Using Mobile Phone Location Data to Estimate the Association of Daytime Racial and Economic Segregation With Hypertensive Disorders of Pregnancy in Metro Atlanta

Background: Hypertensive disorders of pregnancy (HDPs) are a group of conditions that contribute to maternal morbidity and mortality, disproportionately affecting minority and low-income individuals. Complications from these HDPs may result in poor outcomes for mother and child during pregnancy or post-pregnancy. Disparities in HDPs may be associated with concentrated racial and income inequality in an individual’s activity space, such that the social exposure of places is not only formed by residence but also by daily mobility of the population at large. A novel aspect of women’s residential neighborhood environment can be explored by measuring daytime population mixing. Anonymized mobile phone data can be used to estimate social mixing during the daytime for income and racial/ethnic groups by describing the undercurrents of population mobility in neighborhoods. Objective: To estimate the association of daytime racial and income segregation with HDPs among women who gave birth from 2018 to 2019 in metro Atlanta. Methods: Data for women who gave birth in Metro Atlanta in 2018 and 2019 were abstracted from vital statistics records to determine HDP diagnosis and maternal residence by census tract. HDPs were grouped into pre-pregnancy hypertension, gestational hypertension, and eclampsia. Aggregated and anonymized mobile-phone data from opted-in devices were used to estimate median household income density in each census tract and racial/ethnic density, as a function of where devices moved between the hours of 7am and 10pm. Racial and economic composition was assigned by the device’s inferred home census tract. The Index of Concentration at Extremes (ICE) was calculated as the mixing of mobile devices in each census tract, by race and income. Tracts were categorized as concentrated disadvantage for an ICE score less than -0.3 and concentrated privileged for an ICE score greater than 0.3. Logistic regression models were used to estimate odds of HDPs in census tracts with micro-segregation and income inequality during the daytime, compared to not at extremes. At no point was mobility data linked to individual health records. Results: There were 122,482 births included in the maternal cohort, and 8.1% of births occurred among women with a diagnosed HDP. Black women had a higher rate of HDPs compared to White women (9.7% versus 6.9%). Census tracts with higher daytime micro-segregation were associated with higher odds of HDPs compared to tracts not in the extremes (OR = 1.42; 95% CI= 1.35, 1.49). Census tracts with higher daytime income inequality were also associated with higher odds of HDPs (1.26;1.20, 1.33). Conclusions: The odds of maternal HDPs were higher among women in census tracts with greater racial daytime segregation and more income disparity, compared to tracts that were not identified as extremes. Describing and distinguishing social epidemiologic patterns for HDPs by population mobility may provide a better understanding of the social context of who is most at risk.