Introduction to logic models
Published 7 August 2018
© Crown copyright 2018
This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk.
Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.
This publication is available at https://www.gov.uk/government/publications/evaluation-in-health-and-well-being-overview/introduction-to-logic-models
Introduction to logic models
A logic model is a graphic which represents the theory of how an intervention produces its outcomes. It represents, in a simplified way, a hypothesis or ‘theory of change’ about how an intervention works. Process evaluations test and refine the hypothesis or ‘theory of change’ of the intervention represented in the logic model.
The design of, and terms used in, logic models vary. However, they commonly include aspects which summarise an intervention’s:
- inputs/resources
- implementation/outputs
- outcomes/impact
- context
- relationships between them
The Medical Research Council (MRC) ‘Process evaluation framework’ (Moore and others 2015) outlines the main aspects of an intervention that a logic model should represent to inform evaluation.
- Implementation refers to how a service or intervention gets delivered and what gets delivered in practice.
- Mechanisms of impact relate to the mechanisms through which the intervention works and produces changes in the intervention recipients.
- Outcomes are the changes that the intervention is ultimately trying to bring about for recipients, such as weight loss or diabetes prevention.
- Context refers to factors external to the intervention that might influence how the intervention operates.
Logic models identify, describe and arrange these critical aspects of an intervention to represent how the intervention produces change, with arrows often used to indicate causal relationships between the aspects.
Logic models are useful for evaluation because they can help prioritise and structure data collection and analysis to explore the main aspects of an intervention and relationships between them. These data can be used to help to explain how the intervention works to achieve its outcomes, or sometimes why it does not work.
A logic model might identify the delivery and content of children’s drama sessions which encourage child-parent discussions about how to eat more healthily as an ‘implementation’ aspect, and changing family attitudes towards diet as a subsequent ‘mechanism of impact’ aspect, with an arrow indicating a causal relationship between the two.
Based on this logic model, an evaluator might collect programme records and observational data on the quantity, quality and content of the drama sessions delivered to children, and interview families about their changing attitudes towards food (including questions about the causal relationship between the drama sessions and their attitudes towards food). By collecting data on these aspects of the logic model, the evaluator is able to explore main aspects of the ‘theory of change’ of the intervention.
Ideally, logic models should be developed during intervention development, or in the early stages of planning an evaluation, so that they can be used to inform the design of the process evaluation and data collection. However, researchers can develop a logic model to describe an intervention at any stage of an evaluation, including retrospectively - for example, the EPODE (Ensemble Prévenons l’Obésité Des Enfants or Together let’s prevent childhood obesity) logic model.
How to create a useful logic model
Ideally, process evaluations are ‘theory based’ evaluations, where the evaluation is supported by the underpinning theory of how the intervention works (see Process Evaluation section). The logic model should draw on and summarise this theory. In some cases, an intervention manual may also be useful to specify critical aspects related to the implementation/delivery of the intervention in more detail.
Logic models are usually designed by a study team rather than a single researcher. They tend to require multiple perspectives and careful discussion to identify the critical aspects of an intervention, and to consider how they are related. Just as stakeholders such as patients, clients and staff delivering an intervention can contribute to the development of an intervention, stakeholders can also contribute to the construction of a logic model.
Some approaches, such as the ‘theory of change’ model promoted by the Aspen Institute emphasise stakeholder involvement, while some other approaches design the logic model with researchers only.
Good logic models share several features
They do not include detail about absolutely everything that happens in an intervention, but summarise the aspects that are critically important in explaining how the intervention produces the changes that it is aiming to achieve. Therefore, it is important when creating a logic model to decide what these critical aspects are.
They ensure, as far as possible, that ‘implementation’, ‘mechanisms of impact’, ‘outcomes’ and ‘context’ are separated in a logic model because these are conceptually distinct. An evaluator will probably have different research questions about these different aspects of the intervention and may want to use different methods to answer these questions
They represent the ‘cause and effect’ relationships between different aspects of an intervention using arrows.
Their inclusion should therefore be considered carefully; if an arrow is present, the study team should be able to articulate what causal relationship it represents.
They are not just conceptual maps which represent intervention aspects and the relationships between them; they represent a process of change. Therefore, the time order of ‘cause and effect’ represented in the logic model is important. Cause always comes before effect. The aspects should be arranged so that the arrows between them show how one aspect (such as delivery of training material) causes a subsequent aspect (such as improved skills).
They will also represent relationships between different aspects as specifically as possible rather than, for example, listing all ‘implementation’ aspects and linking this list with one arrow to a list of ‘mechanisms of impact’ aspects. This is particularly true for interventions which are large or complex and have many different components, with different theorised mechanisms.
A ‘list’ approach in a logic model does not illustrate the causal pathways or theory of the intervention in much detail. Therefore, it will not be useful in guiding the data collection and analysis.
There are also several features of logic model design which are highly variable, depending on the intervention. A good logic model is designed and adapted so that it accurately represents the underpinning theory of the specific intervention, rather than being based on a standard template.
Logic models may include more detail on one aspect of the intervention than others. The following examples demonstrate how this can be true for different aspects.
In some interventions, the implementation process may be complex, involving different stages of staff training and re-organisation of service delivery systems. For this type of intervention, the logic model might include a number of aspects to describe the different aspects and stages of implementation.
In other types of interventions, such as those using mobile phone apps, the implementation process is likely to be much simpler and the logic model would not need to include much detail on this aspect of the intervention. However, the mobile phone app may be designed with multiple ‘mechanisms of impact’ (for example, goal-setting, social support, description of behaviour, feedback, and self-monitoring).
Some interventions may only impact on outcomes at one point in time, and this will be represented in the final section of a logic model. In other studies, short, medium and long-term outcomes may all be of interest, and therefore the ‘outcomes’ part of the logic model might extend to several sections.
An additional factor to consider relates to the level of intervention. For example, whether the intervention is targeted at an individual, family, organisational or community level, as this will affect the content of the logic model at:
- an individual level: a logic model for an individual level intervention might include detailed aspects describing psychological processes which underpin its mechanisms of impact (such as changes in individuals’ social identity and attitudes)
- a community level: a logic model for a community intervention would include aspects describing group or community level changes (such as increased participation rates in a local sports facility following a physical activity promotion campaign)
Early development of a logic model is advisable. It helps to focus the process evaluation on the most important research questions and make best use of limited resources for data collection and analysis.
Several researchers who contributed to case studies included in the MRC process evaluation guidance were asked what they would have done differently in their studies. They commented that they would have developed the intervention theory and logic model earlier, to allow them to address the major process evaluation questions in a more focused way.
Limitations of logic models
Logic models tread a fine line between being simple, easy to understand and use, and reflecting the complexity of the real world. A good logic model will include the critical aspects of an intervention that contribute to its outcomes.
In practice, these may be hard to identify in advance, and therefore the logic model may not include every factor that explains outcomes. On the other hand, a very complex logic model may become unwieldy and impractical. The evaluator therefore has the tricky task of finding a balance between these 2 demands.
Logic models have also been criticised for representing interventions as linear and mechanistic, and for overplaying the predictability of an intervention. For example, they tend to give the impression of steady change over time, whereas change may occur in jumps at certain points or problems may initially get worse before they get better.
An alternative perspective comes from ‘complexity theory’ which emphasises the unpredictability of processes and characterises interventions as ‘adaptive’ rather than fixed systems (Plsek and Greenhalgh 2001).
Logic models could be adapted to a certain extent to accommodate some aspects of complexity theory, such as feedback loops (these are present in the MRC process evaluation framework). However, logic models will remain relatively mechanistic, linear representations of processes of change which do not reflect the full complexity of the real world but, as noted above, provide a simplicity that has advantages for planning and conducting evaluations.
Categorising aspects of logic models
In practice, it can sometimes be tricky to define and agree on which category (such as ‘implementation’ or ‘context’) an aspect of a logic model falls into. There are no standard definitions for what counts as ‘implementation’, a ‘mechanism of impact’, ‘context’ or ‘outcome’ as these will vary depending on the study.
For example, a school may be part of the context for a child bullying intervention which provides counselling for individuals. Alternatively, if an intervention is designed to change the school culture in order to reduce bullying, then the school is part of the intervention not the context.
To decide where aspects of a logic model fit, it can be helpful to think through the intervention in as simple terms as possible.
Implementation
What are the major features that characterise your intervention delivery (for example, what resources and services is the intervention providing; what activities are intervention staff required to undertake; how is the intervention or service delivered)?
Mechanisms of impact
What processes should be triggered by delivery of the intervention (for example, what changes should happen in intervention recipients that wouldn’t have happened otherwise; how should participants respond to the intervention)?
Outcomes
What are the ultimate aims of the intervention (for example, what changes the intervention is aiming to achieve)?
Context
What external factors may influence the intervention (what is the intervention not addressing; what is beyond the scope of the intervention; what factors in the organisation’s or participants’ environment might affect the intervention)?
In addition, difficulties can arise where aspects of the logic model fall into grey areas on the boundaries between categories. For example, patient acceptability may be considered an aspect of ‘implementation’ since it reflects on the delivery of the intervention. However, it could also be considered an aspect of how patients respond to and engage with an intervention and, therefore, be included in the ‘mechanisms of impact’ category.
Alternatively, a short-term outcome (for example, increased used of a sports facility) could reflect a mechanism of impact in an intervention with an ultimate goal of increasing physical activity. If you are having problems identifying what categories the aspects of your logic model fall into, and if some aspects seem to be on the boundaries, then it can be more useful to concentrate on what the critical aspects of your intervention are.
Concentrate on how they are causally related to each other, rather than getting too caught up in making aspects fit into the ‘implementation’, ‘mechanisms of impact’ or ‘context’ categories exactly.
Developing a logic model for exploratory interventions
If you have designed an intervention, you will have some hypothesis about how it is expected to produce change to resolve the problem you are trying to address. This hypothesis can form the basis of a logic model.
However, there may be some aspects of how the intervention might work or processes that may occur which due to large gaps in knowledge are genuinely unknown. For example, it may be that the intervention is being delivered in a new setting where the contextual factors that could affect the intervention are unknown.
Alternatively, the participant response to the intervention may be very uncertain. In these cases, there may be gaps in logic models which can be specifically explored during the process evaluation with the findings contributing to a more complete logic model at the end of the study.
Useful resources
The following are most relevant to logic models.
Centers for Disease Control and Prevention (2011). ‘Developing an effective evaluation plan’
Julia Coffman (1999). ‘Learning from logic models: an example of a family/school partnership program’
Evaluation Support Scotland (2012). ‘Developing a logic model’
HM Treasury (2011). ‘The Magenta Book: guidance for evaluation’
Health Canada (2013). ‘A guide for first nations on evaluating health programs’
Medical Research Council (2006). ‘Developing and evaluating complex interventions: new guidance’
Medical Research Council (2014). ‘Process evaluation of complex interventions’
National Science Foundation (2002). ‘The 2002 user-friendly handbook for project evaluation’
NHS Health Scotland (2003). ‘LEAP for health: learning, evaluation and planning’
NHS National Observatory on Obesity (2015). ‘Evaluation of weight management, physical activity and dietary interventions: an introductory guide’
NHS National Obesity Observatory (2012). ‘Standard evaluation framework for dietary interventions’
NHS National Obesity Observatory (2012). ‘Standard evaluation framework for physical activity interventions’
NHS National Obesity Observatory (2009). ‘Standard evaluation framework for weight management interventions’
UN Programme on HIV/AIDS (2008). ‘Framework for monitoring and evaluating HIV prevention programmes for most-at-risk populations
US Department of Health and Human Services (2010). ‘The program manager’s guide to evaluation’
University of Wisconsin Extension Program Development and Evaluation: logic models
WK Kellog Foundation (2006). ‘Logic model development guide’
WK Kellog Foundation (2004). ‘Logic model development guide: using logic models to bring together planning, evaluation and action’
World Health Organization (2013). ‘WHO evaluation practice handbook’
Acknowledgements
This work was partially funded by the UK National Institute for Health Research (NIHR) School for Public Health Research, the NIHR Collaboration for Leadership in Applied Health Research and Care of the South West Peninsula (PenCLAHRC) and by Public Health England. However, the views expressed are those of the authors.
Written by Sarah Morgan-Trimmer, Jane Smith, Krystal Warmoth and Charles Abraham.
References
Anderson A (2005). ‘The community builder’s approach to theory of change: a practical guide to theory development’ New York: The Aspen Institute roundtable on community change
Coffman J (1999). ‘Learning from logic models: an example of a family/school partnership program’ Harvard Family Research Project
De Silva and others (2014). ‘Theory of change: a theory-driven approach to enhance the Medical Research Councils’ framework for complex interventions’ Trials 15: 267
Evaluation Support Scotland (undated). ‘Evaluation support guide 1.2: developing a logic model
Moore, G and others (2014). ‘Process evaluation of complex interventions: Medical Research Council guidance’ London: MRC Population Health Science Research Network
Plsek, PE and Greenhalgh, T (2001). ‘The challenge of complexity in health care’ British Medical Journal 323: 625-8.
University of Wisconsin–Extension (undated). ‘Program development and evaluation: logic models’
Van Koperen, TM, and others (2013). ‘Characterizing the EPODE logic model: unravelling the past and informing the future’ Obesity Reviews 14 (2): 162-170
WK Kellog Foundation (2004). ‘Logic model development guide: using logic models to bring together planning, evaluation and action’
Access a series of supporting materials: