SARS-CoV-2 infection during pregnancy is associated with increased risk of hospitalization, intensive care admission,1 preterm delivery, and stillbirth.2,3 Although pregnancy was an exclusion criterion in pivotal trials of COVID-19 mRNA vaccines, recommendations for COVID-19 vaccination include pregnant people. Current guidance from the US Centers for Disease Control and Prevention (CDC)4 and American College of Obstetricians and Gynecologists5 recommends that all pregnant people receive a primary series and booster dose, with a preference for mRNA vaccines. Despite this, COVID-19 vaccine coverage among pregnant people remains low compared with similarly aged nonpregnant individuals.6
Evidence of the benefits of COVID-19 vaccination may increase confidence in vaccination among pregnant people. Moreover, because mRNA vaccines are a new class of vaccines with limited human use before the COVID-19 pandemic, postlicensure effectiveness assessments among pregnant people can shed light on whether immune changes associated with pregnancy affect mRNA vaccine performance. Vaccine effectiveness (VE) estimates among pregnant people are accruing,7-11 yet, to our knowledge, few studies have focused on more severe COVID-19 outcomes or provided variant-specific estimates. We analyzed data from pregnant people in the VISION Network12 to estimate mRNA VE of 2 doses and of a single booster dose against laboratory-confirmed COVID-19–associated emergency department and urgent care (ED/UC) visits and hospitalizations during periods of Delta and Omicron variant predominance.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and was reviewed and approved by the institutional review boards at participating sites and under a reliance agreement between CDC and the Westat institutional review board. This activity was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy (45 CFR part 46 and 21 CFR part 56). This study presented minimal risk to participants because there was no interaction or intervention with patients; therefore, a waiver of informed consent was granted.
Data Source and Study Design
VISION Network methods have been previously published.12 Briefly, VISION leverages electronic medical record data from 306 hospitals and 164 ED/UC facilities across 10 US states to evaluate COVID-19 VE using a test-negative design. This analysis included hospitalizations or ED/UC visits (referred to together as medically attended events) among persons aged 18 to 45 years with a COVID-19–like illness (CLI) diagnosis (based on discharge codes for acute respiratory illness or signs or symptoms of COVID-19) who underwent molecular testing (primarily real-time reverse transcriptase polymerase chain assay) for SARS-CoV-2 during the 14 days before through 72 hours after the medical encounter and who were pregnant at the time of the encounter. Four categories of codes were considered: (1) acute respiratory illness, including respiratory failure, viral or bacterial pneumonia, asthma exacerbation, influenza, and viral illness not otherwise specified; (2) nonrespiratory COVID-19–like illness diagnoses including cause-unspecified gastroenteritis, thrombosis, and acute myocarditis; (3) respiratory signs and symptoms consistent with COVID-19–like illness, including hemoptysis, cough, dyspnea, painful respiration, or hypoxemia; and (4) signs and symptoms of acute febrile illness. One code in any of the 4 categories was sufficient for inclusion. Whether CLI encounters were associated with delivery and pregnancy status at the time of vaccination and CLI encounters was ascertained based on established approaches at each site (eTable 1 in the Supplement). Three health systems used modifications of a published algorithm,13 ,14 4 used delivery flags from pregnancy data tables based on electronic medical records, and 1 relied on International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes for CLI encounters and surrounding medical encounters. To avoid classifying hospitalizations with incidental SARS-CoV-2 infections among persons screened when admitted for labor and delivery as COVID-19–associated hospitalizations, some relatively nonspecific ICD codes for CLI signs and symptoms used in prior VISION publications12 and in the ED/UC outcome for this analysis were excluded from the CLI definition for the hospitalization outcome (eTable 2 in the Supplement). A complete list of diagnostic codes included in CLI case definitions for ED/UC visits and hospitalizations for this analysis is presented in eTable 2 in the Supplement. Data on race (Black, White, and other [Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, multiracial, and other not listed]) and ethnicity (Hispanic and non-Hispanic), which are potential confounders of the association between vaccination and medically attended COVID-19, were obtained from the health systems’ electronic medical record.
The analysis was stratified into 2 periods of predominant variants based on state and local surveillance data: Delta (when the Delta variant accounted for ≥50% of new cases with sequencing data, June 1 to December 15, 2021, depending on site) and Omicron (when the Omicron variant accounted for ≥50% of new cases, December 16, 2021, to February 26, 2022, depending on site) (eTable 3 in the Supplement). Start dates (eTable 3 in the Supplement) varied accordingly and ranged from June 1, 2021, in Colorado to September 11, 2021, in Texas (the earliest date data were available for this site); the end date was June 2, 2022.
To aid in interpretation of the VE results among pregnant people, VE was also estimated for nonpregnant women aged 18 to 45 years in the VISION network for the same time periods and using the same analytic methods. A sensitivity analysis was also conducted excluding participants with any documented SARS-CoV-2 positive result on a molecular test 14 or more days before the current CLI encounter.
Exposures of primary interest were vaccination with 2 doses (14-149 and ≥150 days prior to an eligible CLI event) and 3 doses (7-119 and ≥120 days prior) of mRNA vaccine (either BNT162b2 [Pfizer-BioNTech] or mRNA-1273 [Moderna]), with at least 1 dose administered during the same pregnancy associated with the CLI event. As a secondary analysis, we examined the VE of 2 doses (14-149 and ≥150 days prior) and of 3 doses (7-119 and ≥120 days prior) received any time before the CLI event, including people who only received doses before pregnancy. The following exclusion criteria were applied: recipients of Ad.26.COV2.S (Janssen vaccine [Johnson& Johnson]); recipients of exactly 1 or more than 3 doses of mRNA vaccine; and those for whom fewer than 14 days had elapsed since second dose or fewer than 7 days since third dose.
Outcomes of interest were COVID-19–associated ED/UC CLI visits and COVID-19–associated CLI hospitalizations, defined by a CLI medical encounter (as previously described) with a positive SARS-CoV-2 test result in pregnant persons (and in nonpregnant women aged 18-45 years for reference), during periods of Delta and Omicron variant predominance.
The association between prior vaccination (most recent dose during pregnancy) and laboratory-confirmed COVID-19 medical encounters (ED/UC visits and hospitalizations) was estimated by comparing the odds of prior vaccination among patients with CLI and positive SARS-CoV-2 results (ie, cases) and those with negative results (ie, controls) using logistic regression, with VE estimated as (1 − adjusted odds ratio [aOR]) × 100%. As in previous VISION analyses, ORs were adjusted for age, geographic region (8 regions, representing the composite of counties with facilities for each health system) (eTable 3 in the Supplement), calendar time (days from January 1, 2021, to the CLI event, based either on preceding test or medical encounter date), and local virus circulation.12 Propensity scores were calculated based on the probability of vaccination in test-negative control participants as previously described for VISION,15 with the addition of estimated gestational age (included as both a continuous variable and a categorical variable [as estimated trimester of pregnancy]) because of its potential to confound the association between vaccination and SARS-CoV-2 infection among pregnant people. Factors unbalanced after estimation of propensity scores were included as covariates in the models. The referent group for all VE analyses was 0 doses (unvaccinated individuals). All analyses were stratified by period of predominant SARS-CoV-2 variant. VE estimates were not calculated for strata that included fewer than 20 encounters with prior vaccination.
A standard mean or proportion difference of 0.20 or greater indicated a nonnegligible difference in distributions of characteristics by vaccination or infection status. When calculating SMDs for differences in characteristics across COVID-19 vaccination status, we calculated SMD as the average of the 3 absolute values of the SMD for unvaccinated vs each vaccination status category individually (unvaccinated vs 2 doses 14-149 days earlier, unvaccinated vs 2-dose ≥150 days earlier, and unvaccinated vs 3 mRNA doses ≥7 days earlier). The average SMD calculation comparing negative SARS-CoV-2 test result and positive SARS-CoV-2 test result was generated by directly calculating the SMD for negative SARS-CoV-2 test result and positive SARS-CoV-2 test result.
For VE estimates, 2-sided 95% CIs were calculated for each VE point estimate, and nonoverlapping intervals were considered to be significantly different. Analyses were conducted using R version 4.1.2 (R Project for Statistical Computing).
Among 5492 encounters, there were 4517 eligible CLI ED/UC visits among pregnant persons (eFigure 1 in the Supplement); 885 (19.6%) had a positive SARS-CoV-2 test result; the median age among participants was 28 (24-32) years (Table 1); 537 (11.9%) were non-Hispanic Black, 1189 (26.3%) Hispanic, and 2286 (50.6%) non-Hispanic White. The timing of ED/UC visits during pregnancy was 1218 (27.0%) first trimester, 1647 (36.5%) second trimester, and 1652 (36.6%) third trimester. A total of 2861 visits (63.3%) occurred during Delta variant predominance and 1656 (36.7%) during Omicron variant predominance. Among pregnant persons with a CLI ED/UC visit, 3380 (74.8%) were unvaccinated; among those vaccinated with at least 1 dose during pregnancy, 721 (16.0%) had received 2 doses of mRNA vaccine, including 523 (11.6%) with second dose 14 to 149 days prior and 198 (4.4%) with second dose 150 or more days prior; 416 (9.2%) had received 3 doses, including 351 (7.8%) with third dose 7 to 119 days prior and 65 (1.4%) with third dose 120 or more days prior. Among those vaccinated with documented product information, 769 (67.6%) received BNT162b2 and 321 (28.2%) mRNA-1273. Vaccination was initiated before pregnancy in 589 individuals (51.8%); 341 (30.0%), 192 (16.9%), and 15 (1.3%) received their first dose of mRNA vaccine during the first, second, and third trimesters of pregnancy, respectively.
During Delta variant predominance, the estimated VE of 2 mRNA vaccine doses 14 to 149 days prior against COVID-19–associated ED/UC visits was 84% (95% CI, 69%-92%) (Figure 1); for 2 doses received 150 or more days prior, it was 75% (95% CI, 5%-93%). The estimated VE of 3 doses 7 to 119 days prior was 81% (95% CI, 30%-95%); the sample was not sufficient to estimate VE for 2 doses 120 or more days prior. During Omicron variant predominance, the estimated VEs against COVID-19–associated ED/UC visits were 3% (95% CI, −49% to 37%) for 2 doses within 14 and 149 days and 42% (95% CI, −16% to 72%) for 2 doses 120 or more day prior. The estimated VE of 3 doses 7 to 119 days prior was 79% (95% CI, 59%-89%); the VE for 3 doses 120 or more days prior was −124% (95% CI, −414% to 2%).
Among 975 eligible hospitalizations among pregnant persons, 334 (34.3%) had a positive SARS-CoV-2 result; the median (IQR) age among participants was 31 (26-35) years (Table 2); 118 (12.1%) were non-Hispanic Black, 240 (24.5%) Hispanic, and 440 (45.1%) non-Hispanic White. The timing of hospitalizations during pregnancy was 31 (3.2%) in the first trimester, 161 (16.5%) second trimester, and 783 (80.3%) third trimester. Two-thirds of CLI hospitalizations (638 of 975 [65.4%]) (Table 2) were associated with delivery. CLI hospitalizations among pregnant people who delivered, who did not deliver, and among nonpregnant women were similar in terms of severity characteristics (Table 3). Among CLI hospitalizations associated with delivery, the frequency of preterm delivery (<37 weeks gestation) was 38.3% (245 of 638) overall and 54.9% (90of 164) among those who were SARS-CoV-2 positive. A total of 679 hospitalizations (69.6%) occurred during Delta variant predominance and 337 (34.6%) during Omicron variant predominance. Among pregnant persons with a CLI hospitalization, 670 (68.7%) were unvaccinated; among those vaccinated with at least 1 dose during pregnancy, 199 (20.4%) had received 2 doses of mRNA vaccine, including 121 (12.4%) with second dose 14 to 149 days before and 78 (8.0%) with second dose 150 or more days before; 106 (10.9%) had received 3 doses, including 82 (8.4%) with third dose 7 to 119 days before and 24 (2.5%) with second dose 120 days or more before. Among the vaccinated with documented product information, 187 (61.3%) received BNT162b2 and 111 (36.4%) mRNA-1273. Vaccination was initiated prior to pregnancy in 118 individuals (38.7%) and in the first, second, and third trimesters of pregnancy among 84 (27.5%), 84 (27.6%), and 19 (6.2%), respectively.
During Delta variant predominance, the estimated VE of 2 mRNA vaccine doses 14 to 149 days prior against COVID-19–associated hospitalization was 99% (95% CI, 96%-100%) (Figure 2); for 2 doses 150 or more days prior, it was 96% (95% CI, 86%-99%). The estimated VE of 3 doses 7 to 119 days prior was 97% (95% CI, 79%-100%); the sample was not sufficient to estimate VE for 3 doses 120 or more days or more prior. During Omicron variant predominance, the estimated VE of 2 doses 14 to 149 days prior was 86% (95% CI, 41% to 97%); for 2 doses 150 or more days prior it was 64% (95% CI, −102% to 93%). The estimated VE of 3 doses 7 to 119 and 120 or more days prior was 86% (95% CI, 28% to 97%) and −53% (95% CI, −1254% to 83%), respectively.
A sensitivity analysis among pregnant persons regardless of timing of vaccination (ie, not excluding those who received all doses before pregnancy) yielded VE estimates that were not significantly different from the primary results (eTables 4 and 5 and eFigures 2 and 3 in the Supplement). VE estimates among nonpregnant women were also not significantly different from estimates among pregnant persons, although point estimates varied and confidence intervals were narrower in nonpregnant women because of larger sample size (eTables 6 and 7 in the Supplement and Figures 1 and 2). VE estimates limited to pregnant people with no prior documented SARS-CoV-2 infection were not significantly different from estimates that included all eligible regardless of prior infection (eTables 8 and 9 and eFigures 4 and 5 in the Supplement).
This study found high maternal protection against medically attended COVID-19 in the Delta period for 2 and 3 mRNA doses and highlighted the importance of a booster dose for both ED/UC and hospitalization end points during Omicron predominance. As with other evaluations of mRNA COVID-19 vaccines in adults,16 estimated effectiveness against ED/UC encounters was generally lower than for hospitalizations, and this difference was most marked for 2 doses during the Omicron period. Additionally, decreased protection of the primary series after 150 or more days since the second dose against the hospitalization end point was more evident during the Omicron period than during the earlier Delta period, highlighting the importance of booster doses among pregnant persons, in line with current recommendations.4,5
Prior assessments of VE among pregnant women preceded booster recommendations for this population, focused primarily on the pre-Delta period, had limited sample size to assess more serious COVID-19 end points, and evaluated limited durations since the second dose.7-11 These prior studies suggested the mRNA vaccine primary series among pregnant people was associated with strong protection against COVID-19 disease and SARS-CoV-2 infections in the period shortly after the second dose, similar to VE assessments among general adult populations. The estimates of effectiveness among pregnant women presented in this study provide more current and complete evidence, and continue to suggest that mRNA vaccine administration during pregnancy does not alter vaccine performance despite immune differences between pregnant and nonpregnant people.17 The patterns we observed among VE estimates for pregnant people were similar to those for women aged 18 to 45 years who were not pregnant at the time of their CLI episode, although the estimates for this latter group, due to larger sample size, had tighter confidence bounds; they were also consistent with VE estimates among general adult populations from other studies.16,18,19 The estimated VE among pregnant people receiving at least 1 dose during pregnancy was also similar to that among pregnant people vaccinated before or during pregnancy.
The findings reported here that COVID-19 maternal vaccination appears to protect pregnant people against medically attended COVID-19 is important given accruing evidence of increased risks of severe maternal morbidity and mortality associated with COVID-19 during pregnancy.1-3,20 Although two-thirds of CLI hospitalizations were associated with delivery, COVID-19 hospitalizations among delivering pregnant people were similar in terms of length of stay (>70% greater than 2 days) and had higher proportions of mechanical ventilation and intensive care unit admission than nondelivery and nonpregnant COVID-19 hospitalizations, suggesting that the outcome used for this study did not reflect incidental infection among women admitted for delivery. Moreover, among COVID-19–associated hospitalizations during which delivery occurred, 54.9% delivered preterm, highlighting the frequency of adverse pregnancy outcomes among pregnant women with COVID-19. While we were able to demonstrate apparent significant and substantial protection against COVID-19–associated hospitalizations and ED/UC visits, we did not have a sufficient sample size to assess whether maternal vaccination was associated with attenuation of the COVID-19 course among hospitalized pregnant people with COVID-19, as has been described for mRNA vaccines in the general adult population.21 Additional evaluations are also needed for a full assessment of waning protection after a booster dose, which has been observed among nonpregnant adults.16
Direct maternal benefits of COVID-19 vaccines add to the accumulating picture of the broader benefits that extend beyond the pregnant person to the fetus and newborn. A recent test-negative design study among infants younger than 6 months hospitalized with CLI in the United States22 found that maternal receipt of 2 doses of mRNA vaccine during pregnancy was less common among polymerase chain reaction– or antigen-positive case infants (16%) than among polymerase chain reaction– or antigen-negative control infants (32%), yielding an estimated VE for young infants of 61% (95% CI, 31%-78%). Additionally, a recent safety evaluation of a cohort of more than 46 000 live births in the United States23 found 2 doses of mRNA vaccine in the third trimester to be associated with reduced risk of preterm delivery (adjusted hazard ratio. 0.82; 95% CI, 0.72-0.94), and a large safety study in Ontario, Canada, that focused on a range of pregnancy outcomes and newborn health indicators24 suggested that maternal vaccination might protect against low Apgar scores (an assessment of infant condition soon after birth) and neonatal intensive care unit admission. Pregnant people in the United States have lagged behind similarly aged nonpregnant adults in primary series and booster dose coverage.6 The growing evidence of the benefits of vaccination during pregnancy, along with postlicensure safety information, addresses important knowledge gaps that arose from the exclusion of pregnant individuals from mRNA vaccine clinical trials and could improve acceptance of COVID-19 vaccination among pregnant people.25,26
This analysis is subject to several limitations. First, the methods used to identify pregnancy and determine gestational age were not standardized across sites; misclassification of pregnancy status at the time of vaccination and CLI medical encounter may have occurred. Second, the threshold for ED/UC care-seeking during pregnancy may be lower than for nonpregnant adults and thus the ED/UC outcome may capture milder infection events. Third, clinicians may be more inclined to admit unvaccinated than vaccinated patients (regardless of pregnancy) because of the potential for developing severe disease, which could theoretically bias toward a higher VE for the hospitalization outcome. Fourth, the analysis did not exclude immunocompromised individuals, for whom an additional primary dose of mRNA vaccine is recommended 4 weeks after dose 2; however, immunocompromised status was uncommon among pregnant people with CLI (<1% of ED/UC visits and <5% of hospitalizations). Fifth, COVID-19 vaccine coverage among pregnant people, particularly for a third dose, was low; the potential for bias arising from differences between vaccinated and unvaccinated individuals that affects all observational VE studies may be even more relevant for this population. Although weights for propensity-to-be-vaccinated and relevant covariates were included in VE estimation models, there is still potential for residual or unmeasured confounding. Sixth, case counts during Omicron predominance were lower than during Delta predominance, resulting in more limited power for VE estimation during Omicron, particularly for the hospitalization outcome. Seventh, due to limited sample size we could not adjust VE estimates for county clusters within health systems but did adjust for the catchment region of each health system.12 Eighth, the characteristics of pregnant people included in VISION may not be generalizable to the pregnant population in the United States, although the VISION network covers 8 health systems across 10 states.
In this study, maternal mRNA COVID-19 vaccination, including booster dose, was associated with protection against medically attended COVID-19. VE estimates were higher against COVID-19–associated hospitalization than ED/UC visits and lower against Omicron than Delta, and protection waned over time, particularly during Omicron predominance.