Abstract
OBJECTIVE: Our study examines the association between social vulnerability index (SVI) and pharmacotherapy initiation for gestational diabetes mellitus (GDM). METHODS: We studied a retrospective cohort of pregnant patients with GDM, enrolled in Tennessee Medicaid, who gave birth between 2007 and 2019. Enrollment files were linked to birth and death certificates, state hospitalization registries, and pharmacy claims. SVI, measured at the community level and determined by residential census tract, ranged from 0 to 100 (low to high vulnerability). Multivariable logistic regression assessed the association between SVI and the odds of initiating the most common pharmacotherapies for GDM-insulin, glyburide, or metformin-and adjusted for relevant covariates. SVI was modeled with restricted cubic splines to account for nonlinear associations, using the median Tennessee SVI as a reference. Secondary analysis assessed associations with the SVI subthemes. RESULTS: Among 33,291 patients with GDM, 21.7% (7,209) initiated pharmacotherapy during pregnancy. Patients from areas with higher SVI were more likely to be non-Hispanic Black with higher body mass index, whereas those with lower SVI were more likely to be nulliparous. Multivariable modeling demonstrated a complex nonlinear association between SVI and GDM pharmacotherapy initiation, relative to the reference. Higher SVI was associated with elevated odds of GDM pharmacotherapy initiation (e.g., odds ratio 1.11 [95% confidence interval 1.02-1.22] for SVI 80) and low to medium SVI had variable nonsignificant associations with GDM pharmacotherapy initiation, relative to the reference (lower odds of initiation for values 25-50, higher odds of initiation for values < 25). Secondary analysis demonstrated a nonlinear association between subtheme 3 and the odds of GDM pharmacotherapy initiation. CONCLUSION: Social vulnerability is associated with initiation of pharmacotherapy for GDM, highlighting the possible role of social determinants of health in achieving glycemic control.