The goal of public health is to prevent disease, promote health, and prolong life through collaborative and cross-sector efforts of our society, organizations, communities, and individuals (DeSalvo et al., 2016; U.S. CDC, 2014). Public health practitioners struggle to develop interventions that are universally effective and there is no one-size-fits-all approach (DeSalvo et al., 2016; U.S. CDC, 2014). However, it is well known that our zip code is more predictive of our health than our genetic code (Artiga and Hinton, 2018; Ritchie, 2013). The zip code is related to social determinants of health (SDOH), which include but are not limited to economic and political structures, social and physical environments, and access to health services (Palmer et al., 2019). SDOH has been shown to be associated with major life events and health outcomes (Artiga and Hinton, 2018; Palmer et al., 2019). Examining patterns of SDOH and their association with a particular health event or outcome may allow public health practitioners to identify unique profiles of at-risk populations and neighborhoods, and then tailor intervention strategies accordingly (DeSalvo et al., 2016; Palmer et al., 2019; U.S. CDC, 2014). Considering limited resources, such strategies may also be more efficient by geographically defining the at-risk population and neighborhoods where people live, learn, work, play, and receive healthcare (Artiga and Hinton, 2018; DeSalvo et al., 2016; Palmer et al., 2019). Using teen pregnancy as an example, the authors illustrate how practitioners can use publicly available SDOH from the Census Bureau to identify distinct SDOH profiles at the census tract level.

A strategic priority of the Connecticut Department of Public Health is to promote health equity by reducing health disparities, particularly in maternal and child health. One of the primary objectives is to reduce unplanned teen pregnancies. The 2015–2019 U.S. Pregnancy Risk Assessment Monitoring System (PRAMS) data show that 75.5% of teen births (aged 15–19) were unintended meaning that the mother reported that at the time of conception she did not desire to have a baby (U.S. CDC, 2021, U.S. CDC, 2022). The proportion of induced terminations (abortions) in this age group is also suggestive of unintended pregnancy. During 2015–2019, abortions accounted for about 45% of all reported pregnancies (births, fetal deaths, abortions) among women 15–19 years in Connecticut. Together, these data estimate that 86.2% of teen pregnancies are unintended. Although the teen pregnancy rate, birth rate, and abortion rate have decreased over time, progress has been uneven and race and ethnicity disparity remains (Basch, 2011; Ngui et al., 2017; U.S. CDC, 2021). Additionally, research has shown geographic hotspots related to social determinants (e.g., high poverty, racial segregation) have consistently higher levels of unintended teen pregnancy (Blake and Bentov, 2001).

While it is the pregnancy that is not planned rather than the birth, access to pregnancy intention data is not available resulting in a dependency on teen birth data for developing public health strategies. Using teen birth rates to identify at-risk neighborhoods will not directly represent the teens at risk for pregnancy but rather those who delivered a live birth. The authors argue that there is value in assessing the SDOH patterns of teen births for two reasons. First, while teen births have been in steady decline for decades, the declines have been unequal among demographic groups meaning that tailored intervention strategies can directly benefit areas that retain higher teen birth rates. Second, the SDOH associated with teen birth can still be an adequately strong predictor of overall teen pregnancy risk and intervention strategies employed may help reduce the behaviors that lead to unintended pregnancy.

In recent years, latent class analysis (LCA) and geographic information systems (GIS) have been applied successfully to address public health problems (Arcaya et al., 2014; Blake and Bentov, 2001; Boscardin, 2012; Eastman et al., 2016; Neal et al., 2016; Taylor and Chavez, 2002; Xin et al., 2016). LCA has advantages over the traditional linear analyses, including assessing non-linear patterns or groups of indicators that may relate to health outcomes; simultaneously analyzing multiple measures even when they are highly correlated (Arcaya et al., 2014; Boscardin, 2012). Further, LCA can detect new areas and provide a fuller picture of most vulnerable neighborhoods by simplifying the patterns (Arcaya et al., 2014; Boscardin, 2012). Since teen birth rates often fluctuate due to small numbers, especially for small neighborhoods, LCA may avoid some of the limitations associated with direct rate comparisons (Backus and Mueller, 2021).

The authors use SDOH measures and LCA to identify neighborhoods most vulnerable to teen birth as a use case for how the approach may be adapted to other programs. The objectives of this study were to: 1) employ LCA to differentiate the Connecticut census tracts into distinct latent classes based on social determinants and profile neighborhoods for teen births as precursor to tailoring intervention strategies for unintended teen pregnancy; 2) utilize GIS mapping to visualize the geographic distribution of census tracts by latent class in Connecticut and overlay public middle school locations on the GIS map for the intervention program; and 3) assess whether latent classes were related to the outcome using GEE Poisson regression model.