Validating the current duration approach for measuring infertility prevalence using novel app data from the USA.

study questionAre the assumptions required by the current duration (CD) approach to estimating population infertility met, and does the CD approach produce similar infertility estimates to those from the 'gold standard' incident prospective cohort approach within the same sample using fertility app data? SUMMARY ANSWER: While we find evidence of CD assumption violation, once addressed, the CD approach produces comparable infertility prevalence estimates and patterns to those from an incident prospective cohort design when selecting from the same sample of women. WHAT IS KNOWN ALREADY: The CD approach to documenting population and subgroup infertility is a promising, more feasible, and cost-effective alternative method to the incident prospective cohort design. However, as a field, we do not have sufficient evidence to use the CD approach for the study of population infertility because method assumptions have not been rigorously tested using population-based samples. STUDY DESIGN, SIZE, DURATION: We use prospective cohort data provided by users of smartphone applications in which they record fertility and pregnancy information. The total prospective cohort sample included 167 451 people using the applications between 1 January 2015 and 28 February 2022. To generate valid comparisons, we drew six CD samples from the same data source. The samples contained between 17 196 and 26 259 people measured in designated months. PARTICIPANTS/MATERIALS, SETTING,

methodsParticipants are users of a fertility and pregnancy smartphone application aged 18-44 years who were trying to become pregnant and began using the smartphone application during the study period. We first tested the three assumptions required by the CD approach. We then estimated 12-month infertility prevalence separately using prospective and CD approaches, first for all users, and then for subpopulations stratified by age, education, parity, and level of poverty. We assessed whether estimates of population infertility using prospective and CD approaches were statistically different, overall and for each subpopulation. MAIN RESULTS AND THE ROLE OF CHANCE: We demonstrate clear violation of one CD assumption, and suggestive evidence of one other assumption violation. The prospective survival function resulted in a 12-month infertility prevalence of 36.1% (95% CI 35.8-36.3) while the CD estimate adjusted for assumption violation was 37.0% (95% CI 36.2-37.7). Infertility patterns by user characteristics were similar for prospective and CD estimates, with greater levels of infertility for those 35 years old or greater, those with no children, those with less education, and those residing in areas with greater levels of poverty. LIMITATIONS, REASONS FOR CAUTION: Potential sample selection and censoring do not affect the internal validity of within-sample comparison tests, which answer the study's primary research questions. However, selection and censoring may contribute to sample non-representativeness and the higher than expected levels of infertility relative to the US population. WIDER IMPLICATIONS OF THE

findingsResearchers can potentially learn about patterns of infertility with cross-sectional data alone, provided they statistically address CD assumption violations. This finding facilitates the study of infertility in populations, subpopulations, places, and periods for which infertility estimates do not exist. Although we weighted the samples to reflect the US population sociodemographic characteristics, users of this app may underreport pregnancy and/or may be subfertile, thus, the infertility prevalence estimates reported here should not be interpreted as population-level estimates for the USA. STUDY FUNDING/COMPETING INTEREST(S): This study was supported by grants from the National Institute of Child Health and Human Development (R01HD102207, K01HD107172, P2CHD047873, P2CHD042854). The funders were not involved in the study design, analyses, manuscript writing, or decision to publish. S.M. is an employee of Ovia Health by Labcorp, and owns stock in Labcorp Holdings. Ovia Health created the fertility app from which the data for the analysis were sourced. Other authors have no conflicts of interest to disclose. TRIAL REGISTRATION NUMBER: N/A.