A public map of the 2026 Bundibugyo virus outbreak
An interactive record of how Bundibugyo virus disease spread in 2026, limited to claims a person has approved.
Cases, contacts, and sources on one map
Research
Cannabis retail access and binge drinking in Baltimore
A cross-sectional ecological study of 194 Baltimore City census tracts, asking why medical cannabis dispensaries and high binge drinking prevalence cluster in the same neighbourhoods, as a baseline for adult-use commercialisation.
- Study field
- Public health
- Discipline
- Research
- Year
- 2026
- Duration
- Poster 1 May 2026
The challenge
Cannabis and alcohol may act as substitutes or as complements. At the neighbourhood level, proximity to a dispensary can serve as a spatial proxy for exposure to the cannabis market. Maryland legalised adult-use cannabis in July 2023, and where retail outlets are sited is an environmental justice concern.
In Baltimore City, dispensaries and high binge drinking prevalence appear in the same census tracts. Why they cluster together had not been examined at the tract level before commercialisation. The study describes that co-clustering, to set a baseline for monitoring spatial equity after the adult-use rollout.
What we did
The study is a cross-sectional ecological analysis of 194 of Baltimore City’s 200 census tracts. Six were excluded because PLACES estimates or covariate data were missing.
The outcome was tract-level binge drinking prevalence from CDC PLACES 2025, modelled from BRFSS 2022–2023 by multilevel regression and poststratification (MRP). The exposure was a Gaussian distance-decay score for proximity to the nearest licensed medical dispensary, with a primary bandwidth of 1.0 km and sensitivity bandwidths of 0.5 and 2.0 km.
A directed acyclic graph set out the hypothesised relationships between dispensary proximity, binge drinking, and structural covariates. In it, structural racism (historical redlining) and commercial zoning are unmeasured. Racialised economic segregation, socioeconomic status, population density, housing burden, and liquor outlet density are adjusted for.
Sequential OLS models with robust standard errors added the covariates one group at a time:
- M0: unadjusted.
- M1: racialised economic segregation, measured by the Index of Concentration at the Extremes for White-Black economic segregation (ICE: affluent White residents minus poor Black residents, divided by the total population).
- M2: socioeconomic status (poverty, unemployment, education).
- M3a: demographics not used in the PLACES model (population density, share aged 18–34).
- M3b: demographics that PLACES also uses in poststratification (median age, male share). Comparing M3a with M3b tests whether the modelled outcome makes the adjustment circular.
- M4: housing (severe rent burden, renter-occupied housing).
- M5: liquor outlet density.
Spatial autocorrelation in the residuals was checked with Moran’s I, and spatial error and spatial lag models were fitted as sensitivity analyses.
Data came from CDC PLACES 2025, the American Community Survey 2019–2023, the Maryland Cannabis Administration (dispensaries), and Open Baltimore (liquor licences). Matt, Diaphana’s founder, is the poster’s sole author.
Interpretation#
The poster reads the pattern as shared neighbourhood disadvantage rather than an independent effect of dispensaries. Upstream structural factors outweigh the nearer environmental exposure, which is consistent with Fundamental Cause Theory. ICE serves as a contemporary proxy for the segregation left by Baltimore’s history of redlining. The finding aligns with environmental justice research on the siting of alcohol and cannabis outlets in racialised neighbourhoods, and it points to equity-centred approaches to where cannabis retail is sited.
Limitations#
- Ecological design. The unit of analysis is the census tract. The results describe tracts, not individuals, and do not show that living near a dispensary changes anyone’s drinking.
- Unmeasured confounding. Factors outside the models, such as commercial zoning, may still confound the association. Zoning is drawn in the DAG but could not be measured.
- Proxy exposure. Proximity to the nearest medical dispensary stands in for exposure to the cannabis market. It does not measure cannabis purchase or use.
- Modelled outcome. Binge drinking prevalence is a model-based estimate, not a direct survey count for each tract. Reintroducing the covariates used in that model changed the coefficient by less than 1.2%, so outcome measurement is unlikely to explain the attenuation.
Future directions#
- A baseline for evaluation. The study records the pattern largely as it stood before Maryland’s adult-use market opened in July 2023, so that later data can be compared against it. Some of the source data, such as BRFSS 2022–2023 and ACS 2019–2023, extend past that date.
- Quasi-experimental follow-up. The poster proposes analyses such as a difference-in-differences design, comparing tracts with high and low dispensary access before and after commercialisation. They would test whether the gap in binge drinking prevalence widens or narrows, that is, whether commercialisation makes spatial inequities worse.
- Policy. The results are intended to inform zoning policy centred on environmental justice, and the siting of cannabis retail outlets.
Results
- Unadjusted, dispensary proximity was positively associated with binge drinking prevalence (M0: β = 7.62, 95% CI 5.81 to 9.42, p < 0.001)
- After full adjustment the coefficient fell by 94.9% (M5: β = 0.39, 95% CI −0.52 to 1.29, p = 0.40)
- The largest single step was the entry of ICE, which reduced the coefficient by 66.0%
- Results held across KDE-based and 0.5 to 2.0 km bandwidth specifications, and in spatial error and spatial lag models (β ≈ 0.41)
- No residual spatial autocorrelation remained (OLS residual Moran's I = 0.012, p = 0.34)
- Reintroducing the MRP poststratification covariates changed the coefficient by less than 1.2%
Contact
Write to Diaphana
For questions about our work, or about working with us, write to [email protected].