State of the Art

A total of six (6) State- of-the-Art submissions have been chosen for presentation during the EHPS 2026, through the submission and review process. You can see the titles, authors and abstracts below

Author:

Abstract:
Over 10 million people in Europe have a visible difference (an appearance that differs to ‘the norm’ due to injury, acquired or congenital conditions, or treatment) including burns, craniofacial conditions, alopecia and skin conditions. The psychosocial impact can be extensive and enduring, yet health psychology can make a significant difference to the lives of those affected and their families/carers.

Research exploring the generic and condition-specific challenges facing people impacted by visible differences has repeatedly shown that some could benefit from appropriate support. However, systematic reviews (e.g., Bessell & Moss, 2007; Waite et al., 2024) report a dearth of appropriate readily available, evidence-based and rigorously evaluated interventions. In addition, there is a pressing need to address societal stigma and negative stereotypes, and for better media representation of visible difference.

This presentation will showcase a 9-year programme of applied health psychology research, driven by the priorities of 25 UK charities and conducted in close collaboration with these support organisations and people with relevant lived experience, to develop and evaluate a novel range of acceptable and accessible evidence-based interventions and resources. This includes an e-book, an app, a board game and guidance on social media use. These are now freely available through an easily accessible website (www.VisibleDifferenceSupportHub.com) developed in collaboration with the charities, and where new resources will continue to be added in the future.

This presentation will highlight the value of genuinely involving charitable support organisations and people with relevant lived experience at all stages of the research, from inception to implementation, and how this collaborative approach underpinned the success and impact of this major programme of work. Challenges encountered and overcome along the way will be shared in this illustration of how health psychology can make a real impactful difference to increase understanding and help address the needs of under-represented groups.

Abstract: 
Ontologies—formal, structured frameworks that represents knowledge within specific domains—offer powerful tools for achieving precision in health psychology. By providing standardised vocabularies and explicit relationships between concepts, ontologies facilitate unambiguous communication, data integration and evidence synthesis across studies. They also provide a means of linking knowledge across topic and disciplinary silos. However, their technical complexity and steep learning curve create barriers to adoption by researchers and practitioners.
This talk will discuss findings from a usability study on the Behaviour Change Intervention Ontology (BCIO) that identified barriers to its use. It will use this as a stepping off point for setting out an approach to democratising ontology use through user-centred design. Rather than expecting users to navigate complex ontological structures or master formal logic, the talk will present tools that start from users’ existing conceptual frameworks and language about behaviour and behaviour change, and help link these to the BCIO in a digestible format. The approach involves three key stages: (1) structured prompts that guide users to articulate their research constructs in their own terminology, (2) generative AI that maps user descriptions to candidate ontology classes by analysing semantic similarity and contextual fit, and (3) interactive interfaces in which users can review recommended matches and make informed choices. The system will then generate annotated documents using selected ontology classes, creating machine-readable outputs that support data integration, automated reasoning, and precise knowledge representation.
We illustrate this approach with a user-friendly interface (CACTISS 2.0) that helps engage with the Behaviour Change Technique Ontology and the Mechanism of Action Ontology within the Behaviour Change Intervention Ontology to develop behavioural interventions. By reducing technical barriers whilst preserving ontological rigour, such tools can accelerate the adoption of precision science practices in health psychology. The talk will present opportunities for community engagement in developing these resources.

Authors:

Abstract: 
Background. Interview and focus group analysis requires systematic coding, often by multiple researchers to enhance reliability and reduce subjectivity. This process is resource-intensive and may delay research progress. Large Language Models (LLMs) are increasingly used to support research tasks, yet their suitability for qualitative analysis remains contested. Some argue that interpreting qualitative data is inherently human and cannot be outsourced. We empirically examined whether an LLM can deductively code interview transcripts and how its results compare to human coding.

Methods. We used OpenAI’s GPT-4o (ChatGPT) in an isolated environment to analyze 25 interviews previously coded by two researchers using a Health Belief Model–based codebook. The model received the transcripts and the codebook, along with 5–15 few-shot examples from one coded transcript. Through prompt engineering on 14 uncoded transcripts, we developed the final prompt over 46 iterations, which was then applied to 10 unseen transcripts.

Findings. Across the 10 transcripts, the LLM generated 756 coded fragments. Of these, 224 (30%) matched the human coding exactly. In 280 cases (37%), the LLM assigned a different code, and in 252 cases (33%), it coded fragments not coded by humans. Expert assessment showed that 220 (29%) of the 280 deviating codes were useful alternatives to those assigned by the human coders, and 168 (22%) of the 252 additional codes were also useful. Overall, 612 codes (81%) were either identical to or considered useful relative to the human benchmark.

Discussion. Although the LLM did not fully replicate human coding, the vast majority of its codes were meaningful. Discrepancies also exposed ambiguities in the codebook. LLMs may therefore not only accelerate qualitative analysis but also stimulate critical reflection and improve coding quality. This study suggests that LLMs are promising, cutting-edge tools for qualitative research. Implications, safeguards, and future directions will be discussed.

Authors:

Abstract: 
While cigarette smoking has declined substantially across many Western nations over recent decades, this public health success is increasingly threatened by the rapid uptake of new nicotine delivery formats — particularly e-cigarettes and nicotine pouches — among young people. Understanding why young people initiate and continue using these products is critical for developing effective prevention strategies, yet the psychological determinants of new nicotine use have remained largely untested until recently.
This presentation synthesises findings from a programme of research conducted across Australia, California, and Finland (total N > 2,000), examining social-cognitive predictors of vaping, pouch use, and cigarette smoking in young people aged 15–25. Drawing on an augmented Theory of Planned Behaviour framework, studies incorporated affective and instrumental attitudes, subjective and descriptive norms, perceived behavioural control, habit, implicit attitudes, and nicotine dependence as predictors of nicotine use intentions and behaviour.
Across all contexts and nicotine products, affective (enjoyment-based) attitudes emerged as the most consistent predictor of use, while harm perceptions and instrumental attitudes showed negligible effects. Social norms predicted initial uptake but not frequency of continued use, where habit and dependence became the dominant drivers. Latent profile analysis identified distinct smoker like vaping subtypes, constant, social, and sporadic vapers, characterised by different cue-behaviour relationships. Hurdle modelling further revealed that the determinants of ever trying a product differ meaningfully from those driving frequent use. Comparative analyses across Finland, Australia, and the US indicate these patterns are largely consistent across regulatory contexts, despite substantial differences in nicotine policy.
These findings challenge the dominant public health approach of targeting harm perceptions and suggest that prevention efforts must also focus on affective attitudes and the short-term experiential consequences of nicotine use.

Authors:

Abstract: 

Background

In intervention development, determinants from behavioural theories, mechanisms of action (MoA), and associated behaviour change techniques (BCTs) specify what should influence behaviour, but not how they become deliverable interventions. A translational gap remains for development regarding content, design, and delivery context, requiring a systematic framework.

Methods

We present a framework, illustrated across multiple behavioural domains (including bowel cancer screening, vaping, condom use negotiation), populations, and targeted MoAs:

Step 0. Identify MoAs and BCTs relevant to the population and behaviour
Step 1. Choose feasible mode of delivery (this might take place after Step 2).
Step 2. Translate each (combination of) selected BCTs into message content
Step 3. Design messages to align with mode of delivery
Step 4. Refine content collaboratively while maintaining theoretical fidelity
Step 5. Evaluate messages proof-of-principle randomised controlled trials (PoP RCTs).
Step 6. Select messages shown to change MoAs
Step 7. Implement

Findings

Using the framework enabled translation of BCTs into deliverable messages in 8 illustrations. Many decisions (steps 1-3) shape intervention effectiveness. For example, translating BCT 5.1 ‘Information about health consequences’ allows alternative framings that may differentially influence MoA. Linguistic and contextual input supported population-appropriate messaging. Creative delivery options, including comics, require adaptation beyond written text.
Refined messages (step 4) required researcher and implementer judgements, followed by empirical evaluation assessing change in targeted MoA. Across illustrations, some but not all messages demonstrated measurable change in intended MoA resulting in rejection of some messages.

Discussion

This framework clarifies the translation of BCTs into deliverable interventions with evidence-based messages suitable for implementation. It positions MoA change, assessed via PoP RCTs, as an early indicator of intervention promise and implementation feasibility. The approach functions as a methodological bridge between experimental research, theory-based identification of BCTs and MoAs and large pragmatic trials, supporting cumulative development of scalable prevention interventions.

Author:

Abstract: 
Background: Behavioural science in health policy has been dominated by behavioural insights approaches targeting individual-level influences on diet, physical activity, and other health behaviours. Results have been “disappointingly modest,” prompting calls to move from an individual-based i-frame towards a systems-based s-frame. Yet this shift risks losing individual agency entirely. I introduce a complex systems frame (cs-frame) that relieves this tension, and conceptualises behaviour change as shifts between stable states known as attractors. Attractors have been argued to remedy linear thinking when it is unwarranted, and work as unifying concepts across scales.

Methods: The primary aim is theoretical: to articulate the cs-frame and its implications for health psychology. Two datasets serve as proof-of-concept illustrations. First, an N-of-1 dataset containing 381 weekly measures of occupational health (work motivation, basic needs satisfaction, work engagement) was assessed for non-linear dynamics using stationarity, linearity, and surrogate data tests. Second, a cross-sectional dataset of 26,097 participants’ civil health emergency preparedness items (e.g. food and water suplies, plans with family and neighbours, battery-operated radio) was analysed with CatBoost and vector embedding. These models operationalised personalised recommendations for reaching adjacent attractor states.

Findings: In the time series data, most or all of the 60 variables depicted non-stationarity and non-linearity (p < 0.05), consistent with attractor shifts rather than linear change. The “adjacent possible” attractor proved conceptually useful for producing next-step recommendations at varying scales, from individuals to municipalities.

Discussion: The cs-frame resolves the i-frame versus s-frame tension by treating them as ends of a spectrum. If the scope of investigation is a nation, subsystems contain individuals, neighbourhoods, and municipalities – akin to how years contain days, weeks, and months. Attractors can be understood on all of these scales, and adjacent attractors act as achievable next steps. This perspective reinforces working across scales instead of choosing between them.