This study was conducted in Kaiser Permanente Northern California (KPNC), a large integrated health care delivery system that provides comprehensive medical care to over 4.5 million members. All 15 regional service centers (with 48 associated office facilities) have Obstetrics and Gynecology and Behavioral Medicine/Psychiatry Departments. KPNC members include people covered by employee-sponsored insurance plans, the insurance exchange and Medicaid. Coverage is provided for ~40% of the northern California population. KPNC members are socio-demographically similar to the population living in the geographic area served by the health plan (14). KNPC maintains comprehensive electronic health records (EHR) for all members.

This cross-sectional analysis was conducted using data from the ongoing KPNC COVID-19 Pregnancy Study (15). This analysis includes data collected from pregnant KPNC members between June 22, 2020 and April 28, 2021. Detailed information on the study design and cohort has been published previously (15). Briefly, pregnant KPNC members at least 12 weeks of gestation, 18 years of age or older and English-speaking were identified from the EHR every 2 weeks and invited to complete a brief, web-based survey.

The Institutional Review Board of KPNC approved all study procedures and participants indicated informed consent by completing the survey after reviewing the consent information in the recruitment email.

Race and ethnicity categories were ascertained from self-reported race and ethnicity in KPNC's electronic health records (EHR) and defined as Asian, Black, Hispanic, Other, and White. The Other category included racial and ethnic categories for which there were a small number of participants (Native American, Pacific Islander, Multiracial) as well as Unknown.

Prenatal mental health outcomes were captured in the survey. Depression. The 8-item, Patient Health Questionnaire (PHQ-8) depression screener has been validated for use in perinatal populations (16, 17) and was used to assess depression symptoms in the past 2 weeks. The PHQ-8 is similar to the PHQ-9 which has been validated for use across diverse racial and ethnic patients (18), with the exception that it excludes the question regarding suicide ideation. Scores range from 0 to 24 and scores of 10–24 were categorized as clinically significant depressive symptoms. Anxiety. The General Anxiety Disorder Scale (GAD-7) (19) has been validated in prenatal (20) and racially diverse populations (21, 22), and was used to measure anxiety symptoms in the past two weeks. Scores range from 0 to 21 and scores of 10–21 were categorized as clinically significant anxiety symptoms.

Several COVID-19 pandemic-related stressors were also captured in the survey. Health and healthcare stressors: Participants were asked if a healthcare provider had told them that they had or likely had COVID-19 (COVID-19 in pregnancy; y/n); if a member of their household had or probably had COVID-19 (Household member had COVID-19; y/n); whether their job put them at increased risk of COVID-19 (high-risk employment; y/n); and how distressed they were about changes to their prenatal care due to the COVID-19 pandemic (moderately/extremely vs. not at all/mildly). Economic stressors: All questions referred to the time frame of “since becoming aware of the COVID-19 pandemic.” Participants were asked if they lost their job permanently, temporarily or reduced their work hours (Lost job or reduced hours; y/n), if their spouse/partner lost their job permanently, temporarily or reduced their work hours (Partner lost job or reduced hours; y/n), and if their childcare was impacted such that they had difficulty arranging childcare or had to pay more for childcare (Childcare challenges; y/n). The validated 2-item Hunger Vital Sign screener was used to assess food insecurity (23). Food insecurity (y/n) was defined as responding often true or sometimes true (vs. never true) to either statement: (1) you worried that your food would run out before getting money to buy more or (2) the food you bought did not last and you did not have enough money to get more.

Inverse probability weights were created to account for survey non-response (20% response rate in the eligible sample; Supplementary Table 1). Inverse probability weights for survey response were calculated using predicted probabilities from a logistic regression model including variables expected to be associated with survey response (race and ethnicity category, insurance type, parity, trimester of pregnancy at the start of the COVID-19 pandemic), additional factors strongly associated with depression or anxiety symptoms (maternal age), and all interaction terms. Weights ranged from 1.56 to 59.19 in participants who completed the survey. To limit the influence from observations with extreme weights, weights were truncated at 20 for the 20 participants with weights >20 (24).

Multiple imputation by chained equations (MICE) (25) was used to address potential bias from excluding participants with missing data (complete case analysis) since data were not missing completely at random (26). We performed 100 imputations to impute missing values for people with missing data for covariates included in the regression models (14% missing parity, 1% missing insurance type). After imputation, we excluded participants with missing data for outcomes (n = 876) (27). The final analytic sample included participants with non-missing outcome information (n = 10,930).

The Ward et al. (13) framework for disparity investigation outlines three steps for identifying contributors to racial health disparities: (1) assess if there are differences in outcome prevalence (e.g., prenatal depression and anxiety) between racial and ethnic groups; (2) assess if there are differences in exposure prevalence (e.g., COVID-19-related health/healthcare and economic stressors) between racial and ethnic groups, and (3) evaluate whether the relationship between the exposure (COVID-19-related stressors) and outcome (prenatal depression and anxiety) differs between racial and ethnic groups. Conclusions about COVID-19-related stressors that contributed to racial and ethnic prenatal mental health disparities were based on the results from the analyses in accordance with the Ward et al. guidelines. Below we outline the methods used for each of these steps.

We first assessed racial and ethnic differences in the prevalence of prenatal depression and anxiety. According to the framework, establishment of a difference in the prevalence of the outcome is fundamental to the definition of a racial and ethnic health disparity. If there is no difference in the prevalence, then a disparity does not exist. We used weighted modified Poisson regression models (28) to estimate prevalence ratios and 95% confidence intervals for prenatal depression and anxiety associated with racial and ethnic group, with and without adjusting for potential confounders (maternal age, parity, insurance type, calendar month of survey completion, and trimester at survey completion) for each imputed dataset. Results were combined using Rubin's rules (29). The final point estimate was calculated as the average of the point estimate distribution across regression model results from each imputed dataset; the variance was calculated using both within- and between-imputation variances.

Second, we evaluated racial and ethnic differences in prevalence of COVID-19-related health/healthcare and economic stressors. Weighted percentages and chi square statistics were calculated for each COVID-19-related stressor by racial and ethnic group for each imputed dataset and combined across imputed datasets (30).

Third, we assessed whether the relationship between COVID-19-related stressors and prenatal depression and anxiety differed by racial and ethnic group. An association between the exposure and the outcome in the disadvantaged group was in some cases sufficient to conclude the stressor contributed to the disparity, depending on the prevalence of the exposure in the racial and ethnic groups (13). We used weighted modified Poisson regression models (28) to estimate prevalence ratios and 95% confidence intervals for prenatal depression and anxiety associated with each COVID-19-related stressor stratified by racial and ethnic category. Models were mutually adjusted for all COVID-19-related health/healthcare and economic stressors and potential confounders (maternal age, parity, insurance type, calendar month of survey completion, and trimester at survey completion). To test statistical differences in associations by racial and ethnic groups, we included a cross-product term between race and ethnicity category and each stressor in separate weighted modified Poisson regression models (28) for each stressor and prenatal depression and anxiety, adjusted for all COVID-19-related health/healthcare and economic stressors and potential confounders mentioned above. Tests for interaction generally have less power to assess statistical significance and thus a p < 0.10 is accepted as the cutoff for statistical significance for interaction terms (31). Models were run for each imputed dataset. Results were combined using Rubin's rules (29).

Analyses were conducted in SAS version 9.4 (SAS Institute INC., Cary, NC, USA) and R (version 4.1.2) (32).

Overall, there were significant racial and ethnic differences in the prevalence of all health/healthcare and economic stressors (p < 0.001 for both). For the purpose of identifying COVID-19-related stressors that contribute to racial and ethnic disparities in prenatal mental health, we focus the text of this section on the racial and ethnic groups (Black and Hispanic) with a higher prevalence of prenatal depression or anxiety compared to White participants.

Similar to above, we focus the text of this section on racial and ethnic groups (Black and Hispanic) with a higher prevalence of prenatal depression and anxiety compared the White group.

Among Black participants, distress due to changes in prenatal care was the only stressor statistically significantly associated with prenatal depression (aPR: 2.27, 95% CI: 1.41, 3.64) (Table 3A). This association was similar to that in White participants (aPR: 2.41, 95% CI: 1.98, 2.92). Among Hispanic participants, high-risk employment (aPR: 1.32, 95% CI: 1.01, 1.71), distress due to changes in prenatal care (aPR: 2.76, 95% CI: 2.12, 3.58), and food insecurity (aPR: 1.70, 95% CI: 1.31, 2.19) were statistically significantly associated with prenatal depression. Associations of high-risk employment and distress due to prenatal care changes with prenatal depression were similar to those among White participants (high-risk employment aPR: 1.34, 95% CI: 1.09, 1.64; distress due to changes in prenatal care aPR: 2.41, 95% CI: 1.98, 2.92). The association of food insecurity with prenatal depression in White participants was stronger than that in Hispanic participants (aPR: 2.35, 95% CI: 1.83, 3.00).

Among Black participants, distress due to changes in prenatal care was the only stressor statistically significantly associated with prenatal anxiety (aPR: 3.00, 95% CI: 1.85, 4.84) (Table 3B). This association was similar to that in White participants (aPR: 2.89, 95% CI: 2.38, 3.50). Among Hispanic participants, high-risk employment (aPR: 1.56, 95% CI: 1.20, 2.02), distress due to changes in prenatal care (aPR: 2.82, 95% CI: 2.15, 3.71), and food insecurity (aPR: 1.66, 95% CI: 1.28, 2.15) were statistically significantly associated with prenatal anxiety. Associations of high-risk employment, distress due to prenatal care changes and food insecurity with prenatal anxiety were similar to those among White participants (high-risk employment aPR: 1.32, 95% CI: 1.08, 1.62; distress due to changes in prenatal care aPR: 2.89, 95% CI: 2.38, 3.50; food insecurity aPR: 1.84, 95% CI: 1.43, 2.36).

Applying the Ward et al. framework (13), one scenario where a stressor contributes to a disparity in a mental health outcome is when: (1) there is a disparity in the mental health outcome, (2) the stressor is more prevalent in the more disadvantaged racial and ethnic group compared to the White group, and (3) the stressor is associated with the mental health outcome in the more disadvantaged racial and ethnic group (regardless of whether that relationship is different from that in the White group). Thus, our findings suggest distress due to prenatal care changes contributed to the observed disparities in prenatal depression and anxiety for Black and Hispanic people and food insecurity additionally contributed to the observed disparities in prenatal depression and anxiety for Hispanic people. We note that if we had relied solely on the results from the interaction model, both distress due to changes to prenatal care and food insecurity would not have been identified as contributors to these disparities in prenatal mental health.

Although high-risk employment was also associated with greater prevalence of prenatal depression and anxiety in Hispanic people, the prevalence of high-risk employment and its association with prenatal depression and anxiety were similar in Hispanic and White people (Table 4). Thus, this indicates that high-risk employment is not a contributor to the observed disparities in prenatal depression and anxiety for Hispanic people. However, while it may not contribute to the noted disparity, employment with high-risk of exposure to COVID-19 has been associated with psychological distress in other studies in pregnant people (5) and other non-pregnant populations (7, 40). Fatigue, health worries, and fear have all been reported by people with high-risk employment even previous infectious disease outbreaks (41). Consideration of high-risk employment as a risk factor for psychological distress in pregnancy, especially during infectious disease outbreaks, should be considered in prenatal depression and anxiety screening strategies.

The COVID-19 pandemic forced rapid implementation of hybrid models of care including both in-person and virtual prenatal care visits. Previous research has documented a significant relationship between COVID-19-related changes in prenatal care and greater psychological distress during pregnancy (6–9). In our study, Black and Hispanic people in particular experienced distress due to the changes in prenatal care which contributed to the higher burden of poor prenatal mental health in these groups. Previous research has found some people of color report negative interactions with the healthcare system during pregnancy, including experiences of racism, unmet information needs, and stressful interactions with all levels of staff (42–44). Thus, there is a need for better communication and careful listening during prenatal visits in pregnant people of color. During the COVID-19 pandemic, Black people reported worrying more than White people about having a good birthing experience and receiving good prenatal care (12), suggesting that COVID-19-related changes in prenatal care may have created challenges for patient-provider communication that further exacerbated existing disparities in quality of prenatal care between Black and White people. As healthcare systems continue to make changes to prenatal care and evaluate which care-delivery modifications to implement as standard prenatal care, providers and healthcare system leaders must remain focused on quality of patient-provider communication and must work with Black and Hispanic communities to identify ways to address stressful aspects of changes in prenatal care.

Similar to previous research during the pandemic in pregnant (12) and non-pregnant populations (45), Black and Hispanic people experienced the highest prevalence of food insecurity. Food insecurity is strongly associated with prenatal depression (46, 47) as well as long-term maternal and child health. (48) Our study demonstrated food insecurity contributed to disparities in prenatal depression and anxiety for Hispanic people. However, the association of food insecurity with prenatal depression did not reach statistical significance among Black people, which may be due to the small sample size. Regardless, these findings highlight the high burden of food insecurity and its contributions to poor maternal mental health during pregnancy in populations of color.

This is among the first studies that we are aware of that has evaluated the contribution of food insecurity to racial and ethnic disparities in prenatal mental health. Research prior to the COVID-19 pandemic has documented income and employment status as a contributor to these disparities (10). Yet in our study job loss or reduced work hours by the participant or their partner was not a contributor. It is possible that food insecurity is a greater contributor to these disparities and an important area for future research especially given the potential policy and clinical implications if so. Addressing food insecurity in pregnant people may have a significant impact on both their nutritional and mental health. The high prevalence of food insecurity for Black and Hispanic people during pregnancy highlights the importance of implementing universal screening for food insecurity in prenatal care and systems to connect people identified as food insecure with available federal and community resources.

Although we did not identify a disparity in mental health outcomes for Asian people compared to White people, several COVID-19-related stressors, including high-risk employment, distress due to prenatal care changes, and food insecurity, were strongly related to prenatal depression and anxiety in Asian people. The prevalence of COVID-19-related stressors tended to be lowest among Asian people, which may explain why prevalence of depression and anxiety was also lowest in this group, even though there were strong associations of COVID-19-related stressors and mental health in Asian people. Future research should look at potential protective factors against poor prenatal mental health in Asian people, given the observed lower prevalence of depression and anxiety.