Abortion-providing facility addresses were obtained from the 2018 Advancing New Standards in Reproductive Health (ANSIRH) facility database, which collected data in the same manner as previous databases. The database included 925 facilities operating in 2018. We excluded facilities that were not open (n = 83). We then abstracted the addresses of all Planned Parenthood health centers operating in 2018, both those providing abortion care and those that did not offer abortion care, from the Planned Parenthood Federation of America Web site (n = 600). Health centers that did not offer abortion care were included as sites to which TMAB could be expanded in the exposure scenarios described subsequently. After removing duplicate addresses (n = 351), we included 1091 facilities.
Abortion accessibility was operationalized as the proportion of US women 15 to 44 years of age who lived within a 30-, 60-, or 90-minute drive time of 1 or more abortion-providing facilities. The range of driving times considered here reflects differences in acceptable driving distances across the country owing to geography, rurality, or culture (e.g., abortion stigma or attitudes toward abortion). For example, 30 minutes is commonly used to define network adequacy for primary care, and 90 minutes may be more realistic for those living in rural settings or preferable for abortion seekers who prefer to travel further to protect their anonymity.
Initially, we geocoded facilities' addresses using the ggmap package in R (R Foundation, Vienna, Austria). Before ggmap geocoding, all "&" instances were removed from addresses to improve the process. All addresses were successfully geocoded. Latitudes and longitudes for all addresses not geocoded at the rooftop level and a random sample of 50 addresses geocoded at the rooftop level were manually checked with GoogleMaps, and the 3 inaccuracies (none of which were geocoded at the rooftop level) were corrected and geocoding was rerun.
To calculate abortion accessibility, we used R's osrm package to calculate 30-, 60-, and 90-minute drive time isochrones (i.e., polygons created by connecting all points along a road network that were a 30-minute drive time from that facility) for each facility. We assumed no barriers to interstate travel, so isochrones could cross state lines. To identify the population living within a given drive time to an abortion-providing facility, we used census block shape files and block-level population data from the 2010 US census obtained via the IPUMS (Integrated Public Use Microdata Series) National Historical Geographic Information System. Census blocks are the smallest geographic unit used by the US census (n = 11 078 297). For each census block, we calculated the number of women of reproductive age by summing the counts of women in age categories inclusive of the ages 15 through 44 years.
Using the sf package in R, we identified the intersection between census blocks and isochrones (i.e., the fraction of each census block's area included within the boundary). For each census block, we calculated the number of women 15 to 44 years of age who lived 30 minutes or less, 60 minutes or less, and 90 minutes or less from at least 1 abortion-providing facility, respectively, by multiplying the intersection fraction by the total number of women 15 to 44 years old residing in that census block (similar to the approach used by Pollini et al.).
We then determined the fraction of reproductive-aged US women in each state and in the United States overall who lived within the drive times of interest in each of the 3 scenarios by dividing the number of women 15 to 44 years old within the specified drive time across all census blocks in the area of interest (i.e., country or state) by the total number of women 15 to 44 years old in that area. Accessibility was defined as being within a set drive time of at least 1 abortion-providing clinic (regardless of whether cross-state travel was required). Thus, women who had access only to an abortion clinic in a neighboring state contributed data to the state where they lived.
The reference (unexposed) scenario was abortion accessibility based on the ANSIRH facility database and Planned Parenthood health centers that offered abortion care in 2018 (i.e., existing levels of provision; n = 850), referred to as the "current" scenario. We examined 2 exposure scenarios: (1) expansion of TMAB services to all Planned Parenthood health centers that did not offer abortion in 2018 in states where TMAB was legal (programmatic change; referred to as the "TMAB expansion" scenario) and (2) removal of all state-level TMAB bans (policy change; referred to as the "TMAB ban removal" scenario).
TMAB expansion assumes that, in 2018, TMAB was expanded in states where it was legal (i.e., the current level of provision along with Planned Parenthood health centers in states that did not have a TMAB ban), for a total of 996 abortion-providing facilities. TMAB ban removal assumes that, in 2018, state-level TMAB bans were removed so that in addition to the current level of provision, all Planned Parenthood health centers in all states offered TMAB (i.e., the TMAB expansion scenario along with all Planned Parenthood health centers in states with a TMAB ban), for a total of 1091 abortion-providing facilities (i.e., all facilities).
We used publicly available 2018 data on state TMAB bans from the Guttmacher Institute to identify facilities that met these criteria. In our analyses, we considered expansions of TMAB care only within the Planned Parenthood network given that site-to-site TMAB has been implemented within that network and there are many existing Planned Parenthood health centers that do not offer abortion care.
We calculated the differences in the proportions of US women 15 to 44 years old residing within 30-, 60-, and 90-minute drive times of an abortion-providing facility, respectively, between each exposure scenario and the reference scenario. To assess effect measure modification by population density, we calculated estimates stratified by census block urban versus rural status. All geographies were visualized and processed via the North America Albers Equal Area Conic projection, and analyses were conducted in R version 4.0.2.
Am J Public Health. 2022;112(8):1202-1211. © 2022 American Public Health Association