Reproductive
Fetal loss as a competing event in health policy studies: a case study using Texas’ 6-week abortion ban Simone Wien* Simone Wien Wien Wien Wien Wien Wien Wien Emory University
Policies may shift the burden of preterm birth towards fetal loss: as fetal loss increases, preterm birth (conditioned on surviving to delivery) decreases, making policies appear protective when they may be harmful. Collections of policy studies estimating effects on preterm birth may report heterogenous results due to differential susceptibility to this competing risk from fetal loss. Using Texas’ (TX) 2021 abortion ban as a case study, we compare conclusions when estimating preterm birth alone versus including proxies for fetal loss to evaluate whether accounting for this competing risk alters study conclusions.
Using vital records from 2014-2022 and augmented synthetic controls, we estimated the policy’s average treatment effect on the treated (ATT) for preterm birth and 3 proxies for fetal loss: stillbirth, live birth sex ratio, and live birth small-for-gestational-age sex ratio. Monthly estimates were calculated per 1,000 pregnancies or as ratios, adjusting for state- and time-varying policy and demographic factors. Sensitivity analyses considered states other than TX receiving the treatment during the same period and synthetic controls without augmentation.
Stillbirth and preterm birth increased in TX post-policy, but stillbirth results were more robust to sensitivity analyses. Stillbirths were higher than expected for months 2 (ATT 1.29, 95% CI 0.33, 1.89), and 8 (ATT 1.16, 95% CI 0.06, 1.83) post-policy; preterm births were higher than expected for months 6 (ATT 8.15, 95% CI 2.52, 16.11), and 7 (ATT 8.62, 95% CI 1.04, 16.19). No state diverged as much from the counterfactual in stillbirth as TX, but 6 states had larger divergences in preterm birth. Synthetic controls without augmentation had similar results; no change in sex ratios was observed.
Including stillbirth as a proxy for fetal loss may strengthen conclusions when studying preterm birth. Replication is needed to identify when this approach yields meaningful improvements in policy effect estimation.
