The effect of COVID-19 vaccination and booster on maternal-fetal outcomes: a retrospective multicenter cohort study

Summary Background: COVID-19 infection in pregnant people has previously been shown to increase the risk for poor maternal-fetal outcomes. Despite this, there has been a lag in COVID-19 vaccination in pregnant people due to concerns over the potential effects of the vaccine on maternal-fetal outcomes. Here we examine the impact of COVID-19 vaccination and booster on maternal COVID-19 breakthrough infections and birth outcomes. Methods: This was a retrospective multicenter cohort study on the impact of COVID-19 vaccination on maternal-fetal outcomes for people that delivered (n=86,833) at Providence St. Joseph Health across Alaska, California, Montana, Oregon, New Mexico, Texas, and Washington from January 26, 2021 through July 11, 2022. Cohorts were defined by vaccination status at time of delivery: unvaccinated (n=48,492), unvaccinated propensity score matched (n=26,790), vaccinated (n=26,792; two doses of mRNA-1273 Moderna or BNT162b2 Pfizer-BioNTech), and/or boosted (n=7,616). The primary outcome was maternal COVID-19 infection. COVID-19 vaccination status at delivery, COVID-19 infection-related health care, preterm birth (PTB), stillbirth, very low birth weight (VLBW), and small for gestational age (SGA) were evaluated as secondary outcomes. Findings: Vaccinated pregnant people were significantly less likely to have a maternal COVID-19 infection than unvaccinated matched (p<0.0001) pregnant people. During a maternal COVID-19 infection, vaccinated pregnant people had similar rates of hospitalization (p=0.23), but lower rates of supplemental oxygen (p<0.05) or vasopressor (p<0.05) use than those in an unvaccinated matched cohort. Compared to an unvaccinated matched cohort, vaccinated people had significantly lower stillbirth rate (p<0.01) as well as no difference in rate of PTB (p=0.35), SGA (p=0.79), or rate of VLBW (>1,500 g; 0.31). Vaccinated people who were boosted had significantly lower rates of maternal COVID-19 infections (p<0.0001), COVID-19 related hospitalization (p<0.05), PTB (p<0.05), stillbirth (p<0.01), SGA (p<0.05), and VLBW (p<0.01), compared to vaccinated people that did not receive a third booster dose five months after completing the initial vaccination series. Interpretation: COVID-19 vaccination protects against adverse maternal-fetal outcomes with booster doses conferring additional protection against COVID-19 infection. It is therefore important for pregnant people to have high priority status for vaccination, and for them to stay current with their COVID-19 vaccination schedule. Funding: This study was funded by the National Institute for Child Health & Human Development and the William O. and K. Carole Ellison Foundation.


2022.
A histogram of the counts by gestational week during which people achieved full vaccination status was generated. Another histogram reported the number of days after receiving the second COVID-19 vaccination that people received the third COVID-19 vaccine booster dose.
In addition, the type of mRNA COVID-19 vaccination received by the vaccinated population and the timing the COVID-19 vaccine doses received in relation to the person's pregnancy were reported. The COVID-19 infections the proportions that occurred during the different dominant strains and the highest level of care (none, outpatient, emergency care, or inpatient) was reported for vaccinated and unvaccinated individuals. A Fisher's exact test was performed to evaluate the difference in population distribution between these two cohorts using R stats (version 4.1.1). All graphs were generated using python package matplotlib (version 3.4.2).

Classification models
To evaluate which demographic or social determinants of health features were most predictive of vaccination status at delivery (vaccinated versus unvaccinated) we trained multiple supervised learning models on 20 features. These features included demographics, insurance, habits, comorbidities, and geographical features (Supplemental Table 4).
Race was coded as a binary variable (not identifying = 0; identifying = 1) for the four races for which there were at least 1,000 people identifying as that race. Pregravid BMI was separated into five standard categories with missing values encoded as -1. The correlation of each feature with vaccination status was evaluated using Pearson correlation calculated by the python library scipy (version 1.6.2).
Models were generated using python package sklearn (version 1.0.2) with default settings for logistic regression, gradient boosting regression, and random forest. The models were trained on 80% of the data with 20% of the data withheld for evaluation of the final performance of the model. Model performance was evaluated using mean absolute error, mean squared error, root mean squared error, area under the curve (AUC), and R 2 . Gini feature importance was used to assess the marginal contribution and influence of each feature on the final model providing interpretation of the machine learning models. In addition, on 1,000 randomly selected patients from the test set the Shapley additive explanations (SHAP) was used to evaluate the average marginal contribution of a feature value across all permutations of features providing insight into the degree of influence of the feature on an individual's predicted vaccination status at delivery for the gradient boosting models. A limited version of the top performing model (gradient boosting regression) was also trained using the top six most important features from the full model.
Predictive models were reported following the TRIPOD guidelines. 3

Reverse kaplan-meier curve
A Kaplan-Meier curve reporting the event of a COVID-19 infection was reported over a five-month (152 day) period using python package lifelines (version 0.27.0). This was reported starting at the day of full-vaccination status (14 days after the second dose) for either the mRNA-1273 Moderna or BNT162b2 Pfizer-BioNTech vaccine.
Unvaccinated people were matched to vaccinated people by conception date and T0 was the date the corresponding vaccinated person reached full vaccination status. All people were pregnant for the entire five-month period. A log-rank test was used to evaluate the difference in event occurrence between these cohorts. The graph was plotted to show the percentage of people that had a COVID-19 infection starting at 0% at T0.

Quantitative statistical analyses
Violin plots were used to report the number of days from full vaccination status to breakthrough status for people vaccinated with mRNA-1273 Moderna or BNT162b2 Pfizer-BioNTech as well as those that were boosted or vaccinated, but not boosted. Differences in these populations were evaluated using a Mann-U Whitney test using python package scipy (version 1.6.2). Rates in COVID-19 infections and related hospitalization along with the 95% Confidence Interval (CI) were calculated using the Wilson Score Interval using python package statsmodels (version 0.12.2) for unvaccinated, vaccinated, one mRNA dose, vaccinated with Ad26.COV2.S Johnson & Johnson, and COVID-19-induced immunity cohorts. Rates of preterm birth (born <37 weeks gestation), stillbirth (fetal demise at >20 weeks gestation), low birth weight (birth weight <2,500 g), very low birth weight (birth weight <1,500 g), and small for gestational age (bottom 10th fetal growth percentile at delivery) were reported for unvaccinated, vaccinated, and boosted cohorts. The timing of premature birth was considered by dividing it into four categories.
Term birth is defined as birth >37 weeks gestation, late preterm birth as <37 and >34 weeks gestation, moderate preterm birth as <34 and >32 weeks gestation, very preterm birth as <32 and >28 weeks gestation, and extremely preterm birth <28 weeks gestation. Fetal growth percentile was calculated using the World Health Organization (WHO) Fetal Growth Charts based on fetal sex, gestational age, and weight. 4 Median and the interquartile range (IQR) for gestational days at delivery and birth weight were reported for unvaccinated, vaccinated, and boosted cohorts. A Fisher's Exact Test using R stats (version 4.1.1) was used to evaluate the difference in rates between the unvaccinated and all other cohorts as well as between boosted and vaccinated, but not boosted cohorts for categorical variables. A Mann-U Whitney test using Python scipy (version 1.4.1) was used to evaluate the differences between continuous variables.

Supplemental Results
Efficacy of one mRNA dose, Ad26.COV2.S Johnson & Johnson, There was also a significantly lower rate of maternal COVID-19 infection in people who received one mRNA

Comparing the efficacy of mRNA-1273 Moderna and BNT162b2 Pfizer-BioNTech
There were demographic differences between the type of COVID-19 vaccine (Moderna: n=9,981; Pfizer: n=16,811) received within the vaccinated cohort (Supplemental Table 10). Patients that received BNT162b2 Pfizer-BioNTech were significantly more likely to be: Black or Asian, Non-Hispanic, older, have lower pregravid BMI, have commercial insurance, non-smoker, non-illicit drug user, lower parity, lower gravidity, and live in urban areas, areas with lower socioeconomic vulnerability, lower household composition and disability vulnerability, and higher minority status and language vulnerability. There was no difference in fetal sex, mode of delivery, or housing type and transportation vulnerability level. There was no difference in the rates of chronic diabetes, chronic hypertension, gestational diabetes, or gestational hypertension amongst people receiving either vaccine. People that received BNT162b2 Pfizer-BioNTech had a significantly lower rate of preeclampsia, but no difference in the rate of severe preeclampsia compared to people that received mRNA-1273 Moderna.
Over a five-month period of pregnancy, people vaccinated with either mRNA-1273 Moderna (n=2,373; p<0.01) or BNT162b2 Pfizer-BioNTech (n=3,746; p<0.01) lower rates of COVID-19 infection compared to unvaccinated people matched on conception date (n=6,119; Supplemental Figure 6A). There was no difference in the infection rate between the two vaccines during this same period (p=0.30). 63% of mRNA fully vaccinated patients received BNT162b2 Pfizer-BioNTech (Supplemental Figure 6B). Breakthrough infections in people that were vaccinated with mRNA-1273 Moderna tended to occur significantly more days after achieving full vaccination status than those that received BNT162b2 Pfizer-BioNTech (p<0.01) indicating that they were protected from breakthrough infections for longer on average (Supplemental Figure 6C There was also no difference in the COVID-19-related hospitalization in people that had been vaccinated with mRNA-1273 Moderna versus those vaccinated with BNT162b2 Pfizer-BioNTech (p=1.0).

Supplemental Figure 1. Flow diagram of cohort selection.
COVID-19 mRNA vaccinated (two-dose vaccination of mRNA-1273 Moderna or BNT162b2 Pfizer-BioNTech) and unvaccinated (no COVID-19 vaccines received) cohorts were selected. These were defined using Providence St. Joseph Providence electronic health records, which include imported state vaccination records.

Supplemental Figure 2. Contribution of features towards predicting vaccination status.
Impurity-based feature importance evaluating the contribution of each of the 20 demographic, comorbidity, and geographical features for machine learning models predicting vaccination status at delivery. The importance of a feature is computed as the normalized total reduction of the criterion brought by that feature. The higher the value, the more important the feature. The models evaluated are logistic regression (gold; top left panel), random forest (teal; top right panel), and gradient boosting regression (blue; bottom panel).

Supplemental Figure 3. Shapley permutation explainer of the feature contribution towards classifying vaccination status.
The contribution of all the features in the gradient boosting model towards classifying vaccination status at delivery as measured by the Shapley algorithm and reported as the SHAP value. This value is the average marginal contribution of a feature value across all permutations of features providing insight into the degree of influence of the feature on an individual's classified vaccination status at delivery. Each line represents a feature, and each dot represents a sample. The dot color represents the value of the feature for the sample, with red being a high value and blue being a low value for that feature across all samples. This evaluation was performed on 1000 patients randomly selected from the test set. SHAP=Shapley additive explanations.

Supplemental Figure 4. Propensity score matching reduces the differences in covariates between vaccinated and unvaccinated cohorts.
Propensity score matching using nearest neighbors with replacement was performed using 20 covariates. The standardized mean difference between the vaccinated and the unvaccinated (before; orange) and unvaccinated matched (after; blue) is reported. A person that has not received any COVID-19 vaccination dose regardless of brand prior to delivery and has no recorded COVID-19 infection prior to the start of the current pregnancy.

Unvaccinated Matched
An unvaccinated person that was matched on several covariates to a vaccinated person using propensity score matching.

Supplemental Table 3. Pregnancy-related conditions SNOMED definitions.
The SNOMED Concept IDs used to define common comorbidities and pregnancy related conditions.

Disease SNOMED Concept ID SNOMED Concept Name Diabetes mellitus (chronic)
199223000 Diabetes mellitus during pregnancy, childbirth and the puerperium 199225007 Diabetes mellitus during pregnancy -baby delivered 199227004 Diabetes mellitus during pregnancy -baby not yet delivered 76751001 Diabetes mellitus in mother complicating pregnancy, childbirth AND/OR puerperium 4783006 Maternal diabetes mellitus with hypoglycemia affecting fetus OR newborn 10754881000119100 Diabetes mellitus in mother complicating childbirth 199226008 Diabetes mellitus in the puerperium -baby delivered during current episode of care 199228009 Diabetes mellitus in the puerperium -baby delivered during previous episode of care 106281000119103 Pre-existing diabetes mellitus in mother complicating childbirth 609563008 Pre-existing diabetes mellitus in pregnancy 609564002 Pre-existing type 1 diabetes mellitus in pregnancy 609567009 Pre-existing type 2 diabetes mellitus in pregnancy 609566000 Pregnancy and type 1 diabetes mellitus 237627000 Pregnancy and type 2 diabetes mellitus Hypertension Disorder (chronic) 198941007 Hypertension complicating pregnancy, childbirth, and the puerperium 541000119105 Hypertension complicating pregnancy, childbirth, and the puerperium, antepartum 82771000119102 Hypertension complicating pregnancy 37618003 Chronic hypertension complicating AND/OR reason for care during pregnancy 24042004 Chronic hypertension complicating AND/OR reason for care during puerperium 698640000 Hypertension in the puerperium with pulmonary oedema 367390009 Hypertension in the obstetric context Pre-eclampsia in puerperium 46764007 Severe pre-eclampsia 95605009 Haemolysis-elevated liver enzymes-low platelet count syndrome Severe pre-eclampsia -delivered 198984008 Severe pre-eclampsia -delivered with postnatal complication 198985009 Severe pre-eclampsia -not delivered 198986005 Severe pre-eclampsia with postnatal complication 67359005 Pre-eclampsia added to pre-existing hypertension 198997005 Pre-eclampsia or eclampsia with pre-existing hypertension 198999008 Pre-eclampsia or eclampsia with pre-existing hypertension -delivered 199000005 Pre-eclampsia or eclampsia with pre-existing hypertension -delivered with postnatal complication 199002002 Pre-eclampsia or eclampsia with pre-existing hypertension -not delivered 199003007 Pre-eclampsia or eclampsia with pre-existing hypertension with postnatal complication 69909000 Eclampsia added to pre-existing hypertension Severe Preeclampsia 198983002 Severe pre-eclampsia -delivered (disorder) 198984008 Severe pre-eclampsia -delivered with postnatal complication 198985009 Severe pre-eclampsia -not delivered (disorder) 198986005 Severe pre-eclampsia with postnatal complication Supplemental  (0), or reported history of illicit drug use (1) Preterm History Previous preterm delivery: no reported history of premature delivery (0), or previously delivered prematurely prior to current pregnancy (1) Parity Number of times a person has given birth to a fetus older than 24 weeks of gestation prior to the current pregnancy: Nulliparity (0 births; 0), Low Multiparity ( Table 5. Demographic, comorbidity, and geographical features association with vaccination status at delivery. Pearson's correlation coefficient and the corresponding p-values (two-tailed) for the correlation of vaccination status at delivery with the 20 features used in the predictive models.