Study population

AWHS is an ongoing longitudinal study of women’s health. The study launched recruitment in November 2019, as previously described. Eligibility for enrollment included women who had menstruated at least once, had a compatible iPhone iOS, downloaded the Research app, lived in the United States, were at least 18 years old, communicated through English, were the exclusive user of an iPhone or iCloud account, and provided informed consent. The study focused on digital recruitment and raising awareness through media engagement to reach potential participants. Primary methods of recruitment included sharing materials through digital and social media, increasing awareness of the study through press and media engagement, and executing recruitment campaigns with organizations focused on research or health (eg, PCOS Challenge [pcoschallenge.org] and Trialfacts [trialfacts.com]). The primary methods of retention included app notifications as a reminder of survey availability and study updates. The study was approved by the institutional review boards at Advarra (CIRB #PRO00037562) and was registered to ClinicalTrials.gov (NCT04196595). After enrollment, participants completed surveys monthly regarding their conception attempts through the Monthly Survey: Menstrual Update (MSMU). Starting in January 2020, the MSMU was available on the Research app on the first Sunday of each month, after at least 4 weeks since enrollment. We assigned February 2020 as a baseline month (the month before the United States declared a COVID-19 national emergency) and evaluated data through March 2021. We included 21,616 participants who met the eligibility criteria, enrolled before March 31, 2021, and responded to at least 1 MSMU. We excluded participants who self-reported hysterectomy at baseline, menopause, current pregnancy, or lactation.

Apple Women’s Health Study surveys and covariates

Survey questions for the AWHS were derived and administered through the Research app as described previously. At enrollment, participants provided their year of birth, race and ethnicity, gender identity, marital status, and educational level. Participants were able to update demographic information during the study. We categorized education into 3 groups: (1) no college education, (2) college education, and (3) graduate degree. The MacArthur Scale of Subjective Social Status (SSS) was assessed at enrollment. The SSS measures relative social status based on the participant’s perceptions of their educational level, socioeconomic status (SES), opportunity, and current life circumstances. ^, The SSS instructs participants to rate themselves on a “social ladder,” with corresponding numbered social status scale rungs of 0 (lowest) to 9 (highest). We categorized SSS scores as low (score of 0–3), moderate (score of 4–5), and high (score of 6–9).

Outcome assessment: attempting conception status

Participants were asked their attempt to conceive status as part of the MSMU with the question: “Did you actively try to get pregnant in the previous calendar month?” Responding participants selected from one of the following responses: (1) Yes, I tried to conceive naturally; (2) Yes, with the help of methods, such as artificial insemination or in vitro fertilization; (3) No; and (4) I prefer not to answer. We defined attempting to conceive as a dichotomous variable. We categorized participants selecting either response 1 or 2 above as attempting to conceive. Apart from that, we designated participants as not attempting to conceive ( Supplemental Table 1 ).

Of 97,052 total MSMU responses from active participants during the follow-up period, 14,543 (15.0%) were missing responses for one or more months. We singly imputed those missing responses with concordant survey responses before and after the response gap (n=14,020 [14.4%]). Supplemental Table 2 summarizes response concordance among these participants ( Supplemental Table 2 ).

Statistical analysis

We used generalized estimating equations (GEEs) to fit logistic regression models for the association between the proportion of participants attempting conception and the month of response while accounting for longitudinal correlation. We computed month-specific odds ratios (ORs) and 95% confidence intervals (CIs) for the dichotomous outcome of attempting to conceive (yes vs no), about the prepandemic month of February 2020. All models were adjusted for age, race and ethnicity, and marital status. GEEs allow for missing nonconcordant responses before and after the response gap (n=523), with the assumption of missing completely at random. We specified an autoregressive covariance structure. Within the main models, we assessed the effect modification by education status, as education may indicate personal agency vs physical risks of exposure and lifestyle associated with the ability to work remotely during the pandemic. We stratified using the SSS as a secondary analysis. We used deidentified data for all analyses and described aggregate results.

We performed a sensitivity analysis to adjust for seasonal trends modeled by sine and cosine functions, as previously described as an applicable test for seasonality. ^, We performed a second sensitivity analysis that was independent of cluster size by assigning AWHS participants to closed subcohorts by quarter. For these models, we restricted the analysis of participants to the first quarter in which they responded to MSMU and assessed whether the participant attempted conception at least once within that quarter. We performed logistic regression and test for trend analysis. Moreover, we stratified by education and SSS separately. We performed data engineering in Python (version 3.6.5; Python Software Foundation, Wilmington, DE) and statistical analyses in R (version 4.0.4; R Core Team, Vienna, Austria).

Principal findings

Among AWHS participants, we evaluated the relationship between the duration of the COVID-19 pandemic and the participants’ attempts to conceive a child. We observed a slight increase in the odds of attempting to conceive as the pandemic began in May 2020, followed by a decline starting in July 2020, compared with prepandemic levels. Our findings suggested a reduction in the odds of attempting to conceive during the pandemic. In March 2021, we observed that the reduction in the odds was the strongest among participants that were not college educated compared with prepandemic levels. Moreover, our results suggested an increase in the odds of attempting to conceive among participants with a graduate degree.

Results in the context of what is known

Our findings were concordant with other studies performed within the context of the pandemic. A previous study from the United Kingdom found that the COVID-19 pandemic affected pregnancy plans for 53% of women, with 72% of those women reporting postponement of pregnancy. However, this study was based on surveys administered at 1 time point during the pandemic (July 2020). Although our results were not directly comparable with the Guttmacher Survey design, the 2020 Guttmacher Survey also reported that 34% of women desired to delay pregnancy or have fewer children because of the COVID-19 pandemic, indicating that women of lower income were slightly more likely to change their plans.

Our primary research question and the focus of this study were based on measuring attempts to conceive, to understand the intent to conceive during the pandemic. Although fertility and birth rates do not exclusively indicate intent to conceive, we can compare these rates with our findings to assess general trends. Our results have comparable trends with publicly available fertility rate data and birth rate data. For example, the US National Center for Health Statistics from the National Vital Statistics System reported some reduction in fertility rates during the pandemic, from the third quarter of 2020 to the third quarter of 2021. In addition, the National Vital Statistics System reported greater reductions in births in the latter half of 2020 compared with the beginning of 2020 (before June), although we found that data for the rest of 2021 and 2022 were not currently available.

Clinical and research implications

Our results may potentially be attributed to increased financial security and flexibility to work from home among advanced-degree workers. Working from home may reduce the potential for exposure to COVID-19 and enable participants to focus on pregnancy planning rather than daily concern about virus exposure at work. In this way, education may serve as a proxy for financial security and personal agency vs physical risks of exposure associated with the ability to work remotely and thereby act as a potential mitigatory factor for the effect of the pandemic. In contrast to our study, other research studies found that the desire for children is generally higher among those of lower educational levels and that women of higher SES had lower net fertility, potentially because those of higher SES respond faster to changing family norms. However, these analyses were performed with data collected from a nonpandemic period.

When comparing results stratified by SSS vs educational level, we observed a trend of reduced odds in attempting conception among those participants with the lowest SSS score ( Supplemental Table 8 ), which was similar to the trend among those participants with no college education. Low SSS and no college education may each indicate factors, such as financial insecurity. Attempts to conceive among those participants with a high SSS were similar to those participants with a low SSS, which was notably different from the findings that we observed for attempts to conceive among participants with the highest educational level. This variation may likely be related to the contextual differences in the SSS, which incorporates measures of relative social status based on current life circumstances, rather than being an indication of objective education or income alone. This distinction between SSS and educational attainment is demonstrated by the matrix in Supplemental Table 12 .

Strengths and limitations

Our study has many strengths. Our analysis allowed for a more microscopic, monthly view of the changes in attempts to conceive. Previous studies have had challenges in gathering population-level data on monthly pregnancy attempts in that they occur approximately ≥10 months before birth. As our digital study was launched before the pandemic, we were able to follow pregnancy attempts during the pandemic. In contrast to existing studies that are cross-sectional and rely on respective reports, as described previously, our study was prospective. In addition, we collected sufficient demographic information to understand the trends in conception attempts stratified by educational level. Our study focused on combining frequently collected survey data and menstrual health indicators to study women’s health.

Our study has several limitations. The generalizability of this study may be limited as the AWHS cohort was restricted to iPhone users and demographic characteristics from this cohort of participants from the AWHS may not be representative of the US population. The AWHS cohort included more White participants (71.6% in the AWHS vs 60.1% in the United States) and fewer Black participants (5.3% in the AWHS vs 13.4% in the United States) ^, than the US population, per the US census. Because of the racial homogeneity of participants in the study population, we were not able to evaluate racial disparities in COVID-19 burden and attempts to conceive. ^, In addition, women who are attempting conception may be more likely to enroll in the AWHS study. However, pregnancy planners were not likely to be oversampled as they were not targeted in recruitment efforts. Moreover, the study was an open cohort, with changes in the study population over time, as demonstrated in Table 2 .

Furthermore, we did not confirm self-reported conception attempts with sexual activity logging. However, the purpose of our investigation did not focus on outcomes, such as time to pregnancy. Another potential concern was attrition bias because of the potential effect of COVID-19 and pandemic-related factors on fertility. If these factors did reduce fertility, we would likely not see a bias because of attrition. However, we might find increases in the level of the month-specific pool of participants reporting yes to attempting conception that may potentially inflate the proportion of those attempting in any given month during the pandemic.

Because of the timing of the study, we were unable to account for seasonality based on historic data before COVID-19, but we incorporated sine and cosine functions to account for seasonality and did not find varying results. Moreover, although we did not collect data on the type of employment, income, or work-from-home status, we were able to evaluate educational level. Although education may be a poor proxy for SES, education is an objective measure and may in its way predict some behavioral factors related to lifestyle, such as work-from-home status. As the SSS was a subjective measure of social status and did not objectively measure education, SES, or income status, we included this assessment as part of a secondary analysis.

Another potential limitation may be related to informative clustering. The longer a participant has attempted to conceive at any point within the study, and stays within the study, the more likely the participant may be to attempt to conceive in any month thereafter. To assess findings independent of cluster size, we performed a sensitivity analysis assigning participants to quarter cohorts and found similar results, although less precise ( Supplemental Tables 10 and 11 ). Finally, we experienced loss to follow-up in this cohort, although we did not observe changes in demographics among participants who were vs were not lost to follow-up.

Conclusions

Overall, these findings indicated a potentially noteworthy trend in the decline in the attempts to conceive related to factors associated with the COVID-19 pandemic, especially among participants with lower educational levels. Barriers to healthcare and loss of financial security have disproportionally affected low-income communities during the pandemic. This research demonstrated the importance of collecting such detailed and granular data for women’s health studies to better understand environmental and social factors that influence health-related decision-making.