Background: Beginning in 2017, the Massachusetts Department of Public Health established a health equity workgroup within in the Office of Statistics and Evaluation (OSE) to build epidemiologist capacity to conduct racial and health equity work. The workgroup was charged with identifying best practices, identifying gaps in program work, and developing recommendations for tools and resources.

Objectives: Epidemiologists identified a need for guidance when framing and contextualizing data on a) health equity as a topic, b) populations that experience health inequities, and c) structural drivers of inequities. The workgroup chose to develop a style guide with the goal of providing example population-specific language and data elements for epidemiologists to use in dissemination products.

Methods: The style guide was organized by population including race/ethnicity, disability, sexual orientation, gender identity, and homelessness. Each population contains a group overview, recommendations for framing group identity, preferred terminology, available data sets containing related variables, and data source-specific terminology.

Results: The workgroup collaborated with partners across MDPH from data and content-specific fields to complete drafts of several guides. Important recommendations for changes in practice were identified (e.g. in the disability guide, the term “congenital differences” is preferred over “birth defects”).

Conclusions: Through the style guide development process, the workgroup learned invaluable lessons about using populations’ preferred terminology as part of epidemiologists’ work. Collaboration across content areas on dissemination products is key to ensuring accurate framing and continuously evaluating language around inequities. Next steps include evaluating current guides, adding populations, and ensuring they are applicable to broader audiences.

Within the field of social epidemiology, attention has been placed on the importance of social structures and systems (macro-level forces) rather than individual behaviors and other proximal risks. Macro-level forces can include economic conditions, early childhood family and educational context to political economic systems and globalization. To develop innovative research incorporating these new directions, a critical question is the following: What methods can be employed to measure complex, multi-faceted social forces that may change over time?

One particular multi-faceted force potentially shaping urban health and equity may be gentrification. We employ a random forest algorithm, a non-parametric tree-based approach using the R package ‘varSelRF’ to identify the variables that best predict a stage of gentrification for each census tract. These data were based on the stages of gentrification developed by UC Berkeley’s Center for Community Innovation (CCI). In this original data set 30 variables captured shifts in market conditions, vulnerable populations, and demographics.

This approach uses random bootstrap samples of the original data set to generate classification trees. When growing a tree, at the node of a tree Random Forest selects a random sub-set of variables, thereby decorrelating trees.

We reached 70% concordance with the CCI identified stages of gentrification for each tract. This is a solid concordance, given their measure accounts for Bay Area regional trends, thus obscuring smaller scale or county-level key attributes.

Based on this approach, the final indicators used were: 1. Adults (25+) with college degree; 2. Low-income households; 3. Median-income households; 4. Renters; 5. Pre-war units; 6. Change in median household income; 7. Change in low household income

These methods illustrate a data-driven approach to measuring key multi-dimensional exposures. This can increase scientific rigor and validity in future studies requiring novel measurement techniques. It can also decrease the risk of obscuring relationships between important social forces and health outcomes or mis-identifying key exposures.

Background: A longitudinal perspective of a Danish cohort (N=9125) followed over a 23 year period examined family's stability and structure, parental characteristics, household characteristics, and the child's social and academic competencies to determine key elements in contributing to adolescent and young adult criminal behavior.

Objectives: Develop multivariate and multidimensional categorical techniques to classify components of criminality based upon parental, child and familial contexts.

Methods: Classification using discriminant analysis techniques identified key correlates. Comparative log-linear analyses of paternal crime, descriptions of the families’ patterns of stability, and socioeconomic status changes over the life of the offspring then revealed predictive gender, family constellation and SES combinations at risk for criminal behavior.

Results: Stepwise discriminant analyses identified five key criminogenic aspects of the family atmosphere during childhood: 1) instability of adult constellations; 2) parent crime; 3) poor economic conditions; 4) work organization and 5) athletic skills of the child. Individual work organization was the single most important factor. Subsequent log-linear analyses reveal family instability at any age is significantly associated with crime, continuation in early adolescence it has the most damaging influence. Important contributors include fathers' criminal record, mothers' contentment, mothers' education, and mothers' efficiency. Academic competencies had significant additive effects, while athletic and work organization had key interactions; specifically, greater athletic skill and ability to work without structure positively “buffer" and prevent crime.

Conclusion: This study contributes effective clues to criminal prevention in unstable family environments and provided a longitudinal perspective on the linkage between the dysfunctional homes and subsequent crime.

background: People who inject drugs are at elevated risk for major bacterial infections (MBIs). Recent increases in MBI prevalence among younger populations highlight the need for well-targeted interventions.

objectives: To identify the geographic distribution and spatial correlates of MBIs in Massachusetts from 2011-2015.

methods: We obtained linked data, comprised of 16 administrative datasets, from the Massachusetts Department of Public Health, and community-level sociodemographic variables from the American Community Survey. We aggregated data by ZIP Codes (n=538) for analyses. Using GIS, spatial, and geostatistical analyses, we first identified significant MBI hotspots. Based on initial spatial analyses, we constructed multivariable and geographically weighted regression (GWR) models to determine factors associated with MBIs.

results: We found that ZIP Codes with lower MBI rates tended have populations with lower median age, poverty rates, fatal opioid overdoses, and inappropriately prescribed opioids (p<0.05). While we noted that ZIP Codes with better “access” to services, such as having shorter distances to pharmacies and highways, had higher MBIs (p<0.05), regional variation existed. Through GWR models, we found that being far away from highways in eastern Massachusetts and from pharmacies in western Massachusetts was significantly correlated with higher MBIs. Coefficients for socioeconomic variables and other risks also varied by region: Fatal overdoses were negatively associated with MBIs in western Massachusetts, and positively associated in northeastern and southeastern regions (p<0.05).

conclusion: Through novel GIS and geostatistical analyses, we identified varying spatial associations between MBIs and risk factors across Massachusetts. Results can inform geographically-focused targeting of MBI interventions.