What ELLY Reveals: AI Thematic Analysis

According to the PLOS One study on the ELLY intervention, qualitative interviews suggested that wellbeing and social connection were more closely associated with positive behaviour change than weight loss alone during the 12-week intervention delivered from September to November 2023.

The ELLY researchers analyzed interviews with 20 participants from two disadvantaged Scottish communities and developed a conceptual logic model connecting intervention components with perceived outcomes. AI thematic analysis can help researchers reproduce this kind of structured synthesis across larger collections of interviews, while human researchers retain responsibility for interpretation, context, and ethics.

• The ELLY intervention reached 75 residents across two Forth Valley communities during 12 weeks from September to November 2023.
• The qualitative component interviewed 20 participants, with 10 participants from each community, at 12 weeks.
• The ELLY study identified interconnected themes including social connectedness, positive behaviour change, healthier diets, perceived mental and physical health improvements, and new skills and knowledge.
• The ELLY intervention included free soup twice weekly, community-asset engagement, optional goal setting, information resources, and a £25 shopping card for participants who earned at least 9 loyalty-card stamps over 12 weeks.
• The authors reported that qualitative evidence alone could not establish definitive causation and recommended longer-term and mixed-methods research.

The ELLY study used semi-structured interviews and framework analysis to understand how participants perceived the intervention and its possible pathways to health and wellbeing outcomes. Researchers conducted approximately 30-minute, face-to-face interviews at participants’ homes or community settings in 2023, using encrypted audio recording and a predefined topic guide.

The ELLY research team recruited 35 participants in community C1 and 40 participants in community C2 during the feasibility study. The interview sample included 14 women out of 20 participants, while 17 of the 20 interview participants lived in areas classified as disadvantaged by Scottish Index of Multiple Deprivation quintiles 1 to 3.

Two researchers independently developed an initial coding framework after reviewing five interviews, and a second researcher independently cross-checked 10% of the coded data. The team resolved disagreements through discussion and reviewed emerging themes in weekly meetings.

The researchers distinguished intervention mechanisms from outcomes. For example, snowball recruitment and community connections were treated as mechanisms, while social connectedness and positive behaviour change were treated as outcomes. The resulting logic model showed that the pathways were complex, interconnected, and not always clearly causal.

One participant told the ELLY researchers, “It’s actually got me going out more, meeting other people, having a laugh, which is important, ” linking participation with social interaction rather than describing weight as the only meaningful outcome. Another participant said, “I feel the ELLY Project was very helpful for me…I did a lot of activities, achieved my goal, which was my weight… Now I walk every night, ” illustrating how participants connected activities, goals, and daily routines.

Community health researchers should analyze intervention experiences as connected pathways rather than as isolated themes. The ELLY study linked local activities, social contact, confidence, dietary routines, and physical activity, showing why a single frequency count cannot fully explain how a complex intervention is experienced.

Researchers should separate participant-reported outcomes from causal conclusions. The ELLY authors stated that “this qualitative data alone is insufficient to establish definitive causation, ” so thematic analysis should clarify perceived mechanisms without presenting perceptions as proof of clinical effectiveness.

Researchers should preserve negative and divergent cases during AI-assisted coding. The ELLY study found that some participants felt excluded by groups, found activities irregular, or considered goals unattainable because of wider life circumstances. A robust analysis must retain these counterexamples instead of allowing dominant positive themes to obscure barriers.

Researchers should compare segments such as community, age, engagement level, health context, and goal-setting experience. Cross-segment analysis can show whether a theme is shared broadly or concentrated among particular participants, while quotations and transcript context help researchers assess meaning.

Public health teams should treat AI outputs as an auditable starting point rather than a final interpretation. The ELLY team used multiple coders, disagreement resolution, field notes, reflexivity, and team review, and AI-assisted workflows should support comparable oversight.

The ELLY study shows that qualitative analysis can reveal how community interventions are experienced through interconnected pathways involving wellbeing, social connection, confidence, diet, and daily behaviour. The study’s 20 interviews, framework analysis, and logic model demonstrate the value of combining detailed participant accounts with structured interpretation.

AI thematic analysis can make this workflow more efficient by organizing transcripts, comparing segments, surfacing supporting excerpts, and helping researchers examine complex relationships. AI does not replace ethical judgment, contextual interpretation, or the need to distinguish participant perceptions from causal evidence.

For public health, UX, policy, and academic teams working with interview data, Evidano provides a practical environment for moving from raw qualitative evidence to reviewable themes and cross-segment findings. Try Evidano for free to analyze your qualitative research data with greater structure and speed.