Quantified Mortality Effect Size of Supervised Injection Sites and Naloxone Access

What the evidence says

The bystander-naloxone literature gives the most consistent quantified picture. Pooling four studies, bystander naloxone administration is associated with higher odds of overdose recovery (OR = 8.58, 95% CI 3.90–13.25), though heterogeneity between the pooled studies is high (I² = 92.09%). A component of that same evidence base reports raw counts: among witnessed overdoses, 0 of 39 people who received lay-dispensed naloxone died, versus 3 of 27 (11.1%) who did not receive it. At larger scale, take-home naloxone (THN) programs report a 96.3% successful reversal rate (2249 of 2336 administrations, 95% CI 95.5–97.1), with a fatal outcome still recorded in 0.8% of cases (1 in 123, 95% CI 0.4–1.2) even where naloxone was given. Separately, delaying enactment of a state naloxone access law by one year is associated with 1.51 more opioid overdose deaths per 100,000 people aged 12 and over in 2018, but the 95% confidence interval (−0.159 to 3.18) spans zero, so the estimate does not rule out no effect.

Training effects are the most precisely estimated outcome in this set: overdose-management training produces a standardized mean difference of 1.35 (95% CI 0.92–1.77) in knowledge test scores versus no training, with no heterogeneity across the five pooled studies (I² = 0.00%). This is a knowledge-outcome finding, not a mortality outcome.

At the community level, overdose mortality decreased in communities where THN was implemented compared to both pre-implementation rates and to communities without THN, though the magnitude of that decrease is not quantified in the source and rests on a single quasi-experimental study; other supporting evidence is observational. A descriptive account from Chicago found overdose deaths fell 20% in 2001, 8% in 2002, and 6% in 2003 versus prior-year rates after THN introduction in 2001, but this comparison had no control group.

For supervised consumption, the facility-level data come from one site (Vancouver's SIF, Milloy 2008): of 766,486 injections and 1004 overdose events recorded on-site, none resulted in death, and naloxone, a 911 call, or an ambulance was used in 453 of those events (45.1%). A Monte Carlo simulation modeling what would have happened absent the facility estimated 8 to 51 averted deaths over the study period, equal to 6.1% to 37.0% of the total overdose mortality burden in the surrounding neighborhood depending on assumptions about non-fatal overdose rates elsewhere (assumed to range 5–30% per year in the literature). The same source states explicitly that it is impossible to know with certainty whether these overdoses would have occurred, and been fatal, outside the facility. A separate review of medically supervised injecting rooms (MSIR) reports an association with decreases in preventable deaths, including management of almost 6000 overdose events, and with lower rates of naloxone administration and ambulance attendance — but without a quantified effect size, confidence interval, and drawn from small samples or limited ecological data.

What is largely absent from these sources is a quantified, controlled mortality effect size for supervised injection sites specifically: the on-site figure is a zero-death count within one facility, and the mortality-averted figures are simulation outputs with wide sensitivity to assumed non-fatal overdose rates, not a measured effect against a comparison group. An earlier narrative review states plainly that supervised injecting rooms had not been methodically evaluated for effectiveness in decreasing fatal and non-fatal heroin overdoses.

Where the evidence splits

Sporer (2006) states that none of the harm-reduction interventions reviewed, including supervised injecting rooms, had been methodically evaluated for effectiveness in decreasing fatal and non-fatal heroin overdoses. This sits against the later facility-level and review data — Milloy (2008), reporting zero deaths among 1004 on-site overdose events and a modeled 8–51 averted deaths, and Stewart (2025), reporting an association between MSIR use and decreased preventable deaths. The claims do not explain this gap beyond publication date: Sporer's assessment predates the outcomes data reported in the 2008 facility study and the 2025 review, and neither later source states that it resolves the earlier "not methodically evaluated" judgment. Whether facility-level counts and simulation-based estimates constitute the "methodical evaluation" Sporer found absent is not addressed by the sources on this page.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

• Whether overdoses occurring inside a supervised injection facility would have occurred, and been fatal, outside the facility is explicitly unmeasured in the one facility study here — the averted-deaths range (8–51) rests entirely on modeling assumptions, not observation.
• No source provides a magnitude or confidence interval for the association between MSIR use and decreased preventable deaths — only direction is reported.
• No controlled or quasi-experimental comparison links naloxone access law timing to overdose mortality with a confidence interval that excludes zero.
• Whether MSIR's association with lower naloxone administration reflects reduced overdose severity/need or reduced external service response is stated as unresolved in the source itself.