This post explains how AI-enabled qualitative analysis can accelerate, scale, and deepen thematic research into help-seeking among people in contact with the criminal justice system. The primary keyword is "qualitative analysis of help-seeking" and this article is written for qualitative researchers, implementation teams, and evaluation leads who need fast, rigorous syntheses of interview and network data. The payoff: concrete methods you can adopt now, extractable statistics and quotes from the PLOS One study, and practical mappings to Evidano workflows that preserve security and researcher control.
Key Takeaways
According to the PLOS One article by Connell et al. (published July 30, 2026), help-seeking for mental health and substance use among people supervised in the community emerges from interacting individual desire, individual ability, and the relational and contextual help-seeking environment.
- The PLOS One study interviewed 50 people between 1 November 2023 and 31 July 2024 and analysed transcripts with a framework approach, confirming relational influences shape help-seeking (50 interviews, Connell et al., 2026).
- Connell et al. (PLOS One, July 30, 2026) recruited 23 participants in LA1 and 27 participants in LA2, showing contextual differences between a more rural LA1 and a deprived urban LA2.
- The PLOS One article highlights prior prevalence estimates cited in the literature: over 40% meeting criteria for a substance use disorder, more than 10% for major depression or psychosis, and 75% for any personality disorder in comparable justice-involved groups (as cited by Connell et al., 2026).
- The PLOS One authors conclude, in their words, that “help-seeking is not an individual behaviour, but strongly affected by relational influences that operate between individuals, across social networks, and via cultural norms” (Connell et al., 2026).
What happened and how the study was measured
The PLOS One study collected 50 semi-structured, audio-recorded interviews and used framework analysis to identify themes of desire, ability, and context affecting help-seeking (Connell et al., 2026).
According to Connell et al. (PLOS One, 2026), interviews were conducted from 1 November 2023 to 31 July 2024, transcribed verbatim, pseudo-anonymised, and coded iteratively using Nvivo with independent double-coding and matrix summaries to ensure consistency.
The PLOS One team combined qualitative interviews with participant-aided social network diagrams to foreground relational and network-level influences that are often missed in survey-only designs.
Direct quotations from the study illustrate lived experience: Participant P13 told researchers, "you just kind of can’t get away from the social, every time you try and do something good basically they will get you on it and you will be wrecked and miss what you are going to do" (P13, quoted in Connell et al., PLOS One, 2026).
Findings Snapshot
| Date / Source | Metric | Value | Implication |
|---|---|---|---|
| 1 Nov 2023–31 Jul 2024 (PLOS One) | Interviews completed | 50 | Rich qualitative sample enabling framework analysis across two local authorities |
| July 30, 2026 (PLOS One) | Geographic split | 23 (LA1) / 27 (LA2) | Contextual contrasts: rural/dispersed services vs urban deprivation |
| Connell et al. citing prior literature (PLOS One, 2026) | Substance use disorder prevalence | Over 40% (in comparable justice-involved samples) | High clinical need, requiring network- and community-level interventions |
Implications for qualitative researchers and evaluation teams
Multi-level, relational analysis should be the default when studying help-seeking in justice-involved populations, because the PLOS One study shows relational norms and resources alter both motivation and capacity to act (Connell et al., 2026).
- Design: According to Connell et al. (PLOS One, 2026), include participant-aided network mapping alongside interviews to capture shared norms and resource flows; the PLOS One team used Network Canvas for this purpose.
- Sampling: The PLOS One authors recruited across contrasting localities (rural LA1 and urban LA2) and recommend purposive recruitment to surface context-specific barriers; the team planned capacity for 60 interviews and concluded after 50 with data richness achieved.
- Analysis: Connell et al. (PLOS One, 2026) used framework analysis with independent double-coding and matrix summaries; qualitative teams should document iterative codebook changes and cross-check at the case-by-theme matrix level.
How Evidano helps: AI-enabled workflows for relational qualitative analysis
Problem: Large interview sets, slow synthesis
Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents.
According to the PLOS One study, teams benefit from rapid cross-case synthesis when examining networked influences (Connell et al., PLOS One, 2026); Evidano automates initial thematic extraction so analysts can focus on interpretation.
Problem: Capturing relational context from sociograms and transcripts
Connell et al. (PLOS One, 2026) show participant-aided network diagrams revealed norms and resource flows that shaped help-seeking; Evidano ingests transcripts and spreadsheeted network attributes, enabling co-occurrence and cross-segment analyses that link quotes to network roles.
Feature mapping: Use Evidano’s thematic, content frequency, and cross-segment analysis to tag quotes by network role, visualize co-occurrence between themes and network positions, and export case-by-theme matrices for reporting. See Evidano features for details.
Problem: Secure, reproducible analysis with sensitive data
Connell et al. (PLOS One, 2026) deposited pseudonymised transcripts in the UK Data Service under controlled access because of sensitivity; Evidano supports encrypted storage and PII redaction during transcription and analysis to match ethical safeguards.
Workflow tip: Combine Evidano’s transcription and PII redaction with coded audit trails so ethics boards can verify who accessed what and when.
FAQ: qualitative analysis of help-seeking
What is the single best method to surface relational drivers of help-seeking?
Answer: Combine interviews with participant-aided social network mapping and a framework analysis, because the PLOS One study shows this approach reveals how network norms and resources shape both desire and ability to seek help (Connell et al., PLOS One, 2026).
Supporting detail: Connell et al. used Network Canvas for mapping and then applied framework analysis to link network features to thematic codes across 50 interviews.
How many interviews do I need to detect network-level patterns?
Answer: There is no fixed number, but the PLOS One team achieved substantive network-level insights with 50 interviews conducted across two contexts (Connell et al., PLOS One, 2026).
Supporting detail: The PLOS One authors planned capacity for 60 interviews, sampled purposively, and stopped at 50 when they judged data richness sufficient for their mixed-methods goals.
Can AI safely accelerate coding of sensitive interview transcripts?
Answer: Yes, when AI tools are configured for privacy, PII redaction, and researcher-in-the-loop validation; the PLOS One study deposited pseudonymised data under controlled access because of sensitivity, demonstrating the need for strong safeguards (Connell et al., PLOS One, 2026).
Supporting detail: Use AI to propose initial codes and cluster quotes, then have trained analysts validate and refine themes to preserve trustworthiness.
How do I turn a relational framework into intervention targets?
Answer: Map themes to levels of intervention (individual, network, community) and prioritise modifiable factors identified across cases, as the PLOS One authors recommend using their framework to design multi-level interventions (Connell et al., PLOS One, 2026).
Supporting detail: For example, target network knowledge gaps with co-produced communications, and support services to adopt trauma-informed relational practice to rebuild trust.
Conclusion & Next Steps
The PLOS One study (Connell et al., 2026) demonstrates that qualitative analysis which makes relational context explicit changes how we understand help-seeking for people in contact with the criminal justice system.
AI-enabled qualitative workflows shorten the path from raw transcripts and sociograms to an actionable framework by automating initial coding, linking quotes to network roles, and producing cross-segment frequency reports that researchers can validate.
If you run qualitative studies with sensitive interviews and network data, adopt multi-method designs and use tools that support PII redaction and encrypted storage to meet ethical requirements (the PLOS One team used controlled deposition for pseudonymised data).
To try an AI-enabled qualitative analysis workflow that supports thematic, cross-segment, and network-linked analysis, Try Evidano for free.