AI Qualitative Analysis: Lessons from ELLY

Researchers evaluating community health interventions need to identify more than whether participants reported improvement. Researchers also need to understand why an intervention worked, which mechanisms connected activities to outcomes, and where implementation barriers remained. The 2026 ELLY study in two disadvantaged Scottish communities shows how qualitative interview analysis can uncover those pathways. The study examined a 12-week, co-designed healthy weight and wellbeing intervention that combined free soup, community activities, goal setting, and a shopping-card incentive. This article explains what the study found and how AI qualitative analysis can help public health, policy, and academic research teams move from interview transcripts to defensible themes, cross-segment comparisons, and conceptual models. The analysis remains research-focused: participant accounts describe perceived outcomes and pathways, but the study does not establish definitive causation or provide clinical diagnoses.

Key Takeaways

According to the 2026 PLOS ONE study on ELLY, qualitative interviews suggested that wellbeing and social connection were more closely associated with positive behaviour change than healthy weight goals alone.

The ELLY study demonstrates why AI qualitative analysis can be useful for complex interventions: researchers must connect participant experiences, intervention components, mechanisms, barriers, and outcomes without treating reported associations as proof of causation.

According to the ELLY study, 75 residents took part in the 12-week intervention delivered from September 2023 to November 2023.
According to the ELLY study, researchers interviewed 20 participants in December 2023 or later, with 10 participants selected from each of two disadvantaged Scottish communities.
According to the ELLY study, participants could receive a £25 shopping card after attending at least 9 community assets over 12 weeks.
According to the ELLY study, framework analysis identified behaviour change, social connectedness, healthier diets, perceived health improvements, and new skills and knowledge as major themes.
According to the ELLY study, the interview sample had a mean age of 51 years, and 14 of 20 interviewees, or 70%, were female.

How the ELLY Study Used Qualitative Interview Analysis

The ELLY study used semi-structured, face-to-face interviews to explore how participants experienced a community-based healthy weight and wellbeing intervention. Researchers conducted interviews at the end of the 12-week intervention delivered from September 2023 to November 2023, and each interview lasted approximately 30 minutes, according to the ELLY study.

The ELLY intervention combined free soup twice each week, access to local activities and groups, optional goal setting, information resources, self-monitoring, and a loyalty card linked to a £25 shopping card. The ELLY study states that the intervention was co-designed with two disadvantaged communities in Forth Valley, Scotland.

The researchers analysed the interviews using framework analysis. According to the ELLY study, two researchers independently developed an initial coding framework after reviewing five interviews, 10% of the data was independently cross-checked, and disagreements were resolved through discussion with a third researcher.

The ELLY study used the interview findings to develop a conceptual logic model. The logic model separated intervention mechanisms, such as incentives and community connections, from reported outcomes, such as social connectedness and positive behaviour change.

The ELLY authors cautioned that qualitative interview data alone cannot establish definitive causation. The ELLY study described the findings as participants’ perceived pathways to outcomes and recommended longer-term, mixed-methods research.

ELLY Findings Snapshot for AI Qualitative Analysis

Date or study period	Metric	Value	Implication for qualitative analysis
September to November 2023, according to the ELLY study	Intervention duration	12 weeks	Researchers needed to distinguish immediate experiences from sustained behaviour change.
November 2023 or later, according to the ELLY study	Interview participants	20 adults, 10 per community	Cross-community comparison was central to interpreting implementation and experience.
September to November 2023, according to the ELLY study	Total intervention participants	75 residents	The interview sample represented a subset of the wider feasibility study.
September to November 2023, according to the ELLY study	Loyalty-card threshold	9 community assets attended over 12 weeks	The incentive created a measurable engagement pathway that could be coded alongside participant narratives.
November 5, 2026, according to the ELLY publication record	Publication date	2026	The peer-reviewed article provides the primary source for the study methods and findings.

What AI Qualitative Analysis Reveals for Public Health Researchers

Which intervention elements appeared most important to participants?

The ELLY interviews indicated that social connection, wellbeing, local activities, and confidence were central perceived mechanisms of change.

One ELLY participant said, "It’s actually got me going out more, meeting other people, having a laugh, which is important." The quotation shows why thematic analysis should preserve participant language rather than reduce findings to outcome counts.

Why should researchers separate mechanisms from outcomes?

The ELLY study separated mechanisms, such as peer recruitment, community activities, and incentives, from outcomes, such as improved social connectedness and healthier routines.

That distinction helps researchers avoid claiming that a reported outcome was directly caused by one intervention component. AI qualitative analysis can support this work by organizing codes into hierarchical themes and linking recurring concepts across transcripts.

What barriers did the ELLY interviews identify?

The ELLY interviews identified irregular activities, poor communication, limited inclusivity, digital access concerns, and difficulty entering unfamiliar groups as barriers.

One participant described a group as "very specifically orientated to different groups" and said that approaching it was difficult. A robust analysis should preserve these negative and minority cases instead of reporting only the dominant positive narrative.

How can researchers improve the reliability of AI-assisted analysis?

Researchers can improve reliability by defining a coding framework, reviewing representative excerpts, checking a subset independently, documenting disagreements, and retaining an audit trail.

The ELLY study used two researchers to create the coding framework, independently cross-checked 10% of the data, and involved a third researcher in resolving disagreements. AI can accelerate coding and comparison, but researchers remain responsible for interpretation, reflexivity, and ethical handling of sensitive data.

How Evidano Helps With AI Qualitative Analysis

Problem: Researchers need to synthesize many interview transcripts

Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interview transcripts, open-ended survey responses, and research documents.

Evidano can support thematic, content, frequency, and cross-segment analyses, helping research teams move from interview material to a structured view of recurring themes and differences between communities. Researchers can explore the platform’s relevant capabilities on the Evidano features page.

Problem: Community differences can disappear in overall themes

Evidano supports cross-segment analysis, which can help researchers compare findings by community, participant group, demographic characteristic, or intervention exposure.

For a study such as ELLY, cross-segment analysis can help distinguish themes shared across both communities from themes shaped by local context, while keeping the interpretation grounded in the source material.

Problem: Coding needs to remain transparent and reviewable

Evidano provides hierarchical visualizations that can display relationships between codes and subcodes, alongside co-occurrence networks and word clouds.

These visual outputs can help researchers inspect whether concepts such as social connection, confidence, activity access, and wellbeing appear together across documents. Visualizations should support, not replace, close reading and methodological review.

Problem: Sensitive health interviews require careful data handling

Evidano offers transcription with PII redaction and states that data is encrypted and never used to train third-party models.

These controls are relevant to public health and community research teams handling interviews that contain information about mental health, disability, weight, income, or other sensitive experiences. Researchers should still apply their own consent, governance, retention, and access requirements.

FAQ: AI Qualitative Analysis for Community Health Studies

What is AI qualitative analysis?

AI qualitative analysis uses artificial intelligence to help researchers organize, code, compare, and interpret text-based research data.

In the ELLY study, the relevant data consisted of semi-structured interviews from 20 participants, and the researchers used framework analysis to identify themes and develop a logic model. AI tools can assist similar workflows, but human researchers must validate interpretations.

What did the ELLY qualitative study find?

The ELLY qualitative study found that participants associated the intervention with positive behaviour changes, stronger social connections, healthier dietary habits, perceived health improvements, and new skills and knowledge.

The ELLY authors reported that wellbeing and social connection appeared more motivating than healthy weight alone, while also stating that the qualitative findings did not establish definitive causation.

How many people were interviewed in the ELLY study?

The ELLY study interviewed 20 participants, with 10 participants from each of two disadvantaged Scottish communities.

According to the ELLY study, the 20 interviewees represented 27% of all intervention participants, while 75 residents took part in the wider feasibility intervention between September 2023 and November 2023.

Can AI prove that a health intervention caused an outcome?

AI qualitative analysis cannot prove that a health intervention caused an outcome when the underlying study design cannot establish causation.

The ELLY study explicitly described the findings as perceived pathways and called for longer-term mixed-methods research to assess whether reported behaviour changes were sustained.

How should researchers use AI with sensitive interview data?

Researchers should use AI with sensitive interview data only when the tool’s security, privacy, consent, access, and retention arrangements meet the study’s governance requirements.

Evidano provides encrypted data handling and PII redaction, but research teams remain responsible for ethical approval, participant consent, data minimization, and institutional compliance.

Conclusion and Next Steps for AI Qualitative Analysis

The 2026 ELLY study shows how qualitative interviews can explain the mechanisms, motivations, barriers, and contextual differences behind a community health intervention. The study’s 20 interviews generated a logic model linking social connection, confidence, local activities, healthier routines, and perceived wellbeing outcomes, while also identifying limits to participation and communication. AI qualitative analysis can make this type of synthesis more efficient by supporting thematic coding, cross-segment comparison, document chat, and visual exploration without removing researcher judgment. Public health and academic teams can use Evidano to organize complex qualitative datasets while maintaining a transparent, research-focused workflow. Try Evidano for free to explore AI-assisted qualitative analysis.

Topics

AI qualitative analysis
AI thematic analysis
qualitative health research
community intervention research
qualitative interview analysis

Keep reading

Browse all articles