Introduction

On March 17th, 2020, the American Society of Reproductive Medicine (ASRM) published a bulletin recommending an indefinite suspension of all new, non-urgent fertility treatments throughout the U.S. due to the COVID-19 pandemic (ASRM, 2020a). Such recommendations were consistent with broader calls in the medical community given the impact of COVID-19 on health systems. However, on April 24th of the same year, the ASRM issued new directions to gradually resume treatments (ASRM, 2020b). Clinic responses to this guidance were heterogeneous and inconsistent (Mehr et al., 2021). At least one study documented declines in December 2020 in assisted reproductive technology (ART) births, which is the earliest effects of clinic closures could be expected given the average length of human gestation (Lisonkova et al., 2022). Overall, prior to COVID-19, the percentage of births due to MAR has been increasing annually; analyses of the U.S. Vital Statistics System show an increase from 1.34% of births in 2011 to 1.97% of births in 2019 (Tierney & Guzzo, 2023) and the more accurate counts of ART births from the National ART Surveillance System show ART comprised 1.5% of births in 2011 (Sunderam et al., 2014) and 2.24% of births in 2019 (authors’ calculation based upon: Martin et al., 2021; Sunderam et al., 2021).

To date, however, no prior work has evaluated the changes in medically assisted reproduction (MAR) births, which is a broad term used to describe births conceived using in vitro fertilization (IVF), use of ovulation medications, and/or use of intrauterine insemination, after December 2020. Based on prior research on overall fertility and treatment utilization trends, it is likely that MAR births will have rebounded from this initial decline. For example, a recent report from the Centers for Disease Control and Prevention using 2021 birth data showed increases in overall births in the U.S. since 2020, with variations in trends by racial/ethnic groups (Hamilton et al., 2021). Similar declines and rebounds in overall fertility have also been documented in other work (Kearney & Levine, 2022). In addition, in a study of commercially insured U.S. women, Zhou and colleagues (2021) documented a decline in infertility treatment utilization between March and April 2020, followed by an increase through November 2020 that surpassed pre-pandemic levels. Thus, it is likely that MAR birth rate declines were short lived but may have varied by sociodemographic groups.

In this study, we aim to provide an assessment of the impacts of the COVID-19 outbreak on MAR birth rates in the U.S. using publicly available, population-level data. Additionally, due to the age-patterning and the well-documented disparities in U.S. ART births (Ethics Committee of the American Society for Reproductive Medicine, 2021; Fujimoto et al., 2010; Jain, 2006; Lazzari & Tierney, 2023; Tierney & Cai, 2019), the analyses also assess whether the findings vary by age, race/ethnicity, or educational attainment.

Materials and Methods

National Vital Statistics System (NVSS) Birth Certificate Data

The NVSS birth certificate data are collected by U.S. States and compiled by the National Center for Health Statistics. The 2003 revision to the U.S. Birth Certificate includes information on the use of medical assistance to become pregnant, including ART and fertility enhancing medications. The U.S. birth certificate also collects data on single year of age, race/ethnicity of the mother, and educational attainment of the mother. The present study used data from January 2017 to December 2021 and defined MAR births as any birth identified in the NVSS data as using ART or any other infertility treatment in Box 41. The analyses included only live births.

Current Population Survey (CPS)

The CPS is a publicly available, nationally representative monthly survey of U.S. households administered by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics (Flood et al., 2020). The CPS dataset was used to estimate the population of U.S. women by age, race/ethnicity, and educational attainment.

Analysis Plan

In the descriptive analyses, we calculated the percentage of births due to MAR overall by month and year and by five-year age groups, race/ethnicity, and educational attainment of the mother. Next, we employed three sets of negative-binomial regression analyses using size of the female population to account for exposure to risk. First, we evaluated whether there were any changes in the incident rates of MAR births in December 2020 relative to all months and years between January 2017 and December 2021 using bivariate and multivariable negative-binomial regression. Controls in the multivariable models included age, race/ethnicity, and educational attainment. Based on this first set of results, we created a three-category variable capturing distinct time periods: the before COVID-19 impacts period (October 2020 and November 2020), the during period (December 2020 and January 2021), and the after period (February 2021 and March 2021). We used this period variable in our second set of analyses using negative-binomial regression models with interaction terms to assess whether the changes in the MAR incident rates varied by age, race/ethnicity, or educational attainment. Finally, we assessed whether there were any changes in the yearly MAR birth incident rates using bivariate and multivariable negative-binomial regression. All analyses were conducted in Stata 17.0 (StataCorp 2021).

Data Restrictions

States that did not report MAR births were excluded from our analyses. However, reporting of MAR has varied over the study period and individual states are not identifiable in the public-use NVSS dataset. To reduce the impact of these changes on the results, we used an exposure variable for population size for included states and used data from 2017 to 2021 because there was less change in state reporting during this period. Specifically, the 2021 dataset excluded South Carolina, while the 2017–2020 datasets excluded South Carolina and Tennessee (Centers for Disease Control & Prevention, 2021). Additionally, MAR births and population counts for implausible age and education combinations were removed (i.e., an under 15 year old with more than a 4-year degree). Due to the low MAR utilization rates among younger women, the negative-binomial regressions excluded women under the age of 20. All data used for this study are publicly available (Flood et al., 2020; National Center for Health Statistics and The National Bureau of Economic Research 2022).

Results

Descriptive Results

Figure 1 shows a cyclical pattern with yearly declines in the percentage of births due to MAR in each September. However, in 2020, there was also a decline in the percentage of births due to MAR in December. Figure 2 shows changes in the percentage of births due to MAR by age (Panel A and B), race/ethnicity (Panel C), and educational attainment (Panel D). Except for younger ages where MAR births were less common and among women with less than a 4-year degree, large declines in the percent of births due to MAR between November 2020 and January 2021 were evident.

Fig. 1
figure 1

Percentage of Live Births due to Medically Assisted Reproduction (MAR) by Month and Year from January 2017 to December 2021

Full size image
Fig. 2
figure 2

Percentage of Live Births due to Medically Assisted Reproduction (MAR) by Month and Year from January 2017–December 2021 and by Five Year Age Groups (Panels A and B), Race/Ethnicity (Panel C), and Educational Attainment (Panel D). The y-axis scale changes across panels to best show the data. NH non-Hispanic, NHOPI Native Hawaiian or Other Pacific Islander

Full size image

MAR Birth Incidence Rates from January 2017 to December 2021

In both the bivariate and multivariable models, all months and years included in the analyses had higher incidence rates of MAR births than December 2020 with the exception of January 2021, where the incident rate ratios did not differ (IRR = 1.05, 95% CI: 0.72–1.53; adjusted IRR = 0.97, 95% CI: 0.77–1.22) (Table 1). Net of controls, in other months and years, the MAR birth incident rates were between 1.61 (95% CI: 1.28–2.01) and 2.28 (95% CI: 1.83–2.84) times the December 2020 MAR birth incident rates. In addition, net of controls, in February 2021, the MAR birth incident rate was 1.54 times the December 2020 incident rate (95% CI: 1.23–1.92), by March 2021, the MAR birth incident rate was 1.95 times the December 2020 incident rate (95% CI: 1.57–2.44), and by April 2021, the MAR birth incident rate was 2.07 times the December 2020 incident rate (95% CI: 1.66–2.59).

Table 1 Bivariate and Multivariable Negative Binomial Regressions of the Association between Births due to Medically Assisted Reproduction (MAR) and Month and Year between January 2017 and December 2021
Full size table

MAR Birth Incidence Rate Changes by Sociodemographic Group and COVID-19-Impacted Periods

Based on the two sets of results above, we created a three-category variable to capture the “before” (October 2020 and November 2020), “during” (December 2020 and January 2021), and “after” (February 2021 and March 2021) periods of decline and rebound of MAR live birth rates for our interactive models.

In the interactive analyses, there were no patterns suggesting the decline or rebound periods differed by race/ethnicity or age group (Supplemental table 1). By contrast, a pattern was observed for educational attainment (Fig. 3). The predicted number of MAR births for the before, during, and after periods by educational group were calculated to ease interpretation of this patterning. For women with less than a 4-year degree, the predicted number of MAR births were 45.4 (95% CI: 35.7–55.1), 38.0 (95% CI: 24.2–38.2), and 44.9 (95% CI: 35.2–54.4) for the before, during, and after periods in comparison with 73.8 (95% CI: 58.1–89.4), 38.5 (95% CI: 30.0–47.0), and 73.1 (95% CI: 57.5–88.7) MAR births for women with a 4-year degree and 161.7 (95% CI: 125.3–198.0), 62.1 (95% CI: 47.6–76.7), and 145.4 (95% CI: 113.2–177.5) MAR births for women with more than a 4-year degree in the before, during, and after periods, respectively. These predicted values show that the decline and rebound in MAR births for women with more than a 4-year degree was steeper than for women with lower levels of education. These predicted counts also show that all educational groups returned to their pre-decline levels by the “after” period.

Fig. 3
figure 3

Predicted MAR Births with 95% Confidence Intervals by Period and Educational Attainment. Analyses are age adjusted, but exclude women 19 and under. MAR Medically Assisted Reproduction

Full size image

MAR Birth Incidence Rates by Year

Net of the controls, we observed that the incident rate of MAR births in 2021 did not differ from 2017, 2018, 2019, or 2020 (2017 IRR = 0.95, 95% CI: 0.89–1.01; 2018 IRR = 0.96, 95% CI: 0.90–1.02; 2019 IRR = 0.95, 95% CI: 0.90–1.02; 2020 IRR = 0.95, 95% CI: 0.89–1.01) (Table 2).

Table 2 Bivariate and Multivariable Negative Binomial Regression of the Incident Rate Ratio of Births due to Medically Assisted Reproduction (MAR) by Year 2017–2021
Full size table

Discussion

Using publicly available population-level data, this study analyzed MAR births in the U.S. during the early stages of the COVID-19 pandemic in 2020–2021. Consistent with prior research that used the NVSS records (Lisonkova et al., 2022), we found a decline in MAR births in December 2020. However, our results also demonstrated that this decline continued into January 2021 and varied by socioeconomic strata. Specifically, the results showed steeper declines in MAR births among women with more than a 4-year degree. This patterning may be the result of more highly educated women suspending treatment in accordance with general public health recommendations, including advice to remain home and to limit health care utilization for non-life-threatening conditions, as observed in other domains of health early in the pandemic (Pampel et al., 2010). Alternatively, the educational differences in declines in MAR births may be due to local restrictions where women with higher levels of education live or other contextual factors such as geographic location or COVID-19 prevalence rates, which are not available in the dataset used.

The results also showed a quick rebound in MAR births in the United States. Indeed, MAR births rebounded as early as February 2021 and pre-pandemic levels were reached by April 2021. While these rebounds did not vary by age or race/ethnicity, they did vary by educational attainment. Specifically, the rebounds were steeper among women with more than a 4-year degree.

These findings have implications for research, practice, and policy. First, for research, the findings demonstrate that the pandemic has led to substantial fluctuations in MAR births in the United States. Importantly, this paper shows that the temporary decline in MAR births documented in previous research (Lisonkova et al., 2022) was ultimately resolved. Specifically, despite the profound impact of COVID-19 on people’s lives and relationships, the yearly incidence rates of MAR births in 2020 and 2021 were not different from previous years, suggesting that individuals were largely prepared to return to treatment when clinics reopened. However, further research is needed to clarify the individual and contextual circumstances that shaped these descriptive trends and the effects of these short-term declines and rebounds on demographic outcomes such as involuntary childlessness. In addition, this study underscores the importance of conducting detailed monthly data analysis, which allowed us to shed light on challenges to care capacity and provide insights into behavior related to MAR use.

Second, for practice, the initial declines in MAR birth rates likely resulted in a short-term increase in demand for services as women who delayed care resumed treatment. While clinics may have been equipped to handle this increase, postponement of births due to the pandemic, or any other reason, may result in more potential patients seeking care who otherwise might not have needed services (Leridon, 2004; Leridon & Slama, 2008; Morgan & Taylor, 2006). As a result, practices may see an increase in older patients or those who have had longer duration of infertility before seeking care. Thus, changes to practice guidelines may be advisable to help patients who delayed pregnancy or infertility care during the pandemic to improve their odds of becoming pregnant and having the families they desire.

Finally, for policy, despite being short lived, the decline in MAR birth rates observed provide a reminder that policies restricting MAR may have effects on family formation because of the time-sensitive nature of infertility and the common use of infertility treatments to have first births (Barbuscia et al., 2019; Lazzari et al., 2021; Tierney, 2022). In other words, even short-term impacts on access to care may result in some foregone childbearing, which should be considered if restrictions on infertility care are needed in the future.

Limitations

This paper has two limitations. First, although this study adjusts for demographic risk, these data do not allow us to conduct analyses on the potential mechanisms of these declines nor how they vary across geography, though such factors have been associated with fertility declines and rebounds throughout the COVID-19 pandemic (Kearney & Levine, 2022). Second, there is well-documented underreporting of ART births in the NVSS data (Moaddab et al., 2016; Thoma et al., 2014; Tierney & Cai, 2019). Unfortunately, there are no other publicly available, population-level estimates of MAR births in the United States that could be used for these demographic analyses. However, prior work has shown that the underreporting is not patterned by maternal characteristics (Thoma et al., 2014). Further, there is no reason to assume that the level of underreporting was higher in the months where declines were observed. Still, the results from this study cannot estimate the true proportion of births due to MAR. Instead, the results provide a unique insight into the patterning of MAR birth rates during this phase of the COVID-19 pandemic. Future research should be pursued by those with access to restricted data to provide population-level estimates of MAR births during COVID-19 overall and across sociodemographic groups.

Conclusion

The present study assessed the impacts of COVID-19 restrictions on MAR births using population-level data. Declines were seen in MAR birth rates in December 2020 and January 2021 that differed from prior monthly trends. The reopening of clinics mitigated the dramatic decline in MAR birth rates during this period, with MAR births quickly rebounding to pre-pandemic levels. Despite the relatively short duration of closures, for some patients, MAR births may not be recovered. It is important for future research to further quantify the effects of these declines and for both medical practices and policy to adapt to changing needs and demands of patients to ensure equitable access to infertility treatment, especially in the wake of pandemic related pregnancy postponement.