Relationship of political ideology of US federal and state elected officials and key COVID pandemic outcomes following vaccine rollout to adults: April 2021–March 2022

Addressing spatial misalignment in population health research: a case study of US congressional district political metrics and county health data

“…To compare the ABRM results to those that would be obtained using standard dasymetric realignment approaches (as described in Section 1), we conduct two additional analyses: a CD-level analysis and a county-level analysis. First, following Krieger et al 5 , we conduct a CD-level analysis in which we (1) use a dasymetric approach to realign county COVID-19 mortality rates to the CD level; (2) collect all adjustment variables at the CD level from the 2016-2020 ACS and 2020 census redistricting files; and (3) fit a Poisson regression model to the CD-level measures. Note that, while this analysis is similar to that of Krieger et al 5 , we use a smaller set of adjustment variables and more updated data sources that were not available at the time of their analyses (i.e., 2020 decennial census data).…”

“…As in Krieger et al 5 , the aim of this case study is to assess associations, not to make causal inference. However, we wish to adjust for several well-established community-level risk factors for COVID-19, and in this study use several social and economic variables also included in the Krieger et al study 5 .…”

“…As in Krieger et al 5 , the aim of this case study is to assess associations, not to make causal inference. However, we wish to adjust for several well-established community-level risk factors for COVID-19, and in this study use several social and economic variables also included in the Krieger et al study 5 . In particular, community impoverishment is an important adjustment variable in our analyses, which we measure using the percent of residents in poverty extracted from the 2016-2020 5-year ACS at the county level.…”

“…Areal spatial misalignment (hereafter simply spatial misalignment for brevity) occurs when areal data on multiple variables are collected using mismatched boundary definitions, and is a common obstacle to data harmonization and analysis across a range of fields. In political metric and health studies, political metric data may inherently arise at the CD level, e.g., representatives' voting patterns, while the most granular health metric data publicly accessible are often aggregated to the county-level (as in Krieger et al 5 ). Analyses seeking to investigate associations in CD and county-level features must confront a complex bi-directional spatial misalignment scenario, as counties can intersect CDs, be fully nested within CDs, or fully contain one or more CDs.…”

“…Dasymetric methods represent one approach to realignment, where a strong and often unrealistic assumption of no within-area variability is imposed for each measure, and values are reapportioned to the target units via population-weighting. Dasymetric methods have previously been used to realign data in the context of political metrics and health research 5 . This type of apportionment-based realignment is deterministic and can introduce measurement error and confounding that bias effect estimates of interest 15,16 .…”