Study overview

The ADAPT study enrolled and prospectively followed an underlying cohort (UC) of participants who were not pregnant (or were having an abortion) at screening (figure 1). We measured the degree to which participants desired to avoid pregnancy multiple times over 1 year using the DAP scale (range: 0–4, with 4 representing a higher DAP) and captured incident pregnancies as they occurred. Participants who experienced new pregnancies during the 1 year UC transferred into a new cohort, the Pregnant and Non-pregnant Matched Comparison Cohort (PMC), where we continue to follow them to document and describe their feelings about the pregnancy, pregnancy decision-making and healthcare-seeking. In addition, we follow them through their pregnancies to investigate differences in health and well-being associated with carrying a pregnancy to term based on the participant’s prepregnancy and postpregnancy preferences.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

ADAPT study schematic. The schematic depicts participant flow through the ADAPT study and the study survey schedule. ADAPT, Attitudes and Decisions After Pregnancy Testing.

Finally, we are following a cohort of non-pregnant women from the UC to serve as a comparison group to pregnant participants and to emulate a hypothetical target trial in which participants are randomised to pregnancy at trial enrolment. Approximately monthly over the course of the UC, as participants experienced and reported incident pregnancies, we selected a set of non-pregnant comparison participants to also follow in the PMC. These participants were frequently matched to those with incident pregnancies (1:1 ratio) on DAP scale score to balance the groups in terms of pregnancy preferences. We also matched comparison participants to pregnant participants on time in the UC until pregnancy, or time at risk of pregnancy, to ensure ‘treatment’ assignment (pregnancy) and initiation of follow-up for the PMC component occurred at about the same time.37 Selection was random among possible ‘eligible’ non-pregnant participants; selected non-pregnant comparison participants who experienced a pregnancy within a year of UC enrolment were themselves assigned non-pregnant matched comparisons. The non-pregnant comparison participants represent a reasonable counterfactual situation of the pregnant participants had they not experienced pregnancy, our exposure of interest for aim 3b. We will compare the outcomes of participants with new pregnancies and new births to those in the non-pregnant comparison group to assess the effect of pregnancy. Specifically, we will investigate differences in health and well-being among those experiencing versus avoiding less desired pregnancy, as well as of experiencing not attaining a pregnancy when desired.

In summary, the ADAPT study has the following aims:

Aim 1: Assess the factors associated with people’s pregnancy and childbearing preferences, how preferences change over time and their associations with contraceptive use, incident pregnancy and feelings about the pregnancy after discovery.

Aim 2: Investigate the options that people consider when they become pregnant and the factors that influence their pregnancy decision-making and ability to access desired reproductive healthcare and services (prenatal, abortion, or adoption).

Aim 3a: Examine the effects of giving birth from a less desired pregnancy, measured prospectively on a continuum, as compared with a more desired pregnancy, on parental and children’s health and well-being.

Aim 3b: Examine the effects of experiencing pregnancy and birth, accounting for pregnancy preferences, on people’s health and well-being, as compared with not experiencing pregnancy.

This is a social science, behavioural study that does not use clinical data or biological markers.

Recruitment strategy and target population

We recruited ADAPT study participants from reproductive and primary care facilities in five US states: Arizona, southeastern California, Nevada, New Mexico and West Texas. We designed the recruitment strategy based on findings from a pilot study (n=800, 2016–2017).34 38 The goals of the pilot were to develop study measures and to identify a feasible strategy to recruit participants prior to recognition of pregnancy, given the importance of establishing prospective pregnancy preferences. We considered recruiting people presenting for pregnancy tests at clinics but found in the pilot that 75% of pregnant respondents had first confirmed their pregnancy by taking an at-home pregnancy test (and 75% of non-pregnant respondents said they would do the same if they suspected pregnancy).38 We considered recruiting pregnancy capable people purchasing at-home pregnancy tests from drug stores, but—like those presenting for care for suspected pregnancies—these people already suspect pregnancy, and we believe pregnancy and childbearing preferences can change on pregnancy suspicion. We concluded that identifying individuals who are pregnant, but do not suspect it, would be infeasible, and thus moved to clinic-based recruitment strategies to identify people likely to experience incident pregnancy, including less desired pregnancy. A significant advantage of clinic-based recruitment is that it provides a clear target population for generalisability and a way to assess representativeness by comparing characteristics of people enrolled to those who are not. We, thus, recruited participants from reproductive health and primary care facilities, including Federally Qualified Health Centers (FQHCs), Planned Parenthood health centres and independent and community clinics. We aimed to select safety-net facilities that varied in patient volume and urbanicity.

We selected the southwestern USA as the study region to capture the experiences of demographically similar populations living in diverse sociocultural and reproductive health policy environments.39 Study states all have higher than average percentages of the population living in poverty.40 Given their proximity to the US-Mexico border, study states have among the highest proportions of the population identifying as Hispanic/Latine (32% Arizona, 42% California, 29% Nevada, 49% New Mexico, 39% Texas vs 18% nationally),41 who experience higher adolescent birth rates relative to people of other races and ethnicities, and high ‘unintended’ birth rates.5 42

Participant eligibility and recruitment

Given the study’s focus on evaluating the effects of pregnancy, including less desired pregnancy, on people’s lives, we selected eligibility criteria to focus on individuals likely to experience an incident pregnancy across a spectrum of pregnancy preferences. To participate, individuals had to be 15–34 years of age (pregnancy rates are relatively lower for those aged 35 and over)1; female, trans-male or gender non-conforming with a uterus; sexually active in the past 3 months with someone who has sperm; and not sterilised, not using a long-acting contraceptive and not pregnant (or having an abortion). Individuals needed to be able to speak and read English or Spanish, reside in a study state or a bordering state, have access to a phone, smartphone or the internet and be willing to be contacted by the research team over 1–4 years.

UCSF research associates (RAs) carried out all recruitment activities. RAs were bilingual (English and Spanish) staff who had experience working with diverse communities in health and social service settings. They received in-depth training and were involved in designing study instruments and protocols. Beginning in March 2019, RAs travelled to selected recruitment facilities, stationed themselves in clinic waiting rooms and interacted with potential participants in person. Recruitment facility front desk staff briefly alerted all English-speaking and Spanish-speaking patients presenting as female, trans-male or gender non-conforming about the study and the presence of the RAs in the waiting room. Facility staff handed patients a voluntary paper eligibility screener in English or Spanish. RAs approached patients in the waiting room and confidentially reviewed the screeners. If a patient was eligible and interested, an RA gave them a recruitment flyer and described the study using a standardised script. If an eligible patient chose to participate, the RA showed them an electronic informed consent form on an iPad, reviewed it with them discretely in the waiting room and obtained verbal consent. The RA assigned the participant a numeric ID and launched a secure electronic informed consent form on a study iPad which the participant signed. Participants also provided their contact information and confidentiality and remuneration preferences. On recruitment days, the RA documented screening, eligibility status and participation for every patient in designated clinic reception areas so that we could rigorously assess generalisability of study participants to the target population.

Due to travel restrictions and clinic safety regulations implemented during the COVID-19 pandemic, we paused recruitment between February and September 2020 and supplemented in-person recruitment with remote approaches beginning in October 2020. These approaches varied slightly to suit recruitment facility requirements, but all targeted the same patient population as the in-person approach. At several facilities, front desk staff handed out paper-based recruitment flyers to patients, posted flyers in the facilities or displayed flyers on monitors in waiting areas (in-clinic recruitment via flyer). Five facilities sent paper-based postcards or used secure patient database systems to disseminate ‘Dear Patient’ messages to potentially eligible adult patients who had previously agreed to receive communications from the institution via text, email or mail or posted study information on their social media accounts (ie, Facebook, Instagram) (remote recruitment via facility outreach). Recruitment flyers, postcards, ‘Dear Patient’ messages and posts included a toll-free phone number, email, URL link and QR code; the latter two linked to an online study description and self-screener. RAs then contacted and recruited prospective participants via phone, screened them for eligibility (or confirmed eligibility for those who had already been screened) and obtained informed consent verbally over the phone.

Data collection procedures

The ADAPT study uses a nested cohort study design (figure 1). All participants initially enrolled into the 13.5-month UC, where they completed a baseline survey, quarterly follow-up surveys (3, 6, 9, 12 months) and interim pregnancy check-in surveys (1.5, 4.5, 7.5, 10.5, 13.5 months). We offered participants the option to self-administer confidential surveys online or to participate in interviewer-administered surveys over the phone; nearly all completed surveys online. RAs sent participants email or smartphone links to online surveys. The baseline survey included questions about sociodemographics, relationships and life context, the DAP scale, and past and current health behaviours. Follow-up surveys included these questions and others about new suspected or confirmed pregnancies. Interim check-in surveys only collected information on incident pregnancies. Participants received a gift card after completing the baseline (US$50) and each UC follow-up (US$20) and check-in (US$5) survey.

As of 2024, participant follow-up is complete for the UC and the data collection is still underway for the PMC Cohort. Participants remained in the UC until they reported a pregnancy, were selected as a non-pregnant matched comparison or exited the study at 13.5 months. Participants who reported an incident pregnancy while they were in the UC moved out of that cohort and into the PMC, where they were followed for 3 years. Only respondents who had confirmed their pregnancy with a urine pregnancy test or healthcare provider moved into the PMC; participants who reported suspected, unconfirmed pregnancies remained in the UC, as did participants who reported early miscarriages.

In the PMC, participants complete surveys at cohort entry, 6 weeks, and 3, 6, 9, 12, 18, 24, 30 and 36 months. PMC surveys ask participants about their feelings about the pregnancy, pregnancy circumstances, care-seeking, and health and well-being outcomes, including mental health symptoms and diagnoses (eg, stress, depression and anxiety), physical health conditions and behaviours (eg, self-rated health, pain and role limitations), well-being, socioeconomic status (eg, food insecurity) and birth outcomes. Non-pregnant matched comparisons follow the same survey schedule as the pregnant participants and are asked about health and well-being outcomes. Participants receive a US$50 gift card for the first and last PMC surveys and a US$35 gift card for those in between. Participants receive information and referrals via pop-up notifications in their survey if their responses indicate a potential need (eg, mental health and interpersonal violence).

To promote study retention, we conduct quarterly US$50 gift card drawings, disseminate electronic newsletters and periodically implement process improvements suggested by the study’s participant advisory group, composed of former study participants.

Patient and public involvement

Patient participants from the ADAPT pilot study were involved in the development and testing of the DAP scale,34 however, they were not involved in the development of the research questions and other aspects of the surveys. 15 former ADAPT study participants served as part of two participant advisory groups and provided input on study processes and experiences to improve protocols and dissemination strategies. Study updates and select initial results are disseminated through electronic newsletters to active participants and those who exited the study and expressed interest in learning about study results. A stakeholder advisory group is being assembled to provide input on interpretation and dissemination of study results.

Target sample size

We determined our target sample size based on the primary study outcome of stress (Perceived Stress Scale, PSS-4, range 0–16) for analyses investigating prepregnancy DAP score as the independent variable for aim 3a. Given that our treatment of pregnancy preferences as a continuum is new and relevant cut-points do not yet exist, we assumed two groups with a 2:1 desired:undesired (or preferred:unpreferred) ratio among incident pregnancies, using a cut-point of 2 on the DAP measure’s 0–4 scale; mean PSS 5.0 for desired and 6.0 for undesired, with a single measurement SD of 2.5 in each group. On average, we assumed four repeated measurements per participant, with an intrasubject correlation of r≤0.7, and thus a SD of the mean PSS of 2.2 points. We built in a penalty for facility clustering design effects (1.14). With an alpha of 0.05 and power of 0.8, we would require a sample size of 148 (desired)+74 (undesired)=222 pregnancies to detect the 1-point PSS-4 difference. We increased the sample size, accounting for 15% of pregnancies ending in miscarriage,43 and 20% of undesired pregnancies ending in termination,44 and then rebalancing the sample to get the 2:1 ratio, for 324 pregnancies (216 desired, 108 undesired). Assuming 18% of participants will experience pregnancy in a year,5 45 46 we would need to enrol 1800 participants to achieve 324 pregnancies. We assumed 80% follow-up between study enrolment and birth (10% loss before baseline, 10% loss before reporting a pregnancy or birth), leading to a final enrolled sample size goal of 2200.

Recruitment facility characteristics

A majority of the 23 reproductive and primary healthcare recruitment facilities were FQHCs and/or Title X funding recipients (52%); 26% were Planned Parenthood health centres; and the others were independent, community or university-affiliated clinics (table 1). All facilities offered reproductive health services and about half also provided general primary care. All offered services in Spanish and served uninsured/underinsured patients; a majority served undocumented patients, adolescents and men. Facilities were located across the five study states, with 26% in Arizona, 23% in California, 23% in Texas, 17% in New Mexico and 13% in Nevada, and they served a range of urbanicities. A median of 74 participants from each of the 23 facilities completed the baseline survey.

Enrolment

We recruited participants for the ADAPT study between 16 March 2019 and 19 October 2022 (figure 2). Among the 10 047 unique individuals eligible for screening, we screened 8727 (87%). Among those, 5810 (67%) were ineligible for participation, most commonly due to current pregnancy (36%), being outside of the eligible age range (30%) or using an intrauterine device or implant (22%). A full 2612 (30%) were eligible, 2168 from in-person recruitment and 444 from remote recruitment. For in-person recruitment, 2000 (92%) eligible people consented to participate, and 1752 (88%) completed the baseline survey. For remote recruitment, 270 (61%) eligible participants consented, and 263 (97%) completed the baseline survey. Altogether, 2270 (87%) eligible participants consented, and 2015 (89%) of them completed the baseline survey and comprise the analysis population.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Flow chart of ADAPT study enrolment. The flowchart depicts numbers of potential study participants screened, eligible, consented, completing the baseline survey and included in the pregnancy and non-pregnant matched comparison cohort of the study. ^aPatients were ineligible if currently pregnant and not planning to start an abortion on the day of recruitment. ^bIn-person recruitment involved a study RA screening and enrolling potential participants in recruitment facility waiting rooms. Remote recruitment involved potential participants learning about the study from the recruitment facility, undergoing screening electronically and undergoing recruitment over the phone with a study RA. ^c35 non-pregnant comparison participants experienced a pregnancy within a year of recruitment. ADAPT, Attitudes and Decisions After Pregnancy Testing; IUD, intrauterine device; RA, research associate.

Based on random effects logistic regression analyses adjusting for facility clustering, participants who enrolled and completed the baseline survey did not differ significantly from eligible people who did not complete the baseline survey in terms of age (mean 24.8 vs 24.9; p=0.15), race and ethnicity (p=0.70) or language (3.2% vs 4.9% Spanish; p=0.14).

Underlying Cohort participant characteristics

The cohort of participants completing baseline (n=2015) was on average 25 years old, with 15% being adolescents aged 15–19 years (table 2). Eight per cent of the sample identified most strongly as black, while 59% identified as Hispanic or Latine, 21% as white and 13% as multiracial or another race. Overall, 46% were living with a main romantic partner; another 36% had a main partner but were not living with them. Over half (56%) had no children, while 20% had one, 14% had two and 10% had three or more children. About half (49%) reported that religion was very or somewhat important to them. Almost one-third (32%) lived below 100% of the federal poverty line, and 37% had experienced food insecurity in the prior month.

The large majority (86%) of the sample was recruited by RAs in person in the recruitment facility. Six per cent were given a study flyer by front desk staff or saw a flyer in the facility and were screened by an RA over the phone, and 9% were adult clinic patients informed about the study by postcard or electronic ‘Dear Patient’ message sent by recruitment facility staff and screened by an RA over the phone. Overall, 29% were recruited in southeastern California, 28% in West Texas, 23% in Arizona, 13% in Nevada and 7% in New Mexico.

In terms of mental health, participants at baseline scored on average 7 (SD: 3) on the PSS47 (range: 0–16); 7 (SD: 6) on the Patient Health Questionnaire48 (range: 0–24) measuring depressive symptoms; and 7 (SD:6) on the General Anxiety Disorder scale49 (range 0–21) (table 3). 30% were currently experiencing moderate or severe depressive symptoms, and 37% had ever been diagnosed with a depressive or anxiety disorder or anorexia or bulimia.50 About 6% had consumed four or more alcoholic beverages weekly, almost daily or daily over the prior month51; 3% had used illicit, street or prescription drugs.52

For well-being, participants scored on average 44 (SD: 7) on Diener’s Flourishing Scale53 (range: 8–56). Over one-quarter (27%) had ever experienced physical intimate partner violence, and about half (49%) had ever experienced emotional physical intimate partner violence.54

For physical health, about half (49%) had ever been diagnosed with a chronic physical health condition. About 37% rated their own health to be fair or poor, and 21% had experienced moderate to severe bodily pain in the last 4 weeks.55

Pregnant and Non-pregnant Matched Comparison Cohort participant characteristics

Over the 13.5-month UC, 335 participants reported experiencing an incident pregnancy, and 336 non-pregnant comparison participants were selected to also be followed as part of the emulated trial portion of the study (aim 3b) (table 4). The two groups did not differ based on DAP scale score (mean: 1.7 prior to pregnancy vs 1.8 prior to becoming a non-pregnant matched comparison, p=0.11). Pregnant participants had been followed as part of the UC for a mean of 0.4 years, vs 0.5 years for the non-pregnant comparisons (p<0.001), owing to the monthly timing of our match selection process. The two groups were generally well balanced according to sociodemographic and reproductive characteristics with one exception: pregnant participants were more likely to have a main partner with whom they lived (60% vs 54% for non-pregnant participants) and less likely to have no main partner (11% vs 20% for non-pregnant participants, overall p<0.01). This difference is likely due to the greater frequency of sex among cohabiting partners.

One-year retention

Overall, 96% of the sample was retained in the study beyond the baseline survey. A full 94% remained in the sample at 3 months, 93% at 6 months, 90% at 9 months and 87% at 1 year. The 1-year retention rate was 85/100 person-years (PY). In a series of Cox proportional hazards models with robust SEs to account for recruitment facility clustering, first year retention was non-differential by baseline characteristics including race and ethnicity, nativity, partnership and cohabitation, parity, in school, employment status, food insecurity, religiosity and DAP score. Retention differed by age group (81/100 PY for 15–17 years, 73/100 PY for 18–19 years, 84/100 PY for 20–24 years, 90/100 PY for 25–29 years, 85/100 PY for 30–34 years, p<0.001) and state of recruitment site (82/100 PY in Arizona, 79/100 PY in Nevada, 85/100 PY in New Mexico, 88/100 PY in Texas and 82/100 PY in California, p=0.02).

Findings to date

An interrupted time-series analysis of the 627 participants recruited prior to the onset of the COVID-19 pandemic investigated changes in pregnancy desires over the year before and the first year of the pandemic (March 2019–February 2021).56 Pandemic onset in the southwestern USA in the Summer of 2020 was associated with a significant but short-term stalling of a prior trend towards a greater desire for pregnancy over time. Participants aged 15–24 years and nulliparous and primiparous participants experienced greater declines in DAP score prior to the summer surge, and greater reversals of decline between Summer and Fall 2020, than did those who were aged 25–34 years and multiparous.