Social determinants of health associated with COVID-19 severity during pregnancy: a multinational cohort study (in the International Registry of Coronavirus Exposure in Pregnancy)

Pregnant and recently pregnant participants enrolled in the International Registry of Coronavirus Exposure in Pregnancy (IRCEP, ircep.pregistry.com) by completing several online survey modules. IRCEP is a multinational, observational cohort study of adult participants who were tested for SARS-CoV-2, regardless of results, or had clinical diagnosis of COVID-19 during pregnancy. Questions concerning pre-pregnancy sociodemographic factors, maternal clinical characteristics, health outcomes, pregnancy/birth outcomes, and infant characteristics were asked. IRCEP was registered with the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) [EUPAS37360] and clinicaltrials.gov [NCT04366986], with additional design and cohort information published previously [15]. The Institutional Review Board of the Harvard T.H. Chan School of Public Health approved this study (IRB20-0622).

Participants with symptomatic COVID-19 any time during pregnancy who completed the demographics panel were included in our analysis (enrollment: June 24, 2020-March 18, 2021).

A focus group of pregnant and recently pregnant women in the United States participated in the review of the study surveys prior to the study launch. They evaluated the questions, the flow of the surveys, and the overall design of the study and provided important feedback that was subsequently used to improve both the content and the format of the web-based data collection tool. In addition, during the conduct of the study, participants were asked to provide comments about each survey module and their responses were carefully considered by the study team for possible use in the study.

Participants self-reported COVID-19 symptoms, clinical diagnoses, and SARS-CoV-2 laboratory test results and were asked to upload de-identified medical records. For primary outcome, a formal COVID-19 test was not required, and a small proportion of the cases (6%) did not have test results available. This was more common at the beginning of the pandemic in some populations and/or countries; therefore, sensitivity analysis was utilized to assess whether availability of a positive test affected the results.

COVID-19 severity: mild, moderate, and severe was the primary outcome. Severity levels were determined based on symptoms and healthcare utilization using an algorithm similar to definitions provided by the Centers for Disease Control and Prevention (CDC) without clinical assessments (Additional file 1: Appendix 1). Severe COVID-19 was defined as admission to an intensive care unit (ICU); use of respiratory assistance; or being hospitalized, and experiencing organ failure, acute respiratory distress syndrome (ARDS), bluish lips/face, difficulty breathing, chest pain, or pneumonia. Participants with moderate COVID-19 had at least one of the following: abnormal chest X-ray/CT scan; ARDS, difficulty breathing, pneumonia, or visited a healthcare facility and experiencing either chest pain or bluish lips/face. Participants with moderate COVID-19 could have also visited a healthcare facility and experienced at least two of the following: high fever (≥ 38.0 °C or ≥ 100.4°F); sore throat/cough/runny nose; fatigue; headache; myalgia; loss of smell/taste; or gastrointestinal symptoms (i.e., abdominal pain, diarrhea, nausea, or vomiting). Lastly, mild COVID-19 was defined as experiencing at least one symptom and not meeting the definition of moderate or severe COVID-19.

Social, demographic, and economic conditions that might influence health status included in our analysis were household size, employment at first COVID-19 symptoms, economic status, highest educational level, and World Bank income classification based on the participant’s country of residence [16]. Healthcare access was assessed by health insurance status and if participant had at least one healthcare visit during their current or recent pregnancy.

Participants reported whether they elected to follow 17 different behaviors that could potentially lower the risk of COVID-19, including wearing a mask, avoiding crowded places, working from home, and handwashing, two weeks prior to diagnosis. These were included in our analysis as some were expected to be related to social and/or employment needs [17, 18].

For completeness, pre-pregnancy self-reported health, pre-pregnancy body mass index (BMI), and medical comorbidities such as asthma, high blood pressure, and any other cardiovascular condition (e.g., heart failure, ventricular septal defects, arrhythmias) were included. Information about pregnancy and pregnancy history were also included, as they could potentially affect COVID-19 severity, including, whether pregnancy was considered high-risk, whether it was first pregnancy, and the trimester of COVID-19 exposure (reported at enrollment).

Continuous variables are described as means and standard deviations (SD) and categorical variables are reported as frequencies and percentages. Stratified analyses were conducted by enrollment timing due to potential differences in COVID-19 severity between groups (pregnant vs. recently pregnant). Recently pregnant was defined as enrolling within 180 days following pregnancy. Prior to analysis, a chi-square test was conducted to examine whether pregnancy status at enrollment (pregnant vs. recently pregnant) was associated with COVID-19 severity. Participants in the recently pregnant group reported a higher frequency of severe COVID-19.

To determine the crude association between social and demographic characteristics and COVID-19 severity, univariate mixed-effects ordinal logistic regression models in the form of a cumulative link mixed effect model were performed. A random effect for country was added to the model to account for global differences during the pandemic. Odds ratios (OR), Wald-type 95% confidence intervals (95% CI), and statistical significance were reported for all parameters. To adjust for multiple testing given the number of social and demographic characteristics examined, Benjamini–Hochberg corrections were used to produce adjusted p-values. Ordinal logistic regression models assumed proportional odds, where odds ratios across all COVID-19 severity categories were equivalent (e.g., odds of having moderate vs mild COVID-19 was the same as the odds of having severe vs. moderate COVID-19). Model diagnostics were performed to test against the assumptions, including underlying proportional odds, or parallel assumptions, that allows for one set of coefficients for the COVID-19 severity levels. The proportional odds model assumption was evaluated using the Brant test [19]. All data analyses were performed using R [20].

In this study of 4,231 pregnant and recently pregnant participants with symptomatic COVID-19 across 41 countries, we demonstrated that social and demographic characteristics are associated with COVID-19 severity during pregnancy. Younger age, lower economic class (at individual and country level), lower educational attainment, and employment setting (food service vs. unemployed) were associated with increased odds of more severe COVID-19. In addition, we found several co-morbidities associated with COVID-19 severity: having a prior pregnancy vs. first pregnancy, pre-pregnancy higher BMI, asthma, and high-risk pregnancy all increase the odds of more severe COVID-19.

These results are similar to other studies in the non-pregnant population which have found a disproportionate burden of SARS-CoV-2 infection among persons living in poverty and those with lower educational attainment [21, 22]. In the pregnant population, various studies have also found a connection between socioeconomic disparities, and maternal morbidity and mortality of mothers infected with COVID-19 [3, 4, 23, 24]. Two studies noted that younger maternal age has been associated with an increased risk of COVID-19 infection, but did not find a relationship between age and COVID-19 severity among infected pregnant participants [6, 11]. Our study found that younger maternal age was associated with more severe COVID-19. Younger pregnant participants may delay seeking medical care assuming they are at a lower risk of severe COVID-19 or other diseases [2]. Another study of pregnant patients with COVID-19 infection who delivered in United States hospitals found that those with more severe illness had older mean age, higher median body mass index, and pre-existing medical comorbidities [25]. In that study, universal screening for infection was not routine across all clinical sites, and asymptomatic positive test results from screening were included in the COVID-19 severity outcome when available. Differences in findings around maternal age and COVID-19 severity could be a function of differences in sampling given IRCEP was an online global study compared to the single country clinical setting that included asymptomatic test results.

Similar to data from Sakowicz [11], our study found that prior pregnancy was also associated with increased risk of severity. We did not assess whether the age or number of living children at home is a risk factor for SARS-CoV-2 infection, or whether the number of children at home is a surrogate marker for other social factors such as crowdedness or increased exposures outside of the home. Children as well as other adults at home may partially explain the severity of infection among pregnant people.

While not the focus of our analysis, our results on comorbidities are congruent with data compiled in a meta-analysis reporting higher BMI (> 30), asthma, and any other pre-existing maternal comorbidity are factors associated with severe COVID-19 disease in pregnancy [26].

Epidemiological studies have reported a high risk of SARS-CoV-2 infection among food workers as a result of contact with one another for extended periods, including at work and shared housing and transportation [27]. The significant association we found in our study of working in the food industry with higher odds of COVID-19 severity is supported by a study by Mutambudzi et al. [28]. After adjusting for age, sex, ethnicity, and country of birth, people working in the food industry still had higher odds of COVID-19 severity [28]. Women make up a higher proportion than men in the food industry. For example, in the United States women comprise approximately 60% of food service workers [29]. We do not have further information on the specific type of employment; i.e., food processing, manufacturing, or agriculture, which limits our ability to contextualize the nuances of the risk. However, regardless of the type of setting, factors that contribute to high risk of infection could be further aggravated by the baseline social characteristics analyzed in this study resulting in a higher degree of COVID-19 severity. Another explanation for the higher odds of COVID-19 severity among people working in the food service industry is that food service workers in our study had lower socioeconomic factors and were younger than other groups. In our study, two-thirds of participants employed in food services reported being either poor or in the lower-middle class versus about one-third to half of participants reporting other occupations. Women employed in food services had a lower median age than other groups. When adjusting for employment status, economic class, and age in the model simultaneously, the adjusted odds ratios of these variables on the outcome of COVID-19 severity were attenuated but remained significant (not shown). No interactions were examined.

Individual lower income and lower educational attainment were associated with COVID-19 severity in Prasannan et al. [6]. In our study, we also found an association between lower country income (World Bank Classification) and COVID-19 severity in pregnant and recently pregnant participants. Those coexisting factors may play a large role in the association we observed. There is limited data from low/middle-income countries (LMICs) on the impact of SARS-COV-2 infection in pregnancy in our study. Pregnant people in LMICs are disproportionately affected by poor COVID-19 outcomes given co-morbidities and more limited healthcare access [4, 30].

Lastly, all the factors highlighted in this analysis (age, education, income, comorbidities) contribute in one way or another to perpetuating the cycle of disease in populations already at risk. COVID-19 has had a profound impact on our society and has exacerbated the maternal health crisis existing in many areas of the world. Pregnant individuals, already facing disadvantaged situations as a result of unfavorable sociodemographic conditions, are placed at higher risk of COVID-19 severity [31]. Recognizing the disparities in pregnant individuals with SARS-CoV-2 infection is imperative to provide an adequate and timely response that decreases additional burden at an individual, societal, and healthcare system level.

A strength of our analysis is its size and breadth, contributing important information on the relationships between social and behavioral characteristics with COVID-19 severity in a multinational registry with 4,231 participants in 41 countries. To our knowledge, IRCEP is the largest COVID-19 pregnancy registry. Although our sample is large, it may have limited generalizability to the general population of pregnant people, in that enrollment was conducted on line and require access to the Internet via computer/mobile phone.

Selection into the study population may be associated with the severity of COVID-19, and enrollment was allowed following birth or a pregnancy loss. We stratified by timing of enrollment (pregnant vs. recently pregnant) given differences in COVID-19 severity and disproportionate numbers of severe COVID-19 in the recently pregnant group. The recently pregnant group is composed of participants who enrolled retrospectively after pregnancy, including those who safely delivered their baby or experienced a pregnancy loss. The retrospective enrollment could have been related to SDHs and/or COVID-19 severity, which could explain any differences across analyses by enrollment timing. Also, those with severe COVID-19 late in pregnancy may not have had the ability to participate in the study until they recovered and, by then, they had delivered. In addition, given that the majority of the sample were those 25 to 34 years old (66%), selection bias may have affected the association between younger age and more severe COVID-19.

Our planned analysis focused on demographic and social characteristics. All of the data was self-reported, and the study had a potential for information bias, including COVID-19 severity and participant COVID-19 risk minimizing behaviors. Participant-reported behaviors to lower risk of SARS-CoV-2 infection seemed to increase the odds of more severe COVID-19, as the analysis conditions on having a symptomatic infection. This finding was thought to be related to collider bias [32], where the relationship between preventive behaviors and COVID-19 severity could have been distorted by conditioning on a positive test or symptoms. Our analysis examining the relationship between SDHs and COVID-19 severity was limited to positive, symptomatic COVID-19 diagnosis to minimize collider bias. As testing was not available to everyone, the inclusion of individuals with negative tests and asymptomatic COVID-19 tests would have been related to risk factors for COVID-19 infection and social factors of interest, including economics and occupation [15, 33]. Lastly, participants may have provided more socially acceptable responses versus admitting to not adhering to risk-minimization behaviors (Additional file 1: Appendix 4).

There were some limitations on the inclusion of SDHs. About 25% of patients eligible for this analysis were excluded due to missing social and demographic characteristics, as the demographic module was optional (Fig. 1). While self-reported socioeconomic status may be somewhat subjective, both lower self-reported wealth and lower World Bank country-level income were associated with increased severity of COVID-19 infection in this multinational study. The IRCEP registry did not capture detailed information on race or ethnicity on a global scale due to challenges in capturing this meaningfully across all the included countries. Race and ethnicity are correlated with other SDHs [34, 35], and the effect of race and ethnicity on COVID-19 severity may be explored in future work relating to specific countries.

In examining the relationship between social and demographic characteristics with COVID-19 severity at enrollment, we selected characteristics prior to pregnancy and prior to COVID-19 diagnosis where possible. However, temporality may not always be precisely known or available (e.g., employment questions during the pandemic in relation to COVID-19 infection and pregnancy). Finally, COVID-19 testing was not available in all regions, and hence SARS-CoV-2 infection status is subject to misclassification. Restricting to those with COVID-19 testing (94% of sample) did not appreciably modify the results. Despite the potential for misclassification, our findings are robust given the requirement for participants in this analysis to exhibit COVID-19 symptoms.

The social and demographic characteristics examined in this study are expected to be correlated with one another and with underlying risk factors for COVID-19 severity. Fitting an adjusted model combining social and demographic characteristics was limited to including occupation, economic class, and maternal age when exploring the increased risk of more severe infection among food service workers given that our primary interest was on individual characteristics. In addition, sample size limitations in the severe COVID-19 category (~ 100 participants) limited the number of characteristics that could be adjusted in multivariable modeling.

Despite the limitations, this study provides important insights into the relationship between demographic factors and SDHs at increased odds of more severe COVID-19 during pregnancy prior to introduction of COVID-19 vaccines.