Abortion access and state variation in observed unintended pregnancy.

OBJECTIVE: The state-level proportion of pregnancies that are unintended is an important social and public health indicator, and comparisons between states inform policy discussions. Unintended pregnancy is measured as a composite of abortions and unintended births, each of which is measured with error. We investigate whether between-state differences in abortion access and demand may bias comparisons between states' unintended pregnancy proportions when pregnancy intentions are misreported. STUDY
DESIGN: We algebraically specify the model currently used to estimate unintended pregnancy, extend it to include underreporting, and simulate the impact of underreporting on observed unintended pregnancy. Comparing the impact of underreporting across states, we identify levels of underreporting at which between-state comparisons are compromised.
RESULTS: We find that underreporting of unintended pregnancies could bias between-state comparisons when reporting of unintended pregnancies is less than 90-95%.
CONCLUSION: Current methods for estimating state-level unintended pregnancy proportions may underestimate unintended pregnancy to a greater degree in places with less abortion, and between-state comparisons may be biased. Estimates of state-level unintended pregnancy proportions would be more comparable if adjustment for completeness of retrospective underreporting were included in the estimation process. IMPLICATIONS: Estimates of unintended pregnancy should be adjusted for nonsampling error and include variances based on sampling and nonsampling error in order to permit robust comparisons between states, between populations, and across time. More research on the fidelity of retrospective reporting of pregnancy intention would facilitate this endeavor.