Do Harm Reduction Vending Machines Work? What the Evidence Shows

Do harm reduction vending machines work? For access, the evidence is consistent: they get naloxone, syringes and test strips to more people, mostly at hours when services are shut, and the people they serve accept them. For health outcomes such as overdose deaths and infections, the evidence is encouraging but observational and thin. This page sets out what the reviews and key studies found, what they do not show, and how a program can design an evaluation that adds to the evidence instead of only counting items. Facts are stated as of October 2026.

Quick answer

• Reviews: a 2025 systematic review of 30 studies (191,242 participants) found high demand, out-of-hours use, acceptance and reach into higher-risk groups. Impact evidence was limited.
• Overdose deaths: Clark County, Nevada saw 229 opioid overdose deaths in the year after naloxone went into machines, against 270 forecast. An association, not proof.
• Out-of-hours reach: 69% of Melbourne syringe machine use fell outside service hours; nearly half of New York State machine items were dispensed after hours.
• Distribution: US programs with syringe machines distributed syringes at a 28% higher rate than those without (RTI/CDC, 2025).
• Preference: rural Kentucky drug users rated machines more acceptable than syringe services programs (4.49 vs 4.17 out of 5).
• Gap: no randomised trials; most studies count dispensing, not use or health outcomes.

What “working” means for a vending machine

Studies measure different things, and the word “works” hides the difference. It helps to separate four levels, from easiest to hardest to prove:

Level	Question	Typical evidence	Strength today
Feasibility	Can a machine run, stay stocked and be used?	Dispense counts, downtime, implementation interviews	Strong
Reach and acceptability	Who uses it, when, and do they find it acceptable?	Time-stamped logs, surveys, focus groups	Strong
Behaviour	Does it change sharing, testing or carrying naloxone?	Self-report surveys, before-and-after comparisons	Moderate, few studies
Health outcomes	Fewer overdose deaths, HIV or hepatitis C?	Time-series on death or infection data	Limited, observational

Most published work sits on the first two rows. That is not a weakness of the machines; it reflects how hard the bottom row is to measure for any community intervention.

What the systematic and scoping reviews found

Zhang et al., 2025 (systematic review, Harm Reduction Journal). The most complete synthesis to date searched four databases to 29 November 2023 and found 45 articles covering 30 studies and 191,242 participants. Twenty studies were outside the US; 18 focused on people who inject drugs; 12 evaluated syringe-dispensing machines. Findings:

• high demand, with “usage mostly occurring outside of traditional business hours”;
• in some comparisons, more syringes and HIV self-tests dispensed than through in-person programs;
• general acceptance by target populations, regardless of the item, and reach into high-risk groups;
• impact: seven articles on syringe machines described less syringe sharing (four) and less drug use (two), and stable or declining drug-related crime (one); HIV self-test machines found detection rates of 1.9% to 17.7%; two articles reported fewer fatal overdoses after naloxone machines.

The authors call impact evaluation “limited” and ask for implementation research on community health outcomes, including overdose.

Russell et al., 2023 (scoping review, 22 studies). Effectiveness results were “mixed between clear effectiveness and inconclusive results”, with positive outcomes from after-hours availability and increased reach. None of the studies reported on race. The review’s best-practice list is a design brief: access up to 24 hours a day, syringe disposal options, the capability to collect data, and anonymity of use.

The overdose studies

Clark County, Nevada (Allen et al., 2022). Trac-B Exchange, a syringe services program, installed three public health vending machines in 2017 and added naloxone in 2019. Researchers modelled monthly opioid-involved overdose deaths among county residents from 2015 to 2020. In the 12 months after naloxone dispensing began, the model forecast 270 deaths; death certificates showed 229, suggesting 41 deaths averted, about 15%. They found a significant immediate drop at launch, followed by a rising trend that ran into the COVID-19 period. The study shows association: it cannot tell whether the naloxone used in a reversal came from a machine, and other prevention work was running at the time.

Ohio (Arendt, 2023). An outdoor machine at the site of a weekly syringe services program, with registration by phone, dispensed 3,360 naloxone doses and 10,155 fentanyl test strips in its first year, more than any other syringe program in the county. Of 637 registrants, 12% had not used harm reduction services before. The machine’s first year was associated with a lower countywide rate of unintentional overdose death and HIV; the author wrote that the association “should be further investigated to assess causality”. A Johns Hopkins summary for opioid settlement planners describes this as a 10% fall in fatal overdoses.

Out-of-hours use and reach

The most consistent finding across 30 years and three continents is timing. Machines are used when services are closed.

• Melbourne (Kerr et al., 2022). Four syringe dispensing machines made 180,989 orders to an estimated 90,488 unique presentations from May 2017 to December 2020; 69% occurred outside the managing needle and syringe program’s hours, and the machines distributed 66% as many syringes as the fixed site.
• New York State (University at Buffalo, 2026). Fifteen anonymous machines dispensed 13,655 items in 14 months, with nearly half outside business hours.
• Sydney (Islam et al., 2008). 50.9% of 167 surveyed users used machines only between 5 pm and 9 am. Younger users preferred machines and cited stigma at staffed outlets.

Volume follows. In the 2023 National Survey of Syringe Services Programs, 16.8% of 529 US programs offered syringes, naloxone or both through machines. After adjusting for other characteristics, programs with syringe machines distributed syringes at a 28% higher rate (p=0.02), and those with naloxone machines distributed naloxone at a 12% higher rate (p=0.08, not statistically significant at the usual threshold), according to an RTI and CDC brief.

Who they reach, and what people want from them

Kentucky preference research. Knudsen, Havens and Young (2025) surveyed 712 people who use drugs in two rural Appalachian counties. Both models scored well, but machines were rated more acceptable and appropriate than syringe services programs (4.49 vs 4.17 on a 5-point scale, p<.001). Preference for machines was strongest among people who do not inject and among people who inject and share syringes without using a program, the groups a staffed service struggles to reach.

Design preferences. In companion focus groups (Marschke et al., 2025), participants ranked access codes above cards or tokens, which can be lost, stolen or traded, and well above fingerprints, which they feared police could access. They wanted a discreet, 24/7 machine with obscured glass and a self-directed callback option for linkage to services.

Acceptability to hosts and neighbours. Host organisations in New York State saw “few challenges” with hosting a machine. In six California veterans’ supportive housing buildings, 60 residents and staff described expanded access and “no perceived increase in visible disorder or safety concerns”, alongside concerns about stigma, privacy, enabling and placement near children (Rife-Pennington et al., 2026). Implementers interviewed across the US reported that placement is “most often determined by feasibility and willingness of host sites” and that maintenance and stocking costs were unanticipated (Rapisarda et al., 2026).

Machines and face-to-face services. In inner Sydney during COVID-19 restrictions, dispensing machine equipment rose 41.1% while face-to-face program visits fell 16.2%, and visits to the co-located primary healthcare clinic rose 59.7% a month (Uthurralt et al., 2022). The authors concluded 24-hour machine access did not reduce targeted primary healthcare use.

Table of studies

Study	Setting	Design	Key finding
Zhang et al., Harm Reduct J 2025	International (20 of 30 studies outside the US)	Systematic review, 30 studies, 191,242 participants	High demand, out-of-hours use, acceptance, reach; 2 articles reported fewer fatal overdoses after naloxone machines
Russell et al., Harm Reduct J 2023	International	Scoping review, 22 studies	Effectiveness mixed or inconclusive; positive after-hours and reach effects; recommends 24/7, anonymity, data capability
Allen et al., Ann Med 2022	Clark County, Nevada	Interrupted time series, overdose deaths 2015–2020	229 deaths vs 270 forecast in the 12 months after naloxone machines; significant immediate drop
Arendt, J Am Pharm Assoc 2023	Ohio, one outdoor machine	Before-and-after program evaluation	10,155 FTS and 3,360 naloxone doses in year one; 12% of 637 registrants new to harm reduction
Kerr et al., Harm Reduct J 2022	Melbourne, 4 syringe machines	Analysis of machine order and keypad data, 2017–2020	69% of presentations outside program hours; keypad demographic data largely invalid
Uthurralt et al., ANZJPH 2022	Inner Sydney, 24-hour machine	Service data comparison, 2020–21 vs prior year	Machine equipment +41.1%; clinic visits +59.7% a month
Islam et al., Drug Alcohol Rev 2008	Sydney	User survey, n=167	50.9% used machines only 5 pm–9 am; 32.8% reported broken or jammed machines
Philbrick et al. (RTI/CDC) 2025	US, national survey plus 12 programs	Survey analysis and interviews	Programs with syringe machines distributed syringes at a 28% higher rate
Knudsen et al., J Stud Alcohol Drugs 2025	Rural Kentucky	Survey, n=712	Machines rated more acceptable than syringe programs (4.49 vs 4.17)
Lynch, Vest et al. (University at Buffalo) 2025–26	New York State, 15 machines	Utilisation data and host interviews	13,655 items in 14 months; nearly half after hours; few hosting challenges
Zhang et al., J Subst Use Addict Treat 2026	Central Pennsylvania, 2 touchscreen machines	Utilisation data, May 2024–May 2025	11,327 items to 2,321 registered clients, plus 4,472 non-registered browsing sessions
Rife-Pennington et al., JAMA Netw Open 2026	6 veterans’ supportive housing buildings, California	Qualitative, 60 interviews	Acceptable; no perceived increase in visible disorder

What the evidence does not show

• Cause and effect on health outcomes. There are no randomised trials. The overdose findings come from time-series studies in single counties, where other interventions and drug-supply changes run in parallel.
• Use after dispensing. A count of test strips or naloxone kits shows that items left the machine. It does not show that a strip was used, a kit carried or an overdose reversed.
• Who the users are. Anonymity is the point, and it limits data. In Melbourne, so much keypad-entered age, gender and postcode data was invalid that only 6% of presentations could be analysed; the authors judged that method “not feasible” for community programs without changes.
• Equity. The 2023 scoping review found no included study reported on race.
• Cost-effectiveness. A 2025 simulation model (Zafarnejad et al.) suggests machines can reduce disability-adjusted life years and societal costs, but that is a model, not observed data.
• Reliability effects. In Sydney, 32.8% of surveyed users reported machines broken or jammed. Uptime is part of the intervention, so report it alongside dispensing.

How to design your program’s own evaluation

The literature grows when programs collect the right data from the start. Plan the evaluation with the machine, not after it.

Evaluation question	Data source	How the machine helps
How much reached the community, when and where?	Dispense log	Every vend records product, quantity, machine, location and time
Was the machine available?	Fault, stock-out and downtime records	Live stock levels with low-stock alerts; fault reports by machine ID
Are we reaching new people?	One optional question: first supply or refill?	Asked on the phone or optional touchscreen before release
Are supplies used?	Optional question: did you use your last kit or strips?	Same; answers stored without identity
Where do users come from?	Optional ZIP code or postcode	One short numeric field, not three
Unique users and repeat use	A code issued after short registration on a phone	Per-code counts and quotas tracked in the cloud
Health outcomes	Health department overdose and infection data	Not from the machine: agree access before launch
Experience and community effects	Interviews with users, host sites and neighbours	Not from the machine

Five rules from the studies above:

1. Ask but never block. Questions should be optional and skippable. NSW needle and syringe program policy says participation in research and evaluation surveys should rest on the client’s “informed and voluntary consent”, and services should avoid “unwanted educational or referral interventions which may discourage future access”.
2. Ask few, simple questions. The Melbourne keypad asked for coded age, gender and postcode before ordering, and most of the data was unusable. One or two questions on a phone screen or touchscreen, with plain answer buttons, is the realistic limit.
3. Keep personal data off the machine. The machine should see only a code; any registration detail sits with the registration service, under a privacy notice and retention period.
4. Get a baseline. Record fixed-site distribution and local overdose figures for at least the 12 months before launch; the Nevada and Sydney studies both depended on pre-launch data.
5. Partner with a research team early. The strongest US evaluations (Nevada, Pennsylvania, New York State, Kentucky) were run with universities, and the RTI brief notes programs that received data support from research partners.

The question design itself — what to ask, how to word it, phone versus touchscreen, what not to ask — is covered in on-screen questions and forms on harm reduction vending machines.

What KioskForce builds for evaluation

On our machines each dispense writes an order record with items, quantities, machine, location, time, order state and receipt ID. Short questions can be answered on the user’s phone or on an optional 7- to 32-inch touchscreen before release, and the cloud dashboard shows live stock with low-stock alerts. Optional temperature monitoring can be logged alongside dispense data. Our compact wall-mount machines can also take an optional millimetre-wave radar traffic counter, in technical preview, that counts people passing without capturing images.

The HIVConnect and MyTest programs use this pattern: a few questions on the user’s phone, a code released at the machine, a per-person quota and, on HIVConnect, an optional follow-up survey. We have not supplied a naloxone, syringe or test-strip machine; the same controls and data are what those programs need to evaluate their work. We build to the data specification you set and do not design or certify the evaluation itself.

What to specify if you are buying one

1. The evaluation questions your funder will ask, written before the hardware is chosen.
2. The data fields: dispense record, optional questions and their wording, and any code or registration step.
3. Where data lives and for how long, agreed with your privacy officer, with nothing personal on the machine.
4. Uptime and stock reporting you need, and who receives the alerts.
5. The site and hours, since out-of-hours access is the effect most studies measure.

Hardware starts at about US$800–1,500 for a compact wall-mount unit; a published code-and-quota software project was US$11,000 plus about US$200 a month. See harm reduction vending machine cost and the vending machine cost guide, or send your brief through the contact page. We sell direct and, in some regions, through local distributors; support is by email.

Where a machine is the wrong answer

• When the outcome you need is engagement. Treatment referral, wound care and testing conversations happen face to face; a machine can extend hours, not replace the conversation.
• When nobody owns restocking and repair. Broken or jammed machines were the most-reported problem in a Sydney user survey, and they undo the out-of-hours advantage.
• When the funder demands named data the users will not give. Registration suppresses use; evaluate with anonymous counts and optional questions instead.

Frequently asked questions

Do harm reduction vending machines work?

For access, yes, consistently. A 2025 systematic review of 30 studies found high demand, use mostly outside business hours, acceptance by target populations and reach into higher-risk groups. For health outcomes the evidence is promising but limited: four articles reported less syringe sharing after syringe machines, and two reported fewer fatal overdoses after naloxone machines. These are observational studies, not randomised trials, so they show association rather than proof of cause.

Do naloxone vending machines reduce overdose deaths?

One county study suggests they may. In Clark County, Nevada, researchers forecast 270 opioid-involved overdose deaths in the 12 months after naloxone was added to public health vending machines; 229 occurred, suggesting 41 averted, with a significant immediate drop at launch. In Ohio, a machine’s first year was associated with a lower countywide rate of unintentional overdose death, and the author called for further study of causality. Neither study can trace the naloxone used in a given overdose back to a machine, and other prevention work ran at the same time.

Do harm reduction vending machines reach new people?

There is good evidence they reach people that staffed services miss. In a survey of 712 people who use drugs in rural Kentucky, vending machines were rated more acceptable than syringe services programs, especially by people who do not inject and by people who share syringes without using a program. In Ohio, 12% of 637 people who registered for a machine had not used harm reduction services before. Sydney research found younger users preferred machines and cited stigma at staffed outlets.

Do vending machines replace face-to-face harm reduction services?

No, and the evidence suggests they need not compete. In inner Sydney during COVID-19 restrictions, equipment from a 24-hour dispensing machine rose 41.1% while face-to-face needle and syringe program visits fell 16.2%, and visits to the co-located primary healthcare clinic rose 59.7% a month. The authors concluded the machine did not reduce use of targeted primary healthcare. A 2025 RTI and CDC brief stresses that machines cannot replace the face-to-face engagement that connects people to other services.

Do harm reduction vending machines increase crime or disorder?

The published evidence has not found that. The 2025 systematic review reported stable or declining drug-related crime in the one article that examined it. A 2026 qualitative study in six veterans’ supportive housing buildings found no perceived increase in visible disorder or safety concerns, although residents and staff raised stigma, privacy and placement near children. The evidence base is small, so programs should track complaints, damage and discarded equipment around their own sites.

What does the evidence not show yet?

It does not yet show cause and effect for overdose deaths, HIV or hepatitis C at population level; there are no randomised trials. Most studies count items dispensed, which shows supplies left the machine, not that they were used. Data on who uses machines is weak because use is anonymous, and one Melbourne study found most keypad-entered demographic data invalid. Cost-effectiveness evidence is sparse, and few studies report on race or ethnicity, according to a 2023 scoping review.

How should a program evaluate its harm reduction vending machine?

Decide the questions before launch and build data collection into the machine. Dispense logs give units by product, site and hour. A few optional, anonymous questions before release, such as first supply or refill and whether the last kit was used, add outcome indicators without blocking anyone. Compare with fixed-site data and local overdose figures, and add interviews with users and host sites. Keep questions short: long or confusing keypad entry produced unusable data in Melbourne.

References

• Zhang A et al. — “Vending machines for reducing harm associated with substance use and use disorders, and co-occurring conditions: a systematic review”, Harm Reduction Journal 22:89 (28 May 2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12121115/
• Russell E et al. — “A scoping review of implementation considerations for harm reduction vending machines”, Harm Reduction Journal 20:33 (2023). https://pmc.ncbi.nlm.nih.gov/articles/PMC10018614/
• Allen ST et al. — “Evaluating the impact of naloxone dispensation at public health vending machines in Clark County, Nevada”, Annals of Medicine 54:2692–2700 (2022). https://pmc.ncbi.nlm.nih.gov/articles/PMC9542801/
• Arendt D — “Expanding the accessibility of harm reduction services in the United States: Measuring the impact of an automated harm reduction dispensing machine”, J Am Pharm Assoc 63:309–316 (2023). https://doi.org/10.1016/j.japh.2022.10.027
• Kerr P et al. — “Analysis of four syringe dispensing machine point-of-access data 2017–2020 in Melbourne, Australia”, Harm Reduction Journal 19:144 (2022). https://pmc.ncbi.nlm.nih.gov/articles/PMC9768389/
• Philbrick S et al. (RTI International / CDC) — “Vending machines: A tool for distributing harm reduction equipment” (2025). https://harmreduction.org/wp-content/uploads/2025/03/Vending-Machines-for-Harm-Reduction-2025.pdf
• Knudsen HK, Havens JR, Young AM — “Acceptability and Appropriateness of Harm Reduction Vending Machines Compared to Syringe Service Programs in Appalachian Kentucky”, J Stud Alcohol Drugs 87:345–356 (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12498109/
• Marschke RM et al. — “Discreet but accessible: a qualitative study … optimal design of a harm reduction vending machine in rural Kentucky”, Harm Reduction Journal 22:88 (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12117748/
• University at Buffalo — “Harm reduction vending machines in NYS expand access to overdose treatment and drug test strips, UB studies confirm” (21 January 2026). https://www.buffalo.edu/ubnow/stories/2026/01/lynch-harm-reduction-vending-machines-nys.html
• Rife-Pennington T et al. — “Harm Reduction Vending Machines in Supportive Housing for Veterans”, JAMA Network Open 9(9):e2635246 (2026). https://pubmed.ncbi.nlm.nih.gov/42804692/
• Zhang A et al. — “‘Smart’ harm reduction vending machines to improve public health: Evaluating the utilization”, J Subst Use Addict Treat (2026). https://pubmed.ncbi.nlm.nih.gov/42035889/
• NSW Health — “NSW Needle and Syringe Program”, Guideline GL2023_002 (18 January 2023). https://www1.health.nsw.gov.au/pds/ActivePDSDocuments/GL2023_002.pdf
• Uthurralt N et al. — “The impact of a 24-hour syringe dispensing machine on a face-to-face needle and syringe program and targeted primary healthcare clinic”, Aust N Z J Public Health 46:524–526 (2022). https://onlinelibrary.wiley.com/doi/full/10.1111/1753-6405.13267
• Islam M, Stern T, Conigrave KM, Wodak A — “Client satisfaction and risk behaviours of the users of syringe dispensing machines: a pilot study”, Drug and Alcohol Review 27:13–19 (2008). https://pubmed.ncbi.nlm.nih.gov/18034377/

Related

• Harm reduction vending machines — the cluster hub.
• On-screen questions and forms — what to ask for your evaluation.
• Naloxone vending machines
• Syringe vending machines
• Fentanyl and xylazine test strip vending machines
• Controlled dispensing on harm reduction vending machines

Talk to us about evaluating a harm reduction vending machine

Send the products, sites and the questions your funder will ask. We will propose a machine, an access model and the data each vend will record.