptions using the Kolmogorov-Smirnov test and visual quantile-quantile (Q-Q) plots. RMSSD and LF/HF ratio distributions were positively skewed, thus both HRV measures were natural log transformed.

Differences in HRV data between instances in which participants reported and did not report use of substances were examined using a random intercept maximum likelihood mixed model that adjusted for repeated observations within participants. The repeated intervals for HRV measures and substance use responses were level-1 observations within (level 2) participants. Initial linear mixed models were constructed for each type of substance use report overall and stratified by pregnancy status, with HRV intervals as the dependent variable and a binary substance use response (any vs. none) as the independent variable. Model least squares means and standard errors (SE) are reported, representing the mixed model mean estimates after adjusting for clustering within subjects. Intraclass correlation coefficients (ICC) values, representing between-subject (BS) variation are also reported for these models as well as the within-subject (WS) variation (1-ICC). Degrees of freedom (DF) and F-value associated with the Type 3 tests of fixed effects are reported for statistically significant results.

These initial models assumed that BS and WS effects were equal. To test if the BS and WS effects were equal, alternative models were created with two substance use variables: 1) for each subject, the mean of the subject’s responses for substance use (essentially the proportion of the instances that substance use was reported) to estimate and BS effect, and 2) for each substance use instance report for a subject, the difference between the instance report (0 or 1) and the mean of the subject’s responses for substance use to estimate WS effect. Initial and alternative models were compared using the log likelihood ratio test, which did not indicate significant differences between models. However, alternative models did demonstrate that differences in HRV measures were due to WS effects, so the alternative models incorporating WS and BS effects are reported. Final multivariable HRV models were random-intercept restricted maximum likelihood mixed models that included each of the substance use reports (alcohol, nicotine, and marijuana). Based on the literature, models were further adjusted for subject (level-2) factors of pregnancy status (pregnant vs. postpartum) and age (Sarhaddi et al., 2022; Shaffer and Ginsberg, 2017). Potential interaction effects between pregnancy status and reports of substance use were examined, but were not significant. Model fixed effects estimates (beta coefficients) and standard errors and Satterthwaite degrees of freedom are reported.

Statistical analyses were conducted using SAS statistical software (Cary, NC version 9.4). Analyses were two-tailed with an alpha level of .05 used to determine statistical significance; however, significance using an alpha level of .10 was also reported.

Results

Seventy-six ENRICH-2 participants (53 pregnant and 23 postpartum) were enrolled. Participants who were pregnant were recruited between the main study V1 (second trimester) and V3 (delivery). Postpartum participants were recruited on average at 13.7 (SD4.3) months after delivery (range: 3-18 months). Of the 76 enrolled participants, all completed the COPE-IS and COVID-19 telephone surveys and 71 completed at least some of the mEMA surveys (minimum=7, mean=37) and also had associated HRV data. Table 1 summarizes socio-demographic characteristics for the final sample of 71 participants. Sixty-nine percent of participants self-identified as white, slightly more than one-half of participants were Hispanic/Latina (54.9%) or had a college or professional degree (52.1%). The majority were married/cohabitating (81.7%) or were employed (66.2%). When comparing ethnicity and race between the 5 participants not included in the final sample to those included, no differences were noted.

Table 1.

Socio-demographic Characteristics of the Study Population (n=71)

Characteristics
	Mean (SD)
Maternal age (years)	30.0 (5.0)
	n (%)
Maternal ethnicity:
Hispanic/Latina	39 (54.9%)
Non-Hispanic/Latina	32 (45.1%)
Maternal race:
White	49 (69.0%)
African American	2 (2.8%)
American Indian/Alaskan Native	4 (5.6%)
Other	16 (22.5%)
Marital status:
Single/separated/divorced	13 (18.3%)
Married/cohabitating	58 (81.7%)
Maternal education:
Less than high school	7 (9.9%)
High school or GED	9 (12.7%)
Some college or vocational school	18 (25.4%)
Bachelor’s degree	22 (31.0%)
Masters, doctoral, or professional degree	15 (21.1%)
Currently employed	47 (66.2%)
Pregnancy status:
Pregnant	49 (69.0%)
Postpartum	22 (31.0%)

Over the 14-day period of mEMA collection, a total of 2,610 mEMA reports were obtained, corresponding to a response rate of 87.5% from 2,982 possible reports. For the HRV measurements, a total of 1,944 post-survey observations were obtained (65.2% of the possible observations). In total, 74.4% of mEMA responses had associated HRV measurements. Of these, 219 were <270 seconds and were omitted from further analysis. Among the remaining HRV observations, 4 RMSSD, 5 HF, and 8 LF observations were identified as outliers, resulting in 1,717 RMSSD and 1,716 LF/HF ratio observations included in analyses.

In response to a COPE survey question about how the COVID-19 outbreak changed stress levels or mental health for participants, 60.6% (n=43) responded “Worsened them moderately” and 11.3% responded “Worsened them significantly”, while 23.9% reported “No change” and 4.2% “Improved them moderately.” Responses did not differ between pregnant and postpartum participants.

Table 2 summarizes changes in substance use since the onset of the pandemic, prevalence during the 30-day period before the baseline phone survey, and prevalence during the 14-day period captured by the mEMA surveys. Some participants (7%) reported an increase in alcohol use since the onset of the pandemic, but some (3%) reported a decrease. Of note, a significantly higher percentage of postpartum (13.6%) compared to pregnant (4.1%) participants reported they drank alcohol more than usual during the pandemic. Overall, some (4%) participants reported a decrease in marijuana use while some (1%) reported an increase. The patterns of tobacco use did not appear to change. The combined prevalence of alcohol use from both a traditional survey (past 30 days) and mEMA (14 days) was 21.1%. This was followed by 16.9% combined prevalence for marijuana use and 8.5% for tobacco use. Two sources of data yielded similar results for the reported alcohol use; however, the estimated prevalence for marijuana use was slightly lower based on mEMA data (9.9%) compared to a traditional survey (12.7%). The pattern was reverse for tobacco use (8.5% on mEMA and 5.6% on a traditional survey).

Table 2.

Changes in Substance Use During Pandemic: Results of the Baseline and mEMA surveys.

Substance	Entire sample (n=71)	Pregnant women (n=49)	Postpartum women (n=22)	P valuec
Alcohol use:
General use during pandemica, N (%)
More than usual	5 (7.0)	2 (4.1)	3 (13.6)	.03
Less than usual	2 (2.8)	0 (0.0)	2 (9.1)
No change	64 (90.1)	47 (95.9)	17 (77.3)
Any use during past 30 daysa, N (%)	13 (18.3)	2 (4.1)	11 (50.0)	<0.0001
Any use during 14 days of mEMA surveyb, N (%)	14 (19.7)	1 (2.0)	13 (59.1)	<0.0001
Any use during past 30 daysa and/or 14 days of mEMA surveyb, N (%)	15 (21.1)	2 (4.1)	13 (59.1)	<0.0001
Marijuana use:
General use during pandemica, N (%)
More than usual	1 (1.4)	0 (0.0)	1 (4.5)	.19
Less than usual	3 (4.2)	3 (6.1)	0 (0.0)
No change	67 (94.4)	46 (93.9)	21 (95.5)
Any use during past 30 daysa, N (%)	9 (12.7)	6 (12.2)	3 (13.6)	>.99
Any use during 14 days of mEMA surveyb, N (%)	7 (9.9)	3 (6.1)	4 (18.2)	.11
Any use during past 30 days and/or 14 days of mEMA survey, N (%)	12 (16.9)	7 (14.3)	5 (22.7)	.50
Tobacco use:
General use during pandemica, N (%)				.53
More than usual	1 (1.4)	1 (2.0)	0 (0.0)
Less than usual	1 (1.4)	0 (0.0)	1 (4.5)
No change	69 (97.2)	48 (98.0)	21 (95.5)
Any use during past 30 daysa, N (%)	4 (5.6)	2 (4.1)	2 (9.1)	.58
Any use during 14 days of mEMA surveyb, N (%)	6 (8.5)	3 (6.1)	3 (13.6)	.29
Any use during past 30 days and/or 14 days of mEMA survey, N (%)	6 (8.5)	3 (6.1)	3 (13.6)	.37

^aResults of the COPE survey

^bNumber of mEMA responses for substance use questions: 2610; participants were queried via mEMA surveys 3 times a day for 14 days (Number of possible EMA responses: 14*3*71=2982)

^cFisher’s exact test

During the 14-day mEMA survey period, there were a total of 104 momentary reports of alcohol use from 14 women (100 reports in 13 postpartum women and 4 reports from 1 pregnant woman). There was a total of 52 momentary reports of marijuana use from 7 women (4 postpartum, 3 pregnant) and 98 of nicotine use from 6 women (3 postpartum, 3 pregnant). The smaller dataset of mEMA survey responses with HRV data was representative of responses for the larger mEMA survey response dataset. Reported use as a percentage of all responses was similar between the datasets, and the number of individuals with HRV data reporting any instances of use was the same as the number of individuals in the larger mEMA dataset, with the exception of number of postpartum participants reporting alcohol use (n=12) and marijuana use (n=3).

Mean HRV indices during the 5-min interval following mEMA completion, stratified by momentary reports of substance use (any vs. none) and by pregnant/postpartum status, are summarized in Table 3. In the entire sample, RMSSD was significantly lower in participants with any momentary reports of alcohol use (mean 38.59, SE 2.29 vs mean 43.29, SE 1.28; F_1,1644=5.54; P=.02) ; however, this association is primarily driven by reports among postpartum women as few instances of alcohol use were reported by pregnant women. The mean LF/HF ratio was significantly higher when momentary alcohol use was reported compared to no use in the entire sample (mean 1.32, SE 0.06 vs mean 1.18, SE 0.02: F₁, 1,639=4.95; P=.03), driven again by reports among postpartum women (mean 1.33, SE 0.07 vs mean 1.21, SE 0.04; F₁, 519=3.7; P=.06). No significant associations were observed between reported instances of marijuana or nicotine use and RMSSD or LF/HF ratio, although there was a trend toward higher RMSSD values in women who reported momentary marijuana use compared to no use.

Table 3.

Association Between mEMA Reported Substance Use and HRV Measures^a

	All		Pregnant women		Postpartum women
	Any Use	No Use	Any Use	No Use	Any Use	No Use
Alcohol
# of HRV obs	60	1657b	2	1173b	58	484
RMSSD	38.59±2.29	43.29±1.27*	28.39±9.59	43.37±1.62	38.91±2.63	43.16±2.01*
LF/HF ratio	1.32±0.06	1.18±0.02*	1.41±0.31	1.16±0.03	1.33 ±0.07	1.21±0.04**
Marijuana
# of HRV obs	29c	1682c	9	1160c	20c	522c
RMSSD	47.71±3.19	42.94±1.29	47.6 ±4.86	43.29±1.62	47.17±4.16	42.23±2.09
LF/HF ratio	1.15±0.09	1.18±0.02	1.11±0.15	1.16±0.03	1.18±0.11	1.23±0.04
Tobacco
# of HRV obs	76	1641d	44	1131d	32	510
RMSSD	44.16±3.34	43.03±1.29	44.89±4.27	43.28±1.63	42.37±5.18	42.60±2.06
LF/HF ratio	1.17±0.08	1.18±0.02*	1.11±0.11	1.16±0.03	1.26±0.12	1.22±0.04

^*p<0.05;

^**p<0.10

RMSSD, root-mean square differences of successive R-R intervals

LF/HF ratio: LF - low frequency (0.04–0.15 Hz); HF - high frequency (0.15–0.40 Hz) spectral components of HRV

^aAssociations are reported for HRV readings across the 5-minute interval following EMA survey completion. Values for HRV measures are least squares means ± standard error.

^bFor LF/HF ratio, the observations were: All, No Use N=1652; Pregnant, No Use N=1168

^cFor LF/HF ratio, the observations were: All, any Use N=30, No Use N=1686; Pregnant, No Use N=1165; Postpartum, Any Use N=21, No Use N=521

^dFor LF/HF ratio, the observations were All, No Use N=1636; Pregnant, No Use N=1126

Note: Increase in HF indicates a more mature parasympathetic system (vagal tone), and lower LF/HF ratio indicates increased parasympathetic activity and a balance between sympathovagal ANS functions

Results for the final multivariable mixed models are shown in Table 4. There was a significant and independent WS effect for mEMA reported instances of alcohol use on LnRMSSD (β=−0.14; P=.005) and on LnLF/HF ratio (β=.14; P=.01) when controlling for pregnancy status and maternal age. BS effects for mEMA reported instances of alcohol use were not significant, nor were WS or BS effects for mEMA reported instances of nicotine or marijuana use.

Table 4:

Association Between mEMA Reported Substance Use and HRV Measures: Results of Multiple Regression Analysis

	LnRMSSD				LnLF/HF Ratio
	Estimate (SE)	DF	t	P value	Estimate (SE)	DF	t	P value
Intercept	3.72 (0.21)	66.6	17.89	<.0001	0.34 (0.12)	63.7	2.84	.006
Within Subject:
Any Alcohol Use	−0.14 (0.05)	1,646.2	−2.83	.005	0.14 (0.05)	1,651.1	2.51	.01
Any nicotine use	0.09 (0.09)	1,643.8	0.97	.33	−0.08 (0.10)	1,644.2	−0.85	.40
Any marijuana use	0.09 (0.08)	1,644.6	1.16	.24	0.05 (0.09)	1,648.0	0.56	.58
Between Subject:
Any Alcohol Use	−0.04 (0.44)	62.7	−0.09	.93	−0.08 (0.25)	60.6	−0.34	.74
Any nicotine use	−0.22 (0.19)	61.2	−1.15	.26	0.04 (0.11)	56.3	0.35	.73
Any marijuana use	−0.15 (0.44)	65.0	−0.34	.74	−0.10 (0.26)	69.3	−0.38	.71
Pregnant (vs Postpartum)	−0.02 (0.08)	62.4	−0.21	.83	−0.08 (0.05)	58.9	−1.77	.08
Age	0.00 (0.01)	67.4	0.00	1.00	−0.01 (0.00)	64.8	−1.55	.13

Ln – natural log transformation

RMSSD, root-mean square differences of successive N-N intervals

LF/HF ratio: LF - low frequency (0.04–0.15 Hz); HF - high frequency (0.15–0.40 Hz) spectral components of HRV

Discussion

This study characterized the effect of the COVID-19 pandemic on baseline self-report (via phone survey) and momentary reports of alcohol and other substance use and ambulatory HRV in pregnant and postpartum women. The use of mEMA and continuous HRV monitoring allowed for examination of variation in substance use between participants as well as variability within individual participants. The use of a within-subject design leveraged a repeated measures design that allowed for direct comparison of participants with themselves longitudinally across contexts (Bolger et al., 2019). We were able to demonstrate relative concordance between the methods of substance use assessment and, concerningly, our findings demonstrate that almost a quarter of participants reported alcohol use and approximately 17% reported marijuana use during the pandemic (traditional survey of past 30 days and/or mEMA for 14 days). Additionally, our findings show a significant within-subject effect of alcohol use on HRV indices, after controlling for the effect of other substances, highlighting the negative effect of alcohol use on physiological balance in pregnant and postpartum women.

Critically, findings from this study demonstrated that alcohol was the most common substance reported to be used “more than usual” by pregnant and postpartum women during the pandemic. However, postpartum women reported much higher use compared to pregnant women. Fifty-nine percent of postpartum women reported alcohol use which is slightly higher than pre-pandemic national estimates for females ages 18 and over (NSDUH, 2018) Only 4.1% of pregnant participants in this study reported alcohol use compared to pre-pandemic national and global prevalence reports of alcohol use during pregnancy of approximately 10% (NSDUH, 2018; Popova et al., 2017).

Our alcohol use findings among pregnant women during the COVID-19 pandemic are consistent with other studies. For example, findings from a small (n = 83) online survey of pregnant women residing in the U.S. demonstrated that 14.5% of participants reported using at least one substance to cope with the pandemic and 5% reported alcohol use (Smith et al., 2021). Similar results were found in a large Canadian study, which demonstrated an alcohol use prevalence rate of 6.7 % among pregnant persons (Kar et al., 2021).

It is concerning that alcohol was the most common substance reported to be “used more than usual” as some women increased alcohol consumption during the pandemic. Prenatal alcohol exposure is associated with a broad spectrum of negative consequences broadly known as Fetal Alcohol Spectrum Disorders (FASD). FASD is associated with impaired neurocognitive function, self-regulation, and adaptive functions, such as impairments in communication or social and daily living skills (Kable and Mukherjee, 2017). Even low-to-moderate drinking during pregnancy has been associated with an increased risk of adverse neurodevelopmental outcomes (Lowe et al., 2022). Furthermore, maternal alcohol use after pregnancy has also been associated with negative developmental outcomes in children. For example, alcohol consumption during breastfeeding has been associated with delayed growth and lower verbal IQ scores (May et al., 2016). Additionally, evidence suggests that children of women with alcohol use disorders are at increased risk of developing depression, anxiety disorders, attention deficit hyperactivity disorder, and substance use disorders (Hill et al., 2011). However, it is important to note that mothers with risky alcohol consumption generally decrease alcohol use during pregnancy (Pryor et al., 2017). Therefore, our findings, and those of other investigators, point to the importance of incorporating assessment of alcohol and substance use among all pregnant and postpartum women.

Our results demonstrate higher rates of marijuana use among pregnant women during the pandemic compared to other studies. Approximately 14% of pregnant participants reported any marijuana use during the past 30 days and/or during the 14 days of momentary assessment. In comparison, results of a small (n = 83) online survey of pregnant women residing in the U.S. in April 2020 found that 7% reported cannabis or marijuana use (Smith et al., 2021) and a large Canadian study found 4.3% cannabis prevalence rates among pregnant persons (Kar et al., 2021). In June of 2021, New Mexico passed the Cannabis Regulation Act making the possession of up to two ounces of marijuana, or its equivalent, legal. New Mexico is the 18^th state to enact such legislation legalizing the adult use of cannabis. Concerningly, findings from longitudinal cohort studies suggest that prenatal marijuana use is associated with adverse neurodevelopmental outcomes that may be long lasting (Goldschmidt et al., 2004, 2016; Marroun et al., 2011). This highlights the critical need for further longitudinal investigation of outcomes associated with prenatal marijuana use.

Second, our findings demonstrate a significant association between alcohol use and autonomic changes. Specifically, between participants, LnRMSSD was significantly lower in the overall sample of participants with any momentary reports of alcohol use. LnRMSSD was also significantly lower in postpartum participants who reported momentary alcohol use. Additionally, LnLF/HF ratio was somewhat higher in instances when alcohol use was reported among postpartum women, although results did not reach statistical significance at alpha=.05. Findings were similar at the WS level. There was a significant, independent effect for individual instances of reported alcohol use on LnRMSSD and on LnLF/HF ratio when controlling for pregnancy status and maternal age. Lower RMSSD reflects a decrease in parasympathetic activity or vagally mediated HRV (Shaffer and Ginsberg, 2017). In pregnant and non-pregnant adults, decreased parasympathetic activity has been associated with stress disorders, such as anxiety and PTSD, and substance use disorders (Braeken et al., 2013; Chalmers et al., 2014; Chang et al., 2012; Minassian et al., 2015; Ralevski et al., 2019). LF reflects both sympathetic and parasympathetic influence while HF is an indicator of parasympathetic influence. A higher LF/HF ratio suggests increased SNS activity, associated with stress and fight-or-flight responses and a withdrawal of PNS activity that is consistent with equilibrium (Shaffer and Ginsberg, 2017). Taken together, findings of lower RMSSD and higher LF/HF ratio are reflective of decreased parasympathetic influence and indicative of altered physiologic homeostasis.

In the general population, a growing body of evidence links acute and chronic alcohol use with changes in HRV (Ralevski et al., 2019; Vaschillo et al., 2008). A recent systematic review and meta-analysis found significantly lower RMSSD in adults with alcohol use disorder compared to healthy adults (Cheng et al., 2019). Another systematic review showed that moderate alcohol consumption was associated with increased RMSSD and decreased LF/HF ratio (Karpyak et al., 2014). However, as alcohol consumption increased, RMSSD decreased and LF/HF ratio increased (Karpyak et al., 2014). Additionally, , a large (n = 4,098) cohort study conducted among adults (age 19-65 years) to examine the effect of self-reported daily alcohol consumption on ANS function assessed using wearable devices during sleep (Pietilä et al., 2018) found a significant negative effect of alcohol on HRV metrics across all alcohol doses (low [0.17 g/kg], moderate [0.45 g/kg], and high [1.1g/kg]). Specifically, alcohol intake was significantly associated with decreased RMSSD and an increased LF/HF ratio (Pietilä et al., 2018).

Mechanisms underpinning the effect of alcohol on HRV are poorly understood. One possibility is the direct effect of alcohol on the ANS. In fact, it has been suggested that acute and chronic alcohol use may actually cause damage to the vagal nerve itself (Guo et al., 1987; Weise et al., 1986). Others have suggested the observed differences in HRV measured in individuals who consume alcohol compared to those who do not is not related to a direct effect of alcohol, but the consequence of associated disorders, such as PTSD, anxiety, or depression (Eddie et al., 2015). Regardless, HRV represents a promising noninvasive measure of physical and psychological health as well as a marker of resilience and self-regulation (Carnevali et al., 2018; Kemp and Quintana, 2013; Mccraty and Shaffer, 2015).

Interestingly, we found no significant associations between reported instances of marijuana use and LnRMSSD or LnLF/HF ratio, although there was a trend toward higher LnRMSSD values at the WS level in pregnant and postpartum women who reported momentary marijuana use compared to no use. These results are consistent with findings from a sample of 276 adults in which marijuana use was not associated with changes in RMSSD during acute stress. Notably, there is a paucity of cannabis research that is inclusive of women (Pabon et al. 2022). However, Pabon (2022) found HF-HRV was significantly and negatively associated with THC dose in healthy adult women. When comparing our findings to these studies it should be noted that they were conducted in in a laboratory setting allowing for examination of dose-response relationships. Our study was conducted in a naturalistic setting making direct comparisons challenging. Therefore, future EMA research should incorporate investigation of dose-response relationships.

Similarly, we found no significant associations between momentary reports of nicotine use and LnRMSSD or LnLF/HF ratio at the BS or WS level in our sample of pregnant and postpartum individuals. By comparison, studies have demonstrated that nicotine use is associated with blunted resting-state HRV compared to those without nicotine use. For example, in a sample of 73 adolescents age 15-16, tobacco use was associated with significant decreases in respiratory sinus arrhythmia, a measure of vagus nerve mediation of parasympathetic function (Conrad, Gorka, and Kassel, 2015). Similarly, in a sample of 42 adult men, 22 who smoked tobacco and 20 who smoked tobacco and consumed alcohol, RMSSD, VLF and HF measures were significantly decreased in the smoking only group compared to healthy controls (Yuksel et al., 2016).

While our findings are preliminary, this work could ultimately inform the development of just-in-time interventions that identify ANS dysregulation in individuals and cue them to implement therapeutic interventions through the use of smartphone apps or more traditional therapeutic modalities. For example, HRV biofeedback (HRV-BFB) targets the underlying physiologic processes associated with emotion regulation and substance use (Eddie et al., 2015; Lehrer et al., 2003). By instructing individuals in this specialized paced breathing technique using biofeedback visualization of their real-time respiratory and cardiac parameters, homeostasis of brain-heart feedback loops is promoted (Eddie et al., 2014; Lehrer et al., 2003). For example, in the general adult population HRV-BFB training as an adjunct to standard treatment for alcohol abuse was associated with significant reduction in cravings for alcohol (Alayan et al., 2019; Eddie et al., 2018). Similarly, among women receiving outpatient treatment for SUD, frequent use of a brief HRV-BFB intervention delivered via a smartphone app was associated with lower levels of craving (Price et al., 2022). Other mobile, technology based, therapeutic interventions, including components of cognitive-behavioral treatment and motivational interviewing, have been shown to decrease substance use in a variety of populations (Fowler et al., 2016).

This study had several potential limitations. First, our participant sample size was relatively small. However, the principal unit of analysis was the mEMA substance use instance and continuous HRV data obtained through the wearable device. It is also possible that our findings are affected by under-report of substance use. We previously demonstrated variability in the accuracy of self-report for different classes of substances among pregnant women in the substance use treatment program (Garg et al., 2016). Interestingly, similar to previous work (Mun et al., 2021), there was a substantial agreement in reported substance use between traditional 30-day phone survey and repeated mEMA measures. Another limitation to our study is that the study encompassed the period from December 2020 to May 2022, and stress due to the COVID-19 pandemic may have varied across that time period. While our study was not designed to examine temporal effects across the pandemic, future research may want to consider incorporating phases of the pandemic into analysis. Similarly, questions about changes in substance use due to the pandemic were broadly asked using the COPE survey to allow comparison to other national surveys, and it is unknown whether pregnancy status changes during the pandemic (pre-pregnancy to pregnancy and postpartum) may have been a confounding influence. Another potential limitation to the study is unique to ambulatory monitoring of HRV compared to in-laboratory research. Ambulatory assessments of HRV are known to be affected by inherent limitations associated with the data capture and transmission (approximately 35% of HRV observations were missing). However, the Garmin Vivosmart^® 4 device has been shown to produce reliable HR data with the mean absolute percentage error below the designated acceptable 10% threshold (Chow and Yang, 2020). Moreover, ambulatory assessment of HRV is associated with unique strengths including assessment in the real-world context, longitudinal repeated measures, and convenience for participants. Additionally, HRV analyses were anchored around the reported instance of substance use on mEMA which allowed for evaluation of the proximal effect of substance use on ANS while accounting for repeated measures. While we have used methods typically employed in the field that query how many drinks a participant has consumed over a given adjacent period (Wray et al., 2014), it is possible that there may be some misclassification of substance use. Finally, although no differences in ethnicity and race were noted among the 5 participants not included in the final sample compared to those included, there is the potential that our findings could be limited by the PPG technology employed in consumer wearable devices. Questions about potential inaccuracies in PPG measurement due to variations in skin type have been raised; however, a results of recent systematic review are inconclusive (Koerber et al., 2022). This information suggests the need for further exploration in larger samples sizes with objective consideration of possible effects on skin tone on results.

Our novel findings provide insight into changes in patterns of substance use among pregnant and postpartum women during the COVID-19 pandemic, including increased alcohol consumption in some participants as well as higher reports of marijuana consumption than previously reported. While further investigation into these patterns is warranted, these findings highlight the continued need for universal screening for substance use to identify high-risk mothers and infants in order to provide them with appropriate support services. Additionally, we provide preliminary evidence that HRV metrics may serve as a clinical biomarker of ANS dysregulation in pregnant and postpartum women who consume alcohol. Importantly, this study involved repeated sampling in the natural context of women’s daily lives thus showing the feasibility of mobile health technology use in pregnant and postpartum women. Emerging findings hold the potential to inform the development of personalized interventions delivered via mobile technology to support prevention of alcohol use, restoration of autonomic balance, greater self-regulation, and improved goal directed behavior.