4. DISCUSSION
By using individual geocoded addresses of COVID-19 cases and deaths in MA, we were able to both evaluate community-level risk for these outcomes at high spatial resolution and distinguish institutional outcomes from those in non-institutional settings in community models of disease over time. Removing institutional cases from our models, especially in the context of mortality endpoints where institutional facilities contributed an appreciable percentage, allowed for a more nuanced understanding of local risk and disease drivers. Additionally, assessing trends over time across both case and mortality outcomes shed light on differential case fatality by subpopulation over time. Overall, our efforts highlight the value of collaboration between state public health departments and academic researchers to access, analyze, and interpret COVID-19 data to maximize its effective use in public health practice.
We observed key disparities in models of both cases and death outcomes associated with the proportion of Black and Latinx populations by tract, findings that parallel those at town-level resolution across a shorter time period as well as findings from other studies. The variability observed in the size of these estimates over time reinforces how these race and ethnicity variables are proxies for social constructs and not inherent or constant risk factors. Broader assessments of the structural factors that result in these disparities, including systemic racism, socioeconomic status, housing inequities, and multigenerational family stressors, are vital to comprehensively understand why these communities have experienced elevated COVID-19 risk. The variable IRRs over time could be coupled with other community-level information to better assess the particular factors driving disparities in each phase of the pandemic.
The finding that the association with % Black residents was smaller during Phase 2 than Phase 1 may indicate that public health policies and other measures enacted by Fall 2020 among communities with higher proportions of Black residents successfully reduced risk relative to other communities. This could include greater availability of testing than existed in the early months of the pandemic. Notably, we did not see as substantial of a reduction for communities with higher proportions of Latinx residents, which points toward the need for a closer look at testing as well as structural barriers to implementing risk-reduction methods (such as ability to self-isolate, work from home, or physically distance) across these communities.
Our study period captured only the early months of vaccination availability in the state, reflecting availability for healthcare workers (beginning on December 15, 2020), residents of long-term care facilities (December 28, 2020), and all individuals aged 75 and over (February 1, 2021). It is possible that our findings related to racial/ethnic patterns for Phase 2 reflect, in part, differential vaccination patterns from early 2021; however, this is unlikely to have a large influence on our overall findings, given the limited number of individuals fully vaccinated by the end of the time period assessed here. Given that LTCF residents were eligible for vaccination for several weeks of the study period, it is plausible that our estimates for the LTCF beds per capita variable is an underestimate of the true association.
Case risk associated with proportion essential workers by tract remained elevated across both time periods in our study. This finding may reflect the limitations of workplace interventions to reduce exposure risk among this workforce, including challenges to physical distancing in essential service jobs, and lack of paid sick leave. Our study period did not overlap with vaccine availability targeting essential workers outside of healthcare settings (March 22, 2021); analyses utilizing more recent data could consider the effect of targeted vaccination on these estimates by tract. The elevated mortality risk observed in association with proportion essential workers by tract, especially in Phase 2, may suggest that communities with more essential workers faced a higher case fatality rate, although it is difficult to conclude with certainty from our data.
One important aspect of modeling cases and deaths separately across pandemic phases is the ability to effectively control for differential testing availability and utilization. This was particularly salient during Phase 2, which saw increased availability of testing to the public and widespread asymptomatic surveillance testing at workplaces and schools. Since testing is not uniformly distributed across all census tracts, IRRs are likely higher in places with more testing, irrespective of case severity. By contrast, COVID-19 deaths in Phase 2 are expected to have much more consistent identification, making comparisons less susceptible to testing biases. The greater mortality risk for Black/Latinx populations, essential workers, and older residents that we observed in Phase 2 may thus provide a more accurate depiction of racial and ethnic inequities in COVID-19 impacts than case rates alone, which could potentially be more useful for informing targeted public health interventions.
By retaining LTCF facilities in non-institutional models, we attempted to estimate the impact of LTCF cases on disease incidence outside of these facilities. Our data indicate that density of long-term care services was not a major correlate of infections or deaths in the community at large (among individuals who were not residing in these institutions; section 3.2). This observation may indicate successful biocontainment within these facilities or limited interaction between residents/employees and community members. However, we lack key information to determine this definitively, including controls within facilities and residential location of LTCF employees. It also may indicate that risk factors for institutional cases and deaths differed in meaningful ways from what was observed at the community level. While beyond the scope of this study, considerations of risk factors for mortality specifically within institutions, and comparisons between these within-institution and community-level risk factors, would be a valuable addition to this literature.
Our analysis is limited by a few key factors. First, census tracts are heterogenous and not distinctly classified by the variables in our models, which complicates the interpretation of our findings. We identified tracts using continuous demographic and occupational characteristics, and the same town may have elevated proportions of some, but not all, of the covariates in our model. As such, our analyses can serve as a guide for understanding differential risk by population subgroups but not to identify specific tracts to target with public health interventions, as would be possible with spatial methods. Additionally, as mentioned previously, limited availability of testing during Phase 1 resulted in testing and diagnosis of only symptomatic cases early in the pandemic, while testing was widely available in Phase 2, both for symptomatic cases and asymptomatic surveillance. It is difficult to concretely assess the directionality of these biases, but these trends may indicate that our data from Phase 1 reflects an underestimate of the true association with the covariate analyzed. In addition, the variables we included might not be all or the strongest predictors of cases or deaths. Another limitation is that patient address information likely contains errors, which cascade into the geocoding process, resulting in misclassification of LTCF residents and tract of residence; however, this is likely non-differential with respect to outcome and, moreover, there is no feasible way to ameliorate this type of error. Our age-related covariates may imprecisely classify risk associated with age. Finally, ACS data were derived pre-pandemic and may not fully reflect conditions during the pandemic, especially with respect to employment and housing.
Our project is strengthened by our geocoded individual-level data, which was facilitated by a cross-sectoral partnership. Using address geocodes of individual patients who were diagnosed or died of COVID-19, we observed significant disparities in both case and mortality burden in association with population proportions of Black and Latinx residents, as well as essential workers at the census tract level in the first year of the pandemic in Massachusetts. Modeling cases and deaths separately, as well as with and without institutional outcomes, allowed us to more comprehensively understand health disparities experienced by vulnerable subgroups during the first year of the pandemic.
DISCLOSURES : The authors have no conflicts of interest to disclose.
AUTHOR CONTRIBUTIONS: JIL, KJP, MPF and JHL conceptualized the project. KRS and XP conducted statistical analyses. PP, KRS, KFT, JIL, EAE, KJL and JHL guided data analysis and data interpretation, with support from all authors. FC, TST and KJL conducted and supervised geocoding. RMK and TST contributed individual-level data from DPH and guided its use and interpretation. JHL and KRS wrote the paper with contributions from all authors.
FUNDING SOURCES: This project is supported by NIH/NIMHD grant P50MD010428 (PIs: JIL and F. Laden) and by an unrestricted gift from Google LLC to support research on COVID-19 and health equity (PI: JIL).
Table 1. Descriptive statistics by phase of COVID-19 cases, deaths, and average census tract characteristics in Massachusetts, USA. 1Reflects the number of cases and deaths that were successfully geocoded to a census tract (total number of individuals excluded due to lack of geocoding is 1,360 [0.27%]).