Abstract
Background:
Opioid use has disproportionally impacted pregnant people and their fetuses. Previous studies describing opioid use among pregnant people are limited by geographic location, type of medical coverage, and small sample size. We described characteristics of a large, diverse group of pregnant people who were enrolled in the Environmental Influences on Child Health Outcomes (ECHO) Program, and determined which characteristics were associated with opioid use during pregnancy.
Materials and Methods:
Cross-sectional data obtained from 21,905 pregnancies of individuals across the United States enrolled in the ECHO between 1990 and 2021 were analyzed. Medical records, laboratory testing, and self-report were used to determine opioid-exposed pregnancies. Multiple imputation methods using fully conditional specification with a discriminant function accounted for missing characteristics data.
Results:
Opioid use was present in 2.8% (n = 591) of pregnancies. The majority of people who used opioids in pregnancy were non-Hispanic White (67%) and had at least some college education (69%). Those who used opioids reported high rates of alcohol use (32%) and tobacco use (39%) during the pregnancy; although data were incomplete, only 5% reported heroin use and 86% of opioid use originated from a prescription. After adjustment, non-Hispanic White race, pregnancy during the years 2010–2012, higher parity, tobacco use, and use of illegal drugs during pregnancy were each significantly associated with opioid use during pregnancy. In addition, maternal depression was associated with increased odds of opioid use during pregnancy by more than two-fold (adjusted odds ratio 2.42, 95% confidence interval: 1.95–3.01).
Conclusions:
In this large study of pregnancies from across the United States, we found several factors that were associated with opioid use among pregnant people. Further studies examining screening for depression and polysubstance use may be useful for targeted interventions to prevent detrimental opioid use during pregnancy, while further elucidation of the reasons for use of prescription opioids during pregnancy should be further explored.
Introduction
The opioid crisis has differentially affected individuals by gender. While there has been a dramatic increase in the use of opioids throughout the United States, 1 –3 higher prevalence estimates among women have been described. 4,5 These previous studies indicating an increasing incidence of opioid use among reproductive-aged women is particularly concerning due to the potential for detrimental effects of opioid exposure on those of reproductive ages, and if pregnant, on the developing fetus.
Differences in opioid use by gender have been noted in the early reproductive years. For example, recent data from the Center for Disease Control's Youth Risk Behavior Surveillance Study (YRBSS) show that prescription opioid misuse in high school girls is significantly higher than in their male counterparts (8.3% vs. 6.1% for current use), and could be as high as 16.1% for lifetime use in girls. 6 Lifetime use of nonmedical prescription opioids specifically among girls in some communities may reach 20%. 7 Several barriers to contraception for reproductive-aged people who use opioids exist, 8 and higher rates of unintended pregnancy have been documented. 9 Opioid-exposed fetuses are subject to the risk of neonatal abstinence syndrome, which has also subsequently increased, 10 as well as other poor neonatal outcomes, including admission to the neonatal intensive care unit, 11 preterm birth, 11,12 stillbirth, 12 and birth defects. 13
To succeed in reducing prenatal exposure to opioids, programs, policies, and practices should be informed by empirical data that represent the diverse communities throughout the United States. However, sentinel studies of pregnancy and characteristics of prenatal opioid exposure have been limited to Medicaid participants, 14,15 a specific state or community 16 –19 or rely solely on self-report of opioid use. 20,21 The concern with relying on estimates from previous studies is that the rates of opioid use in pregnancy have been shown to differ widely across type of medical insurance and geographic location. 10,22 These differences indicate that previous studies do not adequately describe the individuals in the United States who use opioids during pregnancy, limiting the potential impact of public health interventions aimed at reducing use in pregnancy, and subsequently, fetal exposure.
The Environmental Influences on Child Health Outcomes (ECHO) Program is a National Institutes of Health (NIH)-funded consortium of pediatric cohort studies across the United States designed to investigate the effects of early life exposures on five key areas of child health: pre-, peri-, and postnatal health; obesity; respiratory conditions, including asthma; neurodevelopment; and positive health
23
/wellbeing.
24
The goal of ECHO is to leverage existing pediatric studies by combining data collected under cohort-specific protocols, in addition to data collected under a standardized ECHO-wide protocol
25
(
Using ECHO data from 21,905 pregnant people participating in cohorts enrolled at sites across the United States, we sought to describe the epidemiology of opioid use among pregnant people, including demographic and medical characteristics of these individuals, and presence of polydrug use in pregnancy. Secondarily, we estimate the association between characteristics of pregnant people in ECHO and use of opioids during pregnancy.
Results from this investigation may inform policies for primary and secondary prevention strategies to screen for opioid use during pregnancy.
Material and Methods
ECHO program
The ECHO-wide cohort has enrolled more than 57,000 children and 61,000 pregnant participants since its launch in 2016; participants represent diverse racial, sociodemographic, and geographic contexts across the United States. 26 Data from the individual cohorts that were collected before the onset of ECHO were harmonized; harmonization is a process by which data from different sources are combined so that they are comparable for collective analyses and interpretation. The study protocol was approved by the local and/or central ECHO Institutional Review Board. Written informed consent or parent's/guardian's permission was obtained along with child assent as appropriate, for ECHO-wide Cohort Data Collection Protocol participation and for participation in specific cohorts.
Study population
Participants were included in this analysis if there was nonmissing information on maternal prenatal opioid use. Of the 61,749 pregnant participants enrolled in the ECHO study, 21,905 pregnancies from 52 recruiting sites—which represented 32 cohorts—had information on prenatal opioid use and were included in this analysis. Figure 1 depicts the geographic location of ECHO principal investigators contributing to this study.
Map of geographic locations of principal investigators contributing to the sample of 21–905 pregnancies for this study from the Environmental Influences on Child Health Outcomes study.
Prenatal substance use
The primary outcome of interest was opioid use during pregnancy. We defined opioid use as any use of an opioid medication (either prescribed to the pregnant person or otherwise obtained) or heroin during pregnancy, which was ascertained by self-report, medical record abstraction, or maternal toxicology screen. Opioid medication included oxycodone and methadone. Those opioid medications administered during labor and delivery were not considered prenatal opioid use and therefore were not included in our definition of prenatal opioid use. We defined prescription use, nonprescription use (misuse or recreational use), trimester of use during pregnancy, and specific medication or drug types when data specificity allowed. As defined by the Centers for Disease Control, the three critical time periods (“overdose waves”) for opioid prescription behaviors in the United States were used in analyses of calendar time for our sample: 1990–2009, 2010–2012, and 2013–2021. 27
Similarly, we defined individual substance use during pregnancy based on any tobacco, alcohol, marijuana, or the use of other illegal drug use, respectively. The use of illegal drugs included use of any recreational or street drugs, including cocaine, heroin, methamphetamines, 3,4-methylenedioxy-methamphetamine (ecstasy), speed (amphetamine sulfate), acid or lysergic acid diethylamide, and ketamine, excluding marijuana use. We defined tobacco use as: cigarette smoking; e-cigarette smoking (i.e., vaping, Electronic Nicotine Delivery Devices [ENDS], Vape pens or mods); or other forms of tobacco (chewing tobacco/snuff, nicotine patch, nicotine gum/lozenges, cigar, pipe, hookah, Bidi [Beedi]).
Sociodemographic variables
Data capturing the highest level of education for pregnant individuals was harmonized into three categories: less than high school, high school, or some college and above. Prenatal household income data were harmonized into five categories (<$30,000; $30,000–$49,999; $50,000–$74,999; $75,000–$99,000; and $100,000 or more). Race and ethnicity were categorized as non-Hispanic (NH)-White, NH-Black, NH-other, and Hispanic categories. The NH other race category included those who were non-Hispanic and Asian, Alaska Native, American Indian, native Hawaiian, or Pacific Islander. Child's race/ethnicity, if available, was used for those pregnant people who were missing maternal race/ethnicity information. Parity was categorized as having no previous births, 1, 2, and 3+ births before the index pregnancy.
Prenatal marital status was categorized as a binary variable defined as married or living with a partner or not married (widowed; separated; divorced; single, never married; partnered, not living together). Insurance type was harmonized into four categories: government-assisted health care (Medicare, Medicaid, medical assistance, Children's Health Insurance Program, or any other kind of government assistance plan based on income or disability of the biological mother); private insurance, including Tricare/military health care and Indian Health Service; other insurance; and no insurance.
Medical history
Medical history was based on the history of the pregnant person. Depression and anxiety were defined as a self-reported diagnosis of a depressive disorder, or anxiety disorder, or a diagnosis in the medical record (ICD-9 296.2, 296.3, 296.9, 300.0, 300.2, 300.4, 309.8; ICD-10 F32–F34, F39, 40.0–40.2, 40.8, 40.9, 41.0, 41.1, 41.8, 41.9, 42, 43.1), at any point before the pregnancy through 8 weeks postpartum, indicated by medical records or self-report.
Antidepressant use was defined with medical records or self-report of any of the following: selective serotonin reuptake inhibitors (SSRI), serotonin norepinephrine reuptake inhibitor, tricyclic antidepressants, monoamine oxidase inhibitors, n-methyl D-aspartate antagonist, neuroactive steroid gamma-aminobutyric acid-A-receptor positive modulator, and atypical antidepressants.
The SSRI group included the Food and Drug Administration (FDA)-approved medications of Citalopram/Celexa; Escitalopram/Lexapro; Fluoxetine/Prozac, Sarafem, Symbyax; Fluvoxamine/Luvox; Paroxetine/Paxil, Pexeva; Sertraline/Zoloft; Vortioxetine/Trintellix; Vilazodone/Viibryd.
Statistical analysis
We described maternal sociodemographic characteristics, prenatal substance use, and history of depression and anxiety for women who used and did not use opioids during pregnancy. Means and standard deviations (SD) are presented for continuous variables, whereas the number of observations and percent of total observations are presented for categorical variables. We performed chi-square and t-tests or Fisher exact tests for categorical and continuous variables, respectively. Due to the high amount of missing data for certain variables, we also present the number and percent of pregnant people for whom any data are available for each variable of interest. Among those who used opioids during pregnancy, additional details of the opioid use are provided, if available (Table 1).
Characteristics of 21,905 Pregnant People in Environmental Influences on Child Health Outcomes with Opioid Use During Pregnancy
Bold values indicate total number for the given category and percentage of the total for the column.
p-value comparing characteristics between opioid exposed and nonexposed by Chi-square test, Fisher's Exact Test, or t-test.
Includes any recreational, illegal, or street drugs, such as Cocaine, Heroin, Methamphetamine (Meth), MDMA (ecstasy), Speed (amphetamine sulfate), Acid/LSD, Special K (ketamine), and others.
SSRI, selective serotonin reuptake inhibitor.
We estimated the odds of opioid use during pregnancy using logistic random effects regression models that included sociodemographic factors, comorbidities, and prenatal use of other substances when the variable had <50% missingness. Random intercepts for cohort effects were used to account for the differences in selection criteria and sample characteristics between cohorts. Models were fit using the SAS procedure GLIMMIX (SAS Institute Inc., Cary, NC, USA). Unadjusted odds ratios (ORs) and 95% confidence intervals (CI), as well as odds ratios adjusted (aORs) for select characteristics were calculated. Covariates included in the final adjusted models included age at delivery, race/ethnicity, calendar year of child birth, education, parity, history of depression, and prenatal substance use of alcohol, tobacco, marijuana, and any illegal drugs.
Missing data were addressed in our analyses with multiple imputation (MI) using the fully conditional specification with a discriminant function 28 ; imputation was performed for: age at delivery, race/ethnicity, education, history of depression, and prenatal alcohol, tobacco, illegal drug, and marijuana use. Imputation models included the prenatal opioid use outcome, all covariates in the main analysis model, and cohort-ID as a classification variable, and were run by three cohort clusters as defined by cohort selection criteria: general population cohorts (n = 28), cohorts enrolling only in a neonatal intensive care unit (n = 3), and an adoption cohort (n = 1) whose participants were recruited through domestic adoption agencies following the birth of the child. The imputation models presented in Table 2 combine estimates from 25 imputations. There were no substantial differences in the results between the nonimputed values and the multiple imputation values.
Unadjusted and Adjusted Odds Ratios and 95% Confidence Intervals from Missing Indicator and Multiple Adjustment Analyses to Examine Correlates of Opioid Use in Pregnancy Among Individuals Enrolled in the Environmental Influences on Child Health Outcomes Study
Includes any recreational, illegal, or street drugs, such as Cocaine, Heroin, Methamphetamine, MDMA, Speed, Acid/LCDL, Ketamine, and others.
Covariates used in adjusted models: maternal age at delivery, maternal race/ethnicity, calendar year of child birth, maternal education, parity, maternal history of depression, and prenatal substance use of alcohol, tobacco, marijuana, and any illegal drugs.
Using imputed data for all variables in Table 2, including the ECHO cohort ID in the multiple imputation model by running three cohort clusters (recruitment from the general population, neonatal intensive care units, and adoptions) for 25 imputations.
Results
In this study of the 21,905 pregnant people with opioid use information, 591 (2.7%) individuals were determined to have used opioids during pregnancy. Table 1 shows participant demographics, stratified by opioid use. Participants who used opioids during pregnancy were more likely to be non-Hispanic White and have a lower household income with Medicare/Medicaid insurance coverage. The mean age at delivery for both groups of pregnant people was 30 years (SD 5.7 years). Among the study participants who used opioids in pregnancy, 11% had a high school education or less, whereas only 8% of the participants who did not use opioids in pregnancy were in this education category (p = 0.006). Thirty-two percent of the individuals who used opioids were also nulliparous compared with 38% of nonopioid users.
Participants who used an opioid during pregnancy, 45% had a history of depression and 37% had a history of anxiety, compared with 20% and 13% among women who did not, respectively (p < 0.001). Fifteen percent of those who used an opioid during pregnancy also used antidepressants during pregnancy compared with 7% of pregnant people without opioid use. (p < 0.001) In this sample, the use of SSRIs specifically during pregnancy was higher for pregnant people who also used an opioid compared with their nonopioid using peers (10% vs. 5%, p < 0.001; Table 1).
We observed that vast majority (86%) of opioid users in this sample were users of prescription opioids, and only 5% reported using heroin during pregnancy. Table 1 also shows that polysubstance use was more common among opioid users compared with nonusers. Among the individuals who used an opioid during pregnancy, 32% reported ever alcohol use, 39% reported tobacco use, 16% reported marijuana use, and 10% reported use of illegal drugs compared with 19% for alcohol, 11% for tobacco, 5% for marijuana, and 1% for use of illegal drugs among those who did not use an opioid during pregnancy. The total number of methamphetamine and cocaine users was 53 and 63, respectively. We faced high levels of missing data (45% missing) for trimester-specific patterns of opioid use; from existing data, 51% of opioid users identified use during the third trimester.
Table 2 describes the estimates from the unadjusted, adjusted, and multiple-imputation models of having an opioid-exposed pregnancy for individuals in the ECHO. In adjusted analyses, we observed a reduced odds associated with pregnant people younger than 20 years of age compared with those 25–30 years of age (MI-aOR = 0.54 [95% CI 0.30–0.97]). Non-Hispanic Black individuals, non-Hispanic Asians, and non-Hispanics of other races was associated with decreased odds of prenatal opioid use compared with Non-Hispanic White individuals. We did not observe an association with education and prenatal opioid use. Infants born between 2010 and 2012 to our participants had a significantly higher odds of prenatal opioid exposure than did those born earlier; and higher order pregnancies (2 and ≥3 previous births) had a significantly higher odds compared with those from people who are nulliparous. History of depression was positively associated with the use of prenatal opioids (MI-aOR = 2.42 [95% CI: 1.95–3.01]).
With regard to other substance use during pregnancy among our study participants, prenatal maternal use of tobacco was associated with higher odds of prenatal opioid use (MI-aOR = 2.98 [95% CI: 2.38–3.73]), but prenatal use of alcohol and/or marijuana use was not. The use of any illegal drugs was associated with higher odds of opioid use (MI-aOR = 2.0 [95% CI: 1.3–3.1]).
Discussion
Principal findings
In our large, geographically diverse sample, we found that 2.8% of pregnancies were exposed to opioids, and that the overwhelming proportion of this opioid use in pregnancy was reported as originating from a prescription drug. This finding was independent of the calendar year of the pregnancy, indicating that regardless of recent interventions to reduce overprescription of opioids, particularly among those of reproductive ages, 29 prescription opioids remain a threat.
We identified that non-Hispanic White race, polysubstance use, and depression were statistically and clinically significantly associated with opioid use during pregnancy. It is yet unclear how behavioral interventions on substance use and depression could be upstream to opioid use; for example, it is unclear if successful treatment of depression may be beneficial in reducing future opioid use. Further research ought to elucidate the temporality between depression and initiation of substance use in reproductive-aged individuals at risk for opioid use.
Results in the context of what is known
Our results largely support previous studies that have found similar correlates to opioid use in pregnancy. However, some community-specific studies have found high rates of opioid-exposed pregnancies among certain samples of pregnant Black, Indigenous, and women of color, 30,31 which we failed to find in our study. Conversely in our medically and socioenvironmentally diverse multicohort sample of over 20,000 pregnancies, we observed a significantly lower odds of prenatal opioid use for non-Hispanic Black, Asian, and other races; we observed no difference between non-Hispanic White people and Hispanic people with regard to their prenatal use. Dawson et al also found that non-Hispanic White pregnant controls in the National Birth Defects Prevention Study had the highest rates of opioid use in pregnancy in nearly every year from 1998 to 2011. 32 While there is evidence that non-Hispanic White pregnant individuals are more likely to receive medication for opioid use disorder than their African American and Indigenous counterparts, 33,34 there is a paucity of data on similar characteristics of pregnant individuals in the absence of opioid use disorder and neonatal opioid withdrawal syndrome.
We also observed an association between opioid use and other substance use, which is not surprising as polysubstance use among women with opioid use disorder has increased recently. 35 Page et al have reported high levels and frequent use of marijuana in pregnancy among women with opioid use disorder, with particularly high rates in the presence of concomitant alcohol use in pregnancy. 36 Whereas Gaalema found that 18% of pregnant women enrolled in a tobacco cessation program also used opioids during the pregnancy. 37
While our findings about the co-occurrence of opioid use in pregnancy with depression, 38 –40 and polysubstance use during pregnancy are predominantly replicative, we include a large comparison sample of pregnant people who did not use opioids prenatally, and tested models that accounted for multiple sources of variation in opioid use. As a result of this sampling and statistical methodology, the broad generalizability and other strengths to our design strengthen previous findings on determinants of prenatal opioid use.
Clinical implications
The majority of opioid use was reportedly from prescriptions originally, and therefore, future clinical recommendations should be initiated to monitor over-prescription and encourage clinical counseling and other treatment interventions. Our finding that tobacco increased opioid use in pregnancy by nearly three-fold presents a particularly difficult clinical scenario because both substances are associated with preterm delivery. 41 Clinical interventions that quickly identify pregnancies at risk and interventions that ameliorate the risks of multiple threats for a term delivery should be identified.
Research implications
Our findings will help inform future research designed to identify screening tools, including those focused on polysubstance use. We found that tobacco and the use of illegal drugs were independently associated with opioid use in pregnancy. Therefore, screening tools for opioid use could focus on polysubstance use. Such screening tools will enable future research that examines the factors that may lead to polysubstance use, and its interrelationship with depression.
Strengths and limitations
This is the largest observational epidemiology study to examine maternal opioid use during pregnancy. Our sample size allowed us to examine rare exposures—such as polysubstance use—at the community level and when complete data are difficult to obtain, as is the case when assessing illegal (i.e., use of illegal drugs) and stigmatizing behaviors (i.e., prenatal tobacco use) and conditions (i.e., depression). Unfortunately, our use of existing cohort data led to a relatively high degree of missing data because, before the initiation of ECHO, data collection methodology varied across cohorts. We accounted for the missing data with the use of multiple imputation, which have previously been shown to provide valid estimates when data are missing at random. 28 We report the multiple-imputation adjusted results, and we found that these estimates were relatively stable, with similar results to analyses that included “missing” as a predictor category. We also had lack of detail on the timing of exposure during the pregnancy and type of polysubstance use, if present.
The pregnancy data moving forward in ECHO will be collected using a uniform ECHO-wide protocol. Finally, some of our data were collected from self-report and this method has been shown to undercount actual substance use in pregnancy, and that the level of underreporting is greater for substance users than the general population.
Conclusions
Our results indicate that opioid use in pregnancy is more prevalent in non-Hispanic White individuals, pregnancy during an era of rapid opioid availability, and in the presence of a history of depression and polysubstance use. It will be important to evaluate the potential benefits of screening for characteristics that may co-occur with addictive behavior, such as depression and polysubstance use, to identify those at risk for prenatal opioid exposure. While our data were limited on specific details of the type of opioids used, we provide evidence that much of the opioid use during pregnancy originated from prescriptions. Therefore, to address the potential risks with pregnancy, policies to reduce opioid availability and programs focused on addressing prescription use for the management of both pain and opioid use disorder, should be further explored.
Footnotes
Acknowledgments
The authors wish to thank our ECHO colleagues, the medical, nursing, and program staff, as well as the children and families participating in the ECHO cohorts. They also acknowledge the contributions of the following ECHO program collaborators: the Coordinating Center: Duke Clinical Research Institute, Durham, NC, USA: Smith PB, Newby KL, Benjamin DK2; The Data Analysis Center: Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, USA: Jacobson LP; and The Research Triangle Institute, Durham, NC, USA: Parker CB.
Author Disclosure Statement
No competing financial interests exist.
Funding Information
This work was supported by the Environmental influences on Child Health Outcomes (ECHO) program, Office of The Director, National Institutes of Health, under Award Numbers U2COD023375 (Coordinating Center), U24OD023382 (Data Analysis Center), U24OD023319 (PRO Core), and 5UH3OD023320-05 (Aschner), UG3OD023337 (Wright), UH3OD023268 (Litonjua), 4UH3OD023248-03 (Dabelea), 5UH3OD023348-05 (O'Shea), 4UH3OD023328-03 (Duarte), UH3OD023279 (Elliott), 4UH3OD023342 (Newschaffer), UH3OD023289 (Ferrara), 4UH3OD023365-03 (Hertz-Picciotto), 4UH3OD023244-03 (Hipwell), 4UH3OD023275-03 (Karagus), UH3OD023347 (Lester), UH3OD023347 (McEvoy), UH3 OD023389 (Leve), UH3 OD023349 (O'Connor), R01 HD 034568 and UH3OD 023286 (Oken), 4UH3OD023271-03 (Karr, Sathyanarayana), UH3 OD023285 (Paneth), 4UH3OD023282-03 (Gern). The authors wish to thank our ECHO colleagues and contributors, and the following ECHO program collaborators: Coordinating Center: Duke Clinical Research Institute, Durham, North Carolina: Benjamin DK, Smith PB, Newby KL; and the Data Analysis Center: Johns Hopkins University Bloomberg School of Public Health, Baltimore, Maryland: Jacobson LP; Research Triangle Institute, Durham, North Carolina: Parker CB; the ECHO cohorts listed in Supplemental Data 
. Data utilized in these analyses may be available through approval from individual study investigators and the ECHO program. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.