Controlled or Quasi-Experimental Estimates of Supervised Consumption Site Mortality

What the evidence says

The individual-level evidence on supervised consumption sites (SCS) comes from one quasi-experimental study (Lambdin 2022) using propensity score weighting to compare people who used an SCS against people who did not. That study found decreases across several health-service outcomes: people using the SCS were 27% less likely to visit an emergency department (95% CI: 12–46%) and had 54% fewer ED visits overall (95% CI: 33–71%). They were also 32% less likely to be hospitalized (95% CI: 4–57%) and spent 50% fewer nights in hospital (95% CI: 1–85%) — all four confidence intervals exclude zero. On overdose likelihood specifically, the same study found a lower likelihood among SCS users, but did not report a magnitude and the estimate did not clear the threshold conventionally used to rule out chance.

Population-level mortality evidence, reviewed by Gariépy (2025), tells a different story at the provincial scale. A controlled interrupted time series with propensity-score-matched comparison areas in British Columbia found no difference in monthly opioid-related mortality between local health areas that opened an SCS between 2015 and 2017 and matched controls (β = −0.08; 95% CI: −0.23 to 0.09). A synthetic-controls analysis in Ontario likewise found no measured association between SCS booth-hours and opioid-related mortality at the provincial aggregate level (β = 0.000; 95% CI: 0.000 to 0.000).

Below the provincial level, the same sources report decreases. The Ontario synthetic-controls model found decreases specifically in two public health units: London (β = −0.004; 95% CI: −0.006 to −0.002) and Thunder Bay (β = −0.004; 95% CI: −0.007 to −0.0002). An uncontrolled interrupted time series in Alberta found decreases in opioid-related deaths in Edmonton (−5.9 deaths/month; 95% CI: −8.9 to −2.9) and Calgary (−1.7 deaths/month; 95% CI: −4.5 to 0.9, an interval that includes zero), while Red Deer (−0.1 deaths/month) and Lethbridge (0.0 deaths/month) showed no change. These Alberta declines occurred while opioid-related deaths were rising province-wide. A separate uncontrolled comparison of Ontario public health units found SCS-implementing units held roughly stable mortality rates (+0.02 deaths/100,000/month, p = 0.27) while non-implementing units saw rates increase (+0.38 deaths/100,000/month, p < 0.001); the sources note this difference in trajectories was not directly tested statistically.

Two ecological studies add correlational evidence at even finer or coarser resolutions. Across Alberta's seven provincial SCSs, higher visit volume correlated with fewer fentanyl-related overdose deaths province-wide between 2017 and 2020 (r = −0.64, p = 0.03). In Toronto, neighbourhoods within 500 m of an SCS had 67% fewer overdose deaths per 100,000 people after implementation (p = 0.04).

No claim in this set links individual-level SCS use directly to fatal overdose; the individual-level study measures overdose likelihood, ED visits, and hospitalization, while the mortality estimates are drawn entirely from population-level, area-based comparisons.

Where the evidence splits

Controlled and quasi-experimental estimates of opioid-related mortality disagree by geographic scale. At the provincial aggregate level, two independent studies — a controlled interrupted time series with matched controls in British Columbia and a synthetic-controls model in Ontario — found no measured association between SCS implementation and mortality. At the sub-provincial level, using the same underlying Ontario synthetic-controls model, decreases appear in two specific public health units (London, Thunder Bay), and an uncontrolled Alberta time series finds decreases in two of four studied cities (Edmonton, Calgary) but no change in the other two (Red Deer, Lethbridge). The sources attribute part of this split to design: the provincial-level estimates use controlled or synthetic-control designs that account for broader mortality trends, while the local-level Alberta findings come from an uncontrolled design that cannot separate SCS effects from other concurrent changes. The sources also flag that the local London/Thunder Bay finding, being limited to two of several studied PHUs, raises questions about how far it generalizes — but they do not explain why some cities or PHUs show decreases while others in the same province do not.

Thresholds and dose-response

The Toronto ecological study establishes a distance-based dose-response pattern: within 500 m of an SCS, neighbourhoods measured 67% fewer overdose deaths per 100,000 people after implementation (p = 0.04); beyond 500 m, the reduction fell to 24% fewer deaths and the estimate no longer cleared the threshold used to rule out chance (p = 0.38).

Open empirical questions

• No source in this set links individual-level SCS use directly to fatal overdose; the individual-level study measures overdose likelihood, ED visits, and hospitalization, not deaths, and the mortality studies are population-level rather than individual-level.
• Why decreases appear at the city/PHU level (Edmonton, Calgary, London, Thunder Bay) but not at the provincial aggregate (British Columbia, Ontario) is not explained by the sources.
• Why Red Deer and Lethbridge showed no change in opioid-related deaths while Edmonton and Calgary showed decreases is not addressed.
• Whether the Ontario trajectory difference between SCS and non-SCS PHUs (stable versus increasing rates) would hold up under a formal statistical test is not established, since the source notes the comparison was not directly tested.