Table 4 Recommendations for defining infertility in analyses of household survey data based on sensitivity analyses of key components of the infertility definition

One- to two-year exposure periods increase misclassification of fertile unions as infertile; a five-year exposure period is recommended.

Measuring time since first union as a proxy for couple status results in an acceptable error (< 5%) for prevalence estimates, even for women with multiple unions. Temporary separations have little effect on infertility estimates

Current contraception is a sufficient proxy for contraceptive use over the exposure period for women over 30.

Current contraception is not a sufficient proxy for contraceptive use over the exposure period.

Intent has a small influence on prevalence estimates in the surveys analyzed, although this may not be true for high-income settings.

Disregarding intent increases estimates of infertility. Taking into account intent is recommended when measuring the disability of secondary infertility.

Using reported birth is recommended as it is a more reliable measure than reported pregnancies. When using birth as an outcome, some women classified as infertile are pregnant at the time of the survey. The proportion of infertile women who are pregnant is smaller if longer exposure times are used.