CFIR, RE-AIM and Proctor's outcomes: choosing an implementation framework without overcomplicating your study

Implementation science has more frameworks than most researchers can usefully hold, and the practical consequence is a common proposal pathology: a study that names three frameworks, operationalises none, and measures whatever data happened to be available. The way out is to classify your question before you choose anything.

1. Three question types, three tools

2. CFIR: for determinants, when the setting is definable

The Consolidated Framework for Implementation Research organises determinants of implementation across domains covering the innovation itself, the outer and inner settings, individuals involved, and the implementation process. The 2022 update, published in Implementation Science by Damschroder and colleagues, revised constructs based on a decade of user feedback and is now the version to cite; construct definitions are maintained on the official CFIR guide.

CFIR's own guidance is unusually candid about when it is appropriate: when your question is about predicting or explaining implementation outcomes through determinants, when you can define the inner setting being studied — a clinic, a hospital, a municipality — and when your team has implementation-science methodological expertise. If any of those three is missing, CFIR will produce interview transcripts coded into a framework rather than an explanation.

• Use it for: formative assessment before scale-up, explaining variation between sites, and structuring qualitative interview guides.
• Do not use it for: measuring whether the programme worked, or for outcome definition — that is not what a determinant framework does.
• Practical scope control: pre-select the constructs plausibly relevant to your setting, justify the selection in the protocol, and report what you excluded.

3. RE-AIM: for impact and equity of reach

RE-AIM, introduced by Glasgow, Vogt and Boles in the American Journal of Public Health in 1999, evaluates interventions along five dimensions: Reach, Effectiveness, Adoption, Implementation and Maintenance. Its 20-year review in Frontiers in Public Health (2019) adapted the framework to newer science and practice, and a 2021 clarification paper in the Journal of Clinical and Translational Science addresses the misapplications that had accumulated. PRISM — the Practical, Robust Implementation and Sustainability Model — is the companion that brings contextual factors into the same design.

RE-AIM earns its place in low-resource settings mainly through Reach and Maintenance. Reach forces the equity question: of the population that needed this, who actually got it, and who was systematically missed? Maintenance forces the sustainability question that project-cycle evaluations habitually avoid, because it asks about the period after the funding ends.

1. 01Define the denominator for Reach before the intervention starts — retrospective denominators are where equity claims go to die.
2. 02Separate Effectiveness (did outcomes change?) from Implementation (was it delivered as intended, and at what cost?).
3. 03Specify Adoption at the level of the delivering unit: facility, ward or provider, not the abstract 'system'.
4. 04Plan the Maintenance measurement point explicitly — typically six to twelve months after external support ends.

4. Proctor's taxonomy: the distinction that fixes evaluations

Proctor and colleagues, writing in Administration and Policy in Mental Health in 2011, defined implementation outcomes as the effects of deliberate and purposive actions to implement new treatments, practices and services — and separated them explicitly from service outcomes and client outcomes. The taxonomy's commonly used outcomes are acceptability, adoption, appropriateness, feasibility, fidelity, implementation cost, penetration and sustainability. An Implementation Outcome Repository catalogues validated instruments mapped to each of them, which saves considerable time at instrument-selection stage.

A study that measures patient outcomes and calls them implementation outcomes has not evaluated implementation. It has evaluated the intervention and hoped.

The practical payoff is that this distinction lets you report a truthful negative result. If fidelity was low and penetration was minimal, then flat clinical outcomes tell you nothing about whether the intervention works — only that it was never really delivered. Without the taxonomy, that finding gets written up as 'the intervention was not effective', which is both wrong and expensive.

5. Combining frameworks without drowning

A defensible combination for a typical municipal or district study looks like this: CFIR to structure formative determinant assessment and the interview guide; a small set of Proctor outcomes — usually fidelity, acceptability, penetration and cost — as the measured implementation endpoints; and RE-AIM to frame the overall evaluation and force the Reach and Maintenance questions. Three frameworks, each doing one job, with a written justification for every construct and outcome you retained.

• Pre-register or at minimum pre-document your construct and outcome selection, with reasons for exclusions.
• Cap qualitative scope: eight to twelve well-sampled interviews per site analysed properly beats forty analysed superficially.
• Choose instruments that already exist and have been used in comparable settings; adapt and pilot rather than invent.
• Report cost. Implementation cost is the outcome decision-makers ask about first and researchers report least often.