From leverage to a logic model

Outputs / Techniques and logic

BCTs, ERIC strategies and the model

Mapping the coded determinants to behaviour change techniques and implementation strategies, then assembling inputs, activities, outcomes and impact.

From determinants to techniques

By this point the coded determinants exist: TDF domains, the mechanism each one operates through, and CFIR constructs for the organisational level. The mechanisms of action are not a separate framework: they are the TDF extended, the 14 domains plus 12 further constructs from a set of 83 theories of behaviour change (Carey et al. 2019). Two resources then turn the coded material into candidate actions. The Theory and Techniques Tool links those mechanisms to behaviour change techniques. The BCT taxonomy v1 supplies the technique numbers and labels, 93 techniques in 16 clusters (Michie et al. 2013).

Name the mechanismFor each coded determinant, name the mechanism it operates through, using the same list the tool uses, and cite the tool's own entry. If the tool has no entry for that determinant, write "no linkage in the tool" and stop there. Do not substitute a mechanism that merely seems plausible.
List the candidate techniquesTake the number, the published label and the tool's confidence for each link. If a technique is not in the taxonomy it does not go on the list, and no technique number is ever invented.
Mark each candidateCORE, OPTIONAL or DO NOT USE, with the reason. A technique the workshop material contradicts is DO NOT USE.
Say what it looks like hereWho delivers it, to whom, at which step, how often, and what would have to be true for it to work in this service.

The same move happens at the organisational level. Coded CFIR barriers go to the CFIR-ERIC matching tool, which maps 39 CFIR barriers to 73 ERIC implementation strategies (Powell et al. 2015; Waltz et al. 2019). Each recommended strategy carries a tier. Level 1 means more than half the expert panel ranked it in their top seven for that barrier. Level 2 means 20 to 50 per cent did. Never present a level 2 strategy as though it were level 1.

The limits of a matching tool

Both resources are consensus instruments, and both get read as something stronger than they are.

A match is a judgement, not a finding. The matching tool encodes expert agreement about which strategies fit which barriers. It does not encode trial evidence that the strategy works, in your service or in any other.
The mechanism is yours, not the tool's. Neither tool supplies a causal chain from strategy to outcome. If you write that a strategy will work by improving staff confidence, that is your reasoning and it must be labelled as such.
Experts disagree. Waltz and colleagues asked experts to recommend strategies for specific CFIR barriers, and the recommendations varied widely, with only weak agreement for many barriers. That variation is a finding about the tool. It is not noise to be smoothed away.
Selection and evaluation are different activities. A scoping review of studies using the tool found it was nearly always used to generate strategies and rarely used to evaluate whether they worked. Planning is the strong end of the tool; learning is the weak end.
An infeasible recommendation is still a recommendation. If the tool suggests something the service cannot deliver, name it and say why, rather than dropping it quietly or listing it as planned.
Record what you used. The tool is a consensus product that is revised over time, so a report that cites it records the version, or the date the tool was accessed.

The rule that covers all of them. A tool link says that experts associate a technique or strategy with a barrier. It does not say the technique works. Say so on the page where you use it, in the paper where you report it, and in the room.

Assembling the logic model

The logic model is where the rest of the method lands. It has five columns, and the reason for the whole route is that every row can be traced back to a source.

Inputs

The staff, time, money, materials and access the plan requires. Include the unpaid work the journeys revealed: chasing, repeating, translating, paying, arranging, caring. That work is a real input and it is usually invisible on a budget line. If the plan assumes it carries on, say so in the row.

Activities

Each technique and strategy, named exactly as in its taxonomy, with who delivers it, to whom, at which step, and how often. This is the one column where a framework name is required, and it is required so that someone outside the team can recognise what was done.

Short-term outcomes

The determinants you expect to move, each named with its framework and its measure: a TDF domain, a mechanism of action, a CFIR construct. If there is no measure, the row is a wish and it should be labelled as one.

Medium-term outcomes

The behaviour changes expected in each actor, with the measure. Name the actor, meaning the person whose behaviour is meant to change. A behaviour change with no actor attached is not a row.

Impact

The health, wellbeing or service outcome the project exists for, with the chain of reasoning that reaches it from the behaviour change. The chain is a claim, so keep it short enough to argue with.

Label every row

A logic model becomes defensible through one habit. Every row carries a label saying where it came from. There are three labels, and they are not interchangeable.

One label per row. Rows with more than one label are usually two rows.
Label	It means	What it commits you to
Measured in this project	The row rests on data you collected, with the method recorded.	You can state the value and the source of it.
Derived from the map	The row rests on a variable, a link, a loop or a coded touch point that the workshop produced.	You can trace it to the material and to the person who said it.
Assumed	The row rests on the team's reasoning, on a framework, or on convention.	You can say plainly that it is an assumption, and list it.

Collect every assumed row in one place at the end. Keep the labels honest: derived from the map is not a softer word for assumed, and a row does not become measured because a measure exists somewhere in the project. Then state what would falsify the model. A model that cannot fail cannot teach you anything.

Check your understanding & apply it · 4 questions, 2 essays

1A logic model row says the project will improve patients' confidence at booking. Its only support is a coded touch point from the workshop, with a quote from a patient. Which evidence label does the row deserve?

2A tool link says experts associate a technique with a barrier. What else does the matching tool tell you about whether the technique works in your service?

3The CFIR-ERIC matching tool recommends a strategy your service cannot deliver. What does the page say you should do?

4A recommendation carries tier 2. How should it be reported?

5Essay A clinic has run a mapping workshop. The material shows that patients arrive not knowing what their appointment is for, and that this sits behind missed appointments; the page records a coded touch point with quotes. The team plans to send a reminder text two days before each appointment and to train reception staff to explain the purpose of the appointment at booking, reasoning that clearer explanations will reduce missed appointments. Attendance will be recorded at every appointment. Assemble a logic model for this service, with every row labelled.

6Essay Write the short paragraph you would put in a report wherever you use the CFIR-ERIC matching tool, covering what a match does and does not mean and how the recommendations are reported.