Health Disparities
Defining policy exposures when differential enforcement occurs: a case study using Tennessee’s prenatal substance use law Simone Wien* Simone Wien Wien Wien Wien Wien Wien Wien Wien Emory University
The consistency assumption for causal inference of policies depends on uniform policy implementation across social groups. This assumption is frequently violated and can result in misleading policy estimates. In the setting of differential enforcement, non-stratified estimates can mask harmful or beneficial policy effects within sub-populations due to aggregation. We use Tennessee’s (TN) 2014 law criminalizing prenatal substance use (SB1391) as a case study: while disproportionate prosecution of Black women and women in Northeast TN was documented, studies reported only statewide estimates.
We estimated SB1391’s effect on preterm birth, entry to prenatal care, and in-hospital delivery using vital records from 2005-2020. We estimated the stratified race/ethnicity average treatment effect on the treated (ATT) using augmented synthetic controls (comparing TN to a synthetic control) and interrupted time series for regional analyses (comparing Northeast TN to the rest of the state). Sensitivity analyses examined robustness to missing prenatal care data in TN during policy implementation. Regional analyses are ongoing.
Statewide estimates obscured racial differences in decreased prenatal care. In the law’s last quarter, Black women experienced greater decreases (ATT -151 per 1,000 births, 95% CI -223, -81), than the total population (ATT -98, 95% CI -161, -46) and White women (ATT -72, 95% CI -130, -31). Prenatal care returned to predicted levels for all groups once the law ended. Northeast TN had a modest decrease of in-hospital deliveries (differential slope change -0.95 per 1,000 pregnancies per quarter, 95% CI -1.90, -0.01) and preterm births (- 0.33, 95% CI -1.90, -0.01). Black prenatal care estimates remained robust assuming 50% of missing cases received care, while White and total estimates were robust at 25%.
Stratified estimates support evidence of differential implementation of SB1391. Analytical plans for social policy estimation should consider stratification.
