Predictors of Non-Response to a Sexual Health Survey in a North American Preconception Cohort Study.

Pregnancy Study Online (PRESTO) is a web-based prospective cohort study of pregnancy planners (detailed methods described previously 14 ). Briefly, participants were recruited via online advertising, posted flyers, and word of mouth. At baseline, eligible participants identified as female and were aged 21–45 years, residents of the United States (U.S.) or Canada, in a relationship with a male partner, and not using contraception or fertility treatment. Upon enrollment, participants completed a baseline questionnaire that captured detailed information on demographics, behaviors, and reproductive and medical history. Every 8 weeks participants received a follow-up questionnaire to update health history and assess pregnancy status. Beginning in March 2021, we emailed participants 30 days after enrollment and invited them to complete the SQ. The SQ contained 47 questions (most participants answered fewer due to skip patterns), and included two validated scales: the Female Sexual Function Index 6-item scale 15 and the Female Sexual Distress Scale, 16 as well as questions about experiences of pain with intercourse, lubricant use during intercourse, partner sexual function, recent vaginal infections, discussion of sexual function with health care provider, preconception intercourse frequency, and history of medical conditions relevant to sexual functioning. The questions had a variety of response options, including Likert scales, binary responses, and ‘check all that apply’ (see Supplemental Table 1 for more details). Participants were informed that the SQ was optional and they could complete it any time. The study was approved by the Boston University Medical Campus Institutional Review Board and informed consent was obtained online from each participant. This analytic sample included participants who enrolled in PRESTO between the SQ launch (March 2021) and December 2021 ( Supplemental Figure 1 ). We restricted to those for whom at least 11 weeks had passed since enrollment, to ensure sufficient opportunity to complete at least 1 follow-up questionnaire in addition to the SQ. We created 3 groups: (i) participants who only completed the baseline questionnaire but did not complete any follow-up questionnaires or the SQ (baseline only), (ii) participants who completed at least 1 follow-up questionnaire but not the SQ (follow-up only), and (iii) participants who completed the SQ regardless of their response to a follow-up questionnaire (SQ responders). Participants with <11 weeks of follow-up who completed the SQ were included as SQ responders. With these groupings, we attempted to distinguish between non-response due to general loss to follow-up and non-response among otherwise engaged participants. The latter may represent explicit refusal to complete the SQ. We selected a variety of potential predictors of response, ascertained via self-report on the baseline questionnaire. First, we assessed factors related to sexual experience: lifetime number of sexual partners (1, 2–4, 5–9, 10−14, ≥15), relationship duration (<5, 5–9, ≥9 years), and frequency of intercourse in the past month (≤1/month, 2–3/month, 1/week, 2–3/week, ≥4/week). We also evaluated factors that could be important covariates in clinical research related to sexual health (all categorized yes/no): history of diagnosed depression, anxiety, sexually transmitted infections, abnormal Papanicolaou test, polycystic ovarian syndrome, and endometriosis. We evaluated factors relevant to sexual health and fertility: months of pregnancy attempt at study entry (≤1, 2–5, 6–11, ≥12), parity (0 vs ≥1 births), history of ever trying to become pregnant for ≥12 months (“infertility”: yes, no, no prior pregnancy attempts), and whether participants were currently doing anything to try to become pregnant (ie, timing intercourse for the fertile window) (yes/no). We evaluated factors related to the potential cultural salience of the SQ: self-identified race/ethnicity (Hispanic/Latina, non-Hispanic Black, non-Hispanic Asian/Pacific Islander, non-Hispanic mixed race/Alaska Native/Native American/other, Non-Hispanic White), geographic region (Canada, Northeastern U.S., Southern U.S., Midwestern U.S., Western U.S.), and whether participants were born outside of the U.S. or Canada or lived outside of the U.S. or Canada at age 15 (both yes/no). We finally evaluated factors that are often confounders in clinical and epidemiologic research: age (<25, 25−29, 30−34, ≥35 years), marital status (married/unmarried), annual household income in U.S. dollars (<50,000, 50,000–74,999, 75,000–99,999, 100,000–149,999, ≥150,000), educational attainment (<16, 16, ≥17 years), and current employment (yes/no). In separate analyses, we compared SQ responders with baseline-only responders and follow-up only responders. We calculated the probability of response to the SQ within each level of each predictor. We selected one level of each predictor to serve as a reference group and calculated the absolute difference in the probability of SQ response between that category and the other categories of a given predictor. We also calculated 95% confidence intervals for the difference in probability of SQ response.

Our study included 1,491 participants. Of these, 302 (20.3%) completed the baseline questionnaire only, 259 (17.4%) completed a follow-up questionnaire but not the SQ, and 930 (62.4%) completed the SQ. The SQ responder group included 84 participants who had not yet been enrolled for 11 weeks but did complete the SQ and 32 participants who did not fill out a follow-up questionnaire, despite being eligible. When we compared SQ responders with baseline-only responders, many predictors were strongly associated with SQ non-response, resulting in an absolute percentage difference >10 ( Table 1 ). As months of pregnancy attempts at study entry increased, the percentage of responders decreased (those with ≥12 months of pregnancy attempt at entry had an absolute difference of −44.4% (95% confidence interval [CI] −56.9%, −31.1%) when compared with those with attempt times of ≤1 month. SQ response increased with older age and decreased with lower education and income. The probability of SQ non-response was highest in the highest category of intercourse frequency (≥4 times/week) compared with other categories. Participants who were born or lived outside the U.S. or Canada at age 15 years had a lower probability of SQ response, as did those who identified as non-Hispanic Black or Hispanic/Latina. When we compared SQ responders with follow-up questionnaire responders, we observed few strong predictors of SQ non-response ( Table 2 ). Participants who identified as Hispanic/Latina or non-Hispanic Asian/Pacific Islander had absolute reductions of 10.0% (95% CI −23.7%, 3.7%) and 12.5% (95% CI −37.5%, 12.9%), respectively, in SQ response compared with participants who identified as non-Hispanic White. Probability of SQ response increased with age, with those aged ≥35 having a 12.2% higher response probability than those <25 years (95% CI −5.0%, 29.4%). Those who reported having intercourse ≥4/week had the lowest probability of SQ response compared with lower frequency categories, but the absolute differences were small.

Systematic differences between research participants and non-participants (sometimes called volunteer bias) can affect the validity of study findings in 2 distinct ways. First, when factors related to participation are also related to the exposure and outcome under study, selection bias can distort observed measures of association. 1 , 2 This bias can occur during initial enrollment or, in the case of a prospective cohort study, during follow-up. 1 Second, enrollment of a study population that is systematically different from the target population can limit generalizability. 3 Concerns that sex research participants are meaningfully different from non-participants because of the private and taboo nature of sexual behavior are well-documented. 3 , 4 As such, a wealth of studies have evaluated factors that influence participation in sex research. Such studies have approached the question in 2 primary ways: (i) comparing participants who completed a study about a non-sexual topic to participants who completed a study about sexual behavior 5 – 7 or (ii) administering a primary study with some sexuality measures included and then asking participants whether they would volunteer for subsequent sexuality studies, typically with increasing invasiveness. 3 , 8 – 12 These studies consistently find that participants report more sexual knowledge, experience, interest, partners, and liberal sexual attitudes than non-participants. 3 , 8 – 10 , 13 However, most studies were conducted before the proliferation of web-based studies 3 , 8 – 10 , 13 and have focused on psychological attributes. Less is known about how participation may differ “online” or may relate to factors relevant to epidemiologic research about sexual health, such as medical history. We evaluated predictors of response to the “Sexual Health and Wellbeing Questionnaire” (SQ), a supplemental survey added to an established web-based prospective cohort study, focusing on predictors of response that may be important in epidemiologic or clinical studies of sexual health. By assessing whether there are strong predictors of response, we aimed to assess the potential for threats to internal or external validity.

This descriptive study evaluated predictors of non-response to the SQ, an optional survey about sexual health added to an established web-based preconception cohort study. When we compared SQ responders (with or without completion of at least 1 follow-up questionnaire) with those who completed at least 1 follow-up questionnaire but not the SQ, few factors were strongly related to SQ non-response. However, when SQ responders were compared with baseline only responders, there were many factors strongly associated with SQ non-response. These findings suggest that among members of a cohort study who may have intentionally refused to complete the SQ, non-response did not appear to be strongly related to evaluated demographic or clinical factors. Instead, the bulk of SQ non-response seemed to be related to general loss-to-follow-up as opposed to specific refusal of the SQ. Previous studies assessing participation in sex research have focused primarily on psychological attributes and sexual attitudes, 3 , 8 − 12 making a direct comparison with our results difficult. When we evaluated lifetime number of sexual partners, those reporting more partners had slightly higher SQ response, as would be expected given research demonstrating that higher levels of sexual experience and/or interest are positively related to participation in sex research. 3 , 8 , 12 Contrary to expectation, those in the highest category of intercourse frequency had slightly lower SQ response, though the absolute differences were small. This may imply that intercourse frequency in our study population of pregnancy planners is not a reliable proxy for higher sexual experience and/or interest. Given that the majority of participants completed the SQ, our results are consistent with a more recent study that found that most participants enrolled in a general psychology study were also willing to complete a sexuality-related online survey. 12 Additionally, recent national sexual health surveys from many European countries, including Germany, 17 Ireland 18 and the United Kingdom 19 have successfully recruited generally representative samples. Our findings extend these results to an existing longitudinal cohort study in the U.S. and Canada. National surveys are resource-intensive and often cost prohibitive. The fact that we did not observe large systematic differences between responders and non-responders indicates that the addition of questions or supplemental surveys related to sexual health to ongoing studies may be a feasible way to expand the reach of epidemiologic and clinical sex research in the absence of, or in addition to, a national survey. Those who identified as Hispanic/Latina or non-Hispanic Asian/Pacific Islander had lower SQ response, which warrants further investigation of the role of ethnicity and/or culture in sex research participation. Additionally, clinicians should consider evaluating whether sexual health measures administered to patients are acceptable cross-culturally. Study limitations include that we had only rough proxy variables for sexual attitudes and experience (ie, intercourse frequency, lifetime number of sexual partners). Thus, we may have missed more nuanced differences in personality traits between responders and non-responders. However, in an epidemiologic context, personality traits may only be of interest insofar as they affect the risk of an exposure or an outcome via their associations with behaviors and comorbid conditions. Finally, our study focuses on a generally highly-engaged population of pregnancy planners which may limit generalizability. Because pregnancy planning is related to sexual intercourse, it is possible that this population is more willing to answer questions related to sexual behavior. Our findings also may not generalize to clinic-based sex studies. Despite these limitations, the implications of our study are that among an established cohort study, participation in an optional sexual health survey resulted in a response rate >60% and did not seem to be strongly related to key demographic, lifestyle, relationship, or clinical factors, which may assuage concerns about threats to internal and external validity due to non-response. Though we observed strong predictors of non-response when we compared SQ responders with baseline only responders, our findings seemed to be related to general loss to follow-up in the study. In conclusion, our findings echo other recent studies 11 , 12 that suggest that online sexuality surveys are broadly acceptable.