How to use the Theoretical Domains Framework

By Masood Khan

In a previous post, we broke down one of the most commonly used frameworks in implementation science, the Theoretical Domains Framework (TDF). And while the TDF is useful for identifying barriers and facilitators that influence behaviour change in knowledge translation and implementation studies, implementation researchers still face challenges in its application, including lack of time, resources, and practical guidance on its use [1].

In response, Atkins and colleagues outlined a step-by-step guide for using the TDF in qualitative implementation research. This post summarizes those guidelines [2].

1. Select the Target Behaviour(s)

According to Atkins et al., the first step in using the TDF effectively is to identify which behaviour(s) need to be changed to achieve the desired implementation objective, based on existing evidence. A key consideration at this stage is whether or not the target behaviours are dependent on any other behaviours. If so, those determinant behaviours will also need to be targeted. If there are many interdependent behaviours, and targeting a large number of behaviours is not feasible, prioritize one or two of them.

When selecting target behaviours, they should meet the following criteria:

• The behaviour is important for bringing about the desired change;
• The behaviour is changeable/modifiable;
• Change in the target behaviour is not likely to have negative effects on other behaviours; and,
• The behaviour can be measured.

2. Be Specific

After identifying what target behaviours need to be changed, the next steps are to specify:

• Who needs to modify behaviour (i.e. healthcare practitioners, patients, or other stakeholders);
• When they need to do it (i.e. pre- or post-surgery);
• Where they need to do it (i.e. the treatment or intervention context);
• How often they need to do it (i.e. frequency); and,
• With whom they need to cooperate (i.e. patients, managers, etc.) to achieve the modified behaviour.

For instance, a clearly specified target behaviour statement could be something like:

Physiotherapists [the who] should measure balance using a specific test instrument [the target behaviour] with in-patient [the where] stroke patients [the with whom] upon discharge [the when].

It is important to specify the behaviour in terms of the target behaviour — focusing on positive intentions and the intended behaviour change (‘look where you want to go’) — rather than the problem behaviour. For example, a recommendation that focuses on the positive would indicate that physicians should recommend exercise [target behaviour] for back pain, instead of recommending they reduce over-prescription of painkillers [problem behaviour)]— even though both recommendations are targeted to reduce over-prescription.

3. Select a Research Methodology

Selection of an appropriate research methodology depends on the research question and how much is known about the research problem. For topics with little available evidence, or for exploratory studies, qualitative methods are generally more appropriate.

On the other hand, quantitative methods are useful for descriptive studies when more is known about the topic and potential influencing behaviours, and the aim is to identify those behaviours [1].

Although the TDF has been mostly used in qualitative studies —using questionnaires [3], interviews [4], and focus groups [5] — it has also been used to in mixed-methods studies [6] and quantitative studies such as systematic reviews [7], though there is no guidance yet for using the TDF in quantitative research.

The remaining steps relate to studies using interviews/focus groups as data collection tools in a qualitative methodological setting.

4. Determine the Sampling Strategy

The overall target population of the study will include those identified in step 1 (selecting the target behaviour), in addition to any relevant stakeholders. These may include patients and caregivers (for individual-level behaviour change), or groups such as organizations and hospitals (for organizational-level behaviour change).

Researchers should aim to specify a minimum sample size beforehand and to include diverse perspectives to improve the validity of collected data (i.e. ensuring the data truly represent what is being measured). The sample size depends on many things, but if using qualitative research methods, having a minimum sample size of ten participants for interviews, and of three participants for focus groups, is recommended [8].

5. Develop an Interview Guide

An interview guide should be in language relevant to the target population, and piloted with representative participants to ensure comprehension, usefulness, and appropriateness.

The interview questions should:

• Be open-ended;
• Focus on the target behaviour;
• Probe to elicit details; and
• Cover all the TDF domains that may be relevant.

The interview questions don’t need to sequentially follow the TDF domains. Instead, Atkins et al. recommend flexibility in the order of questions to maintain the natural flow of conversation.

6. Determine Data Collection Methods

Based on the research methodology, data can be collected through interviews, focus groups, observation, and/or document analysis. For interviews, probing and follow-up questions are often used to ensure in-depth data is gathered and to gain an understanding of how the TDF domains relate to the target problem(s).

For example, if asking a question about how confident a participant is in performing a behaviour, a good probing/follow-up question could be “what makes you confident in performing this behaviour?” Open-ended questions like these provide participants an opportunity to further explain their experiences and thoughts on barriers and facilitators.

7. Define the Data Analysis Plan

While Atkins et al. state that while those analyzing the data don’t necessarily need to have expertise in using particular theories, models, or frameworks, they recommend having at least a basic understanding of the domains and the theoretical constructs they represent. Data can be analyzed both deductively (by coding data into the TDF domains, serving as themes) and inductively (by deriving themes from the data, and then mapping these to the TDF domains) and require the following considerations:

7a. Develop Coding Guidelines

For deductive analysis, a coding guideline should be developed before data collection, and should explicitly outline how the TDF will be applied to the data. The coding guideline should be refined throughout the data collection phase by reviewing the transcripts before data collection is completed. This will help decrease discrepancies between different coders. For deductive analysis, the TDF domains are used as themes and data can be directly coded into the themes.

For inductive analysis, theme development comes after data collection, where themes are allowed to emerge from the collected data based on their frequency/dominance in the data [4]. In the context of the TDF, emerged themes can be matched to TDF domains to see if the TDF was successful in identifying the most recurrent themes. It’s possible not all inductively coded data will fit within the TDF domains. Those that don’t shouldn’t be ‘forced’ into the TDF. Keeping track of themes that don’t fit within the TDF is important to assess how much data the TDF can account for, and where there may be room for improvement.

General flow of inductive (moving from the specific to more general) vs. deductive (moving from general to more specific) processes for qualitative research projects. Modified from https://socialresearchmethods.net/kb/dedind.php

7b. Have a Plan to Resolve Coding Discrepancies

Atkins et al. recommend that data should be coded independently by at least two coders who can then come together to determine their level of agreement on how the data fit within themes. A common discrepancy between coders in using the TDF involves coding the same text to different TDF domains. When discrepancies arise, coders should aim to reach consensus and provide justification for assigning or not assigning a given data item to a domain. In case of failing to reach a consensus, a third person with expertise in using the TDF should be consulted.

Another common problem with coding to the TDF is that some data might fit into multiple TDF domains. In such cases, if the coders fail to reach an agreement on assigning the text to a single domain, it is advised to code the text into all relevant domains of the TDF.

7c. Consider Reliability

Reliability between coders (inter-coder reliability) can be calculated using coefficients such as Cohen’s kappa score and simple percentages. A kappa score greater than 0.6 and percentage greater than 60% across each TDF domain is considered an acceptable level of reliability. Check out this resource for more information on kappa and how to calculate it.

7d. Data Saturation

Data saturation is reached when the addition of new data does not produce any new barriers or facilitators to the target behaviours. Atkins and others recommend concurrently collecting and analyzing data. This way, data collection can be stopped when no new themes emerge, and avoid collecting more data than necessary.

8. Consider How to Report the Findings

There is no recommended method of reporting findings of a TDF-based study. Findings can be reported both through text description [4] (appropriate for qualitative studies to provide detailed information) and in tables [9] (suitable to provide a summary).

Limitations of These Guidelines

The authors note two important limitations to these guidelines: first, they were developed through an informal consensus process, and were based on peer-reviewed publications on the TDF as well as the personal experiences of the group in using the TDF; second, they only pertain to qualitative TDF-based studies, as the TDF has mostly been used in qualitative studies.

Although no formal consensus process was used for developing these guidelines, and their use in quantitative studies has not been looked at, they do provide a useful roadmap for those with little or no experience in using the TDF.

These guidelines are helpful to ensure greater consistency among researchers in applying the TDF in qualitative implementation studies. There is also a need for guidance on how to use the TDF in quantitative studies, and in studies which use TDF along with other implementation frameworks/models.

Let us know your experience with using the TDF, and how you have used it in your implementation research!

References

1. Phillips C, et al. Experiences of using the Theoretical Domains Framework across diverse clinical environments: a qualitative study. J Multidiscip Healthc, 2015; 8:139–146.
2. Atkins L, et al. A guide to using the Theoretical Domains Framework of behaviour change to investigate implementation problems. Impl Sci, 2017; 12:77.
3. Taylor N, et al. The demonstration of a theory-based approach to the design of localized patient safety interventions. Impl Sci, 2013; 8:123.
4. Debono D, et al. Applying the Theoretical Domains Framework to identify barriers and targeted interventions to enhance nurses’ use of electronic medication management systems in two Australian hospitals. Impl Sci, 2017; 12:42.
5. Lynch E, et al. A qualitative study using the Theoretical Domains Framework to investigate why patients were or were not assessed for rehabilitation after stroke. Clinc Rehab, 2016; 31(7):966–977.
6. Thomas S, et al. Use of the Theoretical Domains Framework to Develop an Intervention to Improve Physical Therapist Management of the Risk of Falls After Discharge. Phy Ther, 2014; 11(1):1660–1675.
7. Craig L, et al. Identifying the barriers and enablers for a triage, treatment, and transfer clinical intervention to manage acute stroke patients in the emergency department: a systematic review using the theoretical domains framework (TDF). Impl Sci, 2016; 11:157.
8. Francis JJ, et al. What is an adequate sample size? Operationalising data saturation for theory-based interview studies. Psychol Health. 2010; 25:1229–45.
9. Patey A, et al. Anesthesiologists’ and surgeons’ perceptions about routine pre-operative testing in low-risk patients: application of the Theoretical Domains Framework (TDF) to identify factors that influence physicians’ decisions to order pre-operative tests. Impl Sci, 2012; 7:52.

About the Author

Masood Khan is a Research Coordinator with CHI and the Department of Community Health Sciences at the University of Manitoba.