Written and maintained by CASRAI Editorial Board
Last updated
RE-AIM is a planning and evaluation framework that scores a health, behavioral, or public-health intervention on five separate dimensions: Reach, Effectiveness, Adoption, Implementation, and Maintenance. It was introduced by Russell Glasgow, Thomas Vogt, and Steven Boles in a 1999 American Journal of Public Health paper as a corrective to a specific reporting problem: intervention studies routinely reported a clean effect size on a well-controlled sample and said almost nothing about who was actually reached, whether real-world settings would take the program up, whether it was delivered as designed, or whether any of it lasted once the study ended.
The acronym is well known; the part researchers and evaluators actually get stuck on is not what each word means but what to count and what to divide it by — particularly the difference between Reach and Adoption, which both sound like “how many people used this” but are measured against entirely different denominators. This guide gives each dimension a concrete indicator, states its denominator explicitly, and covers the framework’s PRISM extension and how to report RE-AIM results in a paper or protocol.
The two levels RE-AIM operates at
The single most useful fact for resolving RE-AIM confusion is this: the five dimensions are not five versions of the same question. They split across two levels of analysis.
- Individual level — Reach and Effectiveness describe what happens to people: how many participated, and what changed for them.
- Setting level — Adoption describes what happens to organizations, sites, and staff: how many clinics, schools, or worksites agreed to deliver the program.
- Both levels — Implementation (delivery fidelity at the setting level, receipt of the intended dose at the individual level) and Maintenance (durable individual-level outcomes and whether the setting keeps running the program) span both.
Once the level is fixed, the denominator follows from it, and the recurring Reach/Adoption mix-up mostly resolves itself.
The five dimensions, with a concrete indicator for each
Reach (individual level)
What it measures: the absolute number, proportion, and representativeness of individuals who participate.
Denominator: the full eligible target population — everyone who could have taken part, not just the subset who were actually approached or invited.
Concrete indicator: reach rate = participants enrolled ÷ eligible individuals in the target population, reported alongside a comparison of participant characteristics (age, sex, race/ethnicity, baseline severity) against the eligible population’s characteristics. A high enrollment percentage among people who were approached is not the same claim as a high reach rate against the full eligible population, and representativeness matters as much as the raw percentage — a program that reaches 40% of an eligible population but skews heavily toward the least severe cases has a different external-validity story than one that reaches 40% with a demographic profile matching the eligible population.
Common mistake: reporting the response rate among people who were already approached (a much easier number to look good on) and calling it reach. True reach needs the full eligible population as the denominator, including everyone who was never approached at all.
Effectiveness (individual level)
What it measures: the intervention’s impact on outcomes that matter, including quality-of-life and any negative or unintended effects, not just the single favorable primary outcome.
Denominator: participants who enrolled (an intention-to-treat framing is generally preferred over restricting the analysis to only those who completed the full intervention as designed, since the latter inflates the apparent effect).
Concrete indicator: the between-arm or pre-post effect size on the primary outcome, reported together with at least one broader or negative outcome (adverse events, participant burden, cost to the participant, quality-of-life measure) and broken out by at least one relevant subgroup (for example, by baseline severity or by demographic group) to surface differential effectiveness rather than a single pooled number.
Common mistake: reporting only the outcome the intervention was designed to move, and omitting quality-of-life or negative-effect data even when it was collected.
Adoption (setting level)
What it measures: the absolute number, proportion, and representativeness of settings (clinics, schools, worksites, community organizations) and the staff within them who agree to deliver the program.
Denominator: the full set of eligible settings (every clinic in the network, every school in the district) and, within participating settings, the eligible staff who could deliver it.
Concrete indicator: adoption rate = settings that agreed to implement ÷ eligible settings, reported alongside a comparison of adopting versus non-adopting settings on characteristics likely to matter (size, patient volume, existing resources, urban/rural status). This is the direct contrast with Reach: Reach is an individual-level count against an individual-level denominator (people out of eligible people); Adoption is a setting-level count against a setting-level denominator (sites out of eligible sites).
Common mistake: describing a high participation rate among patients or clients as evidence of “strong adoption.” That is Reach, restated. Adoption is specifically about whether the site and its staff opted in to offer the program at all — a question that can only be answered by counting sites, not people.
Implementation (both levels)
What it measures: the extent to which the intervention is delivered as intended — consistency of delivery across staff and settings, fidelity to the protocol, the time and cost required to deliver it, and any adaptations made along the way. At the individual level, this becomes the dose or content each participant actually received.
Denominator: the protocol-specified components or dose, used as the 100% fidelity benchmark against which what was actually delivered is compared.
Concrete indicator: a fidelity score — the percentage of protocol-specified components delivered as designed, typically assessed through direct observation, session recordings, or a structured checklist audit — reported together with the average time and cost per session or per participant actually required to deliver the program.
Common mistake: treating Adoption as a proxy for Implementation — assuming that because a site agreed to run the program, it was delivered as designed. Those are separate questions; a site can adopt a program and still deliver it with substantial drift from protocol.
Maintenance (both levels)
What it measures: durability. At the individual level, whether outcome changes persist once the intervention period ends. At the setting level, whether the program becomes routine practice — continuing after the study, grant funding, or research team’s involvement ends — rather than lapsing the moment external support stops.
Denominator: at the individual level, participants originally enrolled, assessed at a long-term follow-up (RE-AIM’s convention is six months or more after the intervention ends, distinguishing genuine maintenance from a short-term effect that fades). At the setting level, the settings that adopted the program, assessed for whether they are still delivering it.
Concrete indicator: individual level — the percentage of originally enrolled participants retaining a clinically or practically meaningful outcome change at 6-to-12-month follow-up. Setting level — the percentage of adopting sites still delivering the program 12 or more months after study funding or external support ends, and what it is now funded and staffed by (a permanent budget line versus an unfunded volunteer effort is a materially different maintenance story).
Common mistake: not collecting Maintenance data at all. It is the dimension most often skipped in published RE-AIM applications, mainly for a structural reason rather than a design choice — study funding and the formal follow-up period usually end at the same point the intervention does, and going back to check whether a program is still running twelve months later requires effort and funding the original grant may not have budgeted for.
Reach vs. Adoption, side by side
| Dimension | Level of analysis | Denominator | Concrete indicator |
|---|---|---|---|
| Reach | Individual | Eligible individuals in the target population | Participants enrolled ÷ eligible population, plus a representativeness comparison |
| Effectiveness | Individual | Participants enrolled (intention-to-treat) | Primary-outcome effect size plus at least one negative/quality-of-life outcome, by subgroup |
| Adoption | Setting / staff | Eligible settings and eligible staff within them | Settings that agreed to implement ÷ eligible settings, plus adopter-vs-non-adopter comparison |
| Implementation | Setting and individual | Protocol-specified dose or components | Fidelity score (% of components delivered as designed) plus cost/time per participant |
| Maintenance | Individual and setting | Enrolled participants (individual); adopting settings (setting) | % retaining outcome at 6–12+ months (individual); % of sites still delivering 12+ months post-funding (setting) |
Why the mix-up happens
Reach and Adoption both answer some version of “how many took part,” which is why they get conflated in practice. The fix is to ask which unit is being counted before choosing a denominator: if the unit is people, it is Reach and the denominator is the eligible population of people; if the unit is sites, clinics, schools, or the staff inside them, it is Adoption and the denominator is the eligible population of sites. A study can have high Reach and low Adoption — a program that, wherever it is offered, draws in most eligible patients, but that only a small fraction of eligible clinics ever agreed to offer in the first place. Reporting only the Reach number in that scenario overstates the program’s real-world footprint; the honest external-validity picture needs both.
PRISM: RE-AIM’s contextual extension
PRISM (the Practical, Robust Implementation and Sustainability Model), developed by Feldstein and Glasgow in 2008, extends RE-AIM rather than replacing it — PRISM’s own documentation describes RE-AIM’s five dimensions as the outcomes it produces. PRISM adds four contextual domains that sit around an intervention and interact with it to produce those RE-AIM outcomes:
- Perspectives on the intervention, from both the organization and the people receiving it.
- Characteristics of the implementing settings, staff, and recipients.
- External environment — policy context, resource availability, regulatory factors.
- Implementation and sustainability infrastructure — training, monitoring systems, and the resources available to support ongoing delivery.
RE-AIM alone is usually enough for reporting outcomes; PRISM is worth adding when a study also needs to explain why those outcomes came out the way they did — the same distinction that separates RE-AIM from a determinants framework like the Consolidated Framework for Implementation Research (CFIR). CFIR catalogs the barriers and facilitators that shape implementation; RE-AIM (and PRISM, when the fuller contextual picture is needed) specifies what to measure as a result.
Reporting RE-AIM in a paper or protocol
The clearest way to report RE-AIM results is a single table with the five dimensions as rows and, for each one, the specific indicator used, its denominator, and the resulting value — the same structure as the comparison table above, populated with a study’s actual numbers rather than the generic definitions. Two habits make the difference between a credible RE-AIM report and a cosmetic one:
- State the denominator source in the Methods section, not just the resulting percentage. A reach rate is only interpretable if a reader can see where the eligible-population count came from (a registry, a catchment-area census, an EHR query) — an unsourced denominator is functionally unverifiable.
- Report every dimension that was measured, and say explicitly which ones were not, rather than silently omitting the ones that came out poorly or were never collected. Partial RE-AIM reporting is normal and does not need to be apologized for, but it should be stated as a limitation, not implied by omission.
Specifying RE-AIM outcomes up front, at the protocol or grant-application stage, also makes the framework more useful than retrofitting it onto results after a study has already ended — particularly for Adoption and Maintenance, which require setting-level and long-term follow-up data that are difficult to collect retroactively if they were not planned for from the start.
How RE-AIM relates to hybrid effectiveness-implementation designs
RE-AIM is frequently the outcomes framework used inside a hybrid effectiveness-implementation trial, where a study collects both clinical-effectiveness data and implementation data within a single protocol. In that context, RE-AIM’s individual-level dimensions (Reach, Effectiveness) typically map onto the trial’s clinical-effectiveness aim, while its setting-level dimensions (Adoption, Implementation) map onto the implementation aim — giving a Type II or Type III hybrid design a ready-made structure for specifying what “implementation outcomes” actually means in the protocol, rather than leaving the term undefined. More broadly, RE-AIM sits within the field of implementation science, and its emphasis on setting-level and representativeness data is closely related to the concept of external validity: RE-AIM was explicitly designed to push evaluation beyond internal validity (did the intervention work under controlled conditions) toward external validity (will it work, and be taken up, under real-world conditions and by a representative population). Reach’s representativeness requirement, in particular, only means something once a study has been explicit about how its sample was actually selected — see convenience sampling for how selection method and representativeness interact more generally.
Frequently asked questions
What does RE-AIM stand for?
Reach, Effectiveness, Adoption, Implementation, and Maintenance — a five-dimension framework for planning and evaluating the real-world impact of a health or behavioral intervention, introduced by Glasgow, Vogt, and Boles in 1999.
What’s the actual difference between Reach and Adoption?
Reach is an individual-level measure: participants enrolled divided by the eligible population of people. Adoption is a setting-level measure: sites (and staff) that agreed to deliver the program divided by the eligible population of sites. A study can score well on one and poorly on the other, so they need to be reported separately, not folded into a single participation number.
Do I need to measure and report all five RE-AIM dimensions?
No — partial RE-AIM reporting is common, and Adoption, Implementation, and especially Maintenance are frequently harder to collect than Reach and Effectiveness because they require setting-level or long-term follow-up data. What matters is being explicit about which dimensions were measured and which were not, rather than silently reporting only the ones with favorable numbers.
Is RE-AIM the same thing as CFIR?
No. RE-AIM is an outcomes framework — it specifies what to measure to evaluate an intervention’s real-world impact. The Consolidated Framework for Implementation Research (CFIR) is a determinants framework — it catalogs the barriers and facilitators that explain why implementation succeeded or failed. The two are commonly used together: CFIR to structure the explanation, RE-AIM to structure the outcome measures.
What is PRISM and how does it relate to RE-AIM?
PRISM (Practical, Robust Implementation and Sustainability Model), developed by Feldstein and Glasgow in 2008, extends RE-AIM by adding four contextual domains — perspectives, setting/staff/recipient characteristics, the external environment, and implementation/sustainability infrastructure — that interact with an intervention to produce its RE-AIM outcomes. RE-AIM’s five dimensions sit inside PRISM as the outcomes side of the model.
