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

Area of Science:

• Reproductive Health
• Epidemiology
• Biostatistics

Background:

• The current duration (CD) approach offers a feasible, cost-effective method for assessing population infertility compared to traditional prospective cohort studies.
• However, rigorous testing of CD approach assumptions using population-based samples is lacking, limiting its widespread adoption.

Purpose of the Study:

• To evaluate whether the assumptions of the current duration (CD) approach for estimating population infertility are met.
• To compare infertility estimates derived from the CD approach with those from the gold-standard incident prospective cohort approach using fertility app data.

Main Methods:

• Utilized prospective cohort data from 167,451 users of a fertility smartphone application (2015-2022).
• Generated six CD samples (17,196-26,259 participants) from the same data source for comparison.
• Tested CD approach assumptions and estimated 12-month infertility prevalence using both prospective and CD methods, stratified by age, education, parity, and poverty.

Main Results:

• Demonstrated clear violation of one CD assumption and suggestive evidence of another.
• The prospective cohort method yielded a 12-month infertility prevalence of 36.1%, while the adjusted CD estimate was 37.0%.
• Infertility patterns by user characteristics were similar across both methods, with higher prevalence in older, nulliparous, less educated, and poorer individuals.

Conclusions:

• Despite assumption violations, the CD approach, once adjusted, provides comparable infertility prevalence estimates and patterns to the prospective cohort design within the same sample.
• The CD approach can facilitate infertility research in populations lacking existing estimates, provided assumption violations are statistically addressed.
• App-based estimates may not represent US population-level infertility due to potential underreporting and sample characteristics.