Faster APC Qualitative Analysis: APCs & the Trap

Evidano is an AI-powered qualitative data analysis platform that helps teams reproduce the APC Trap analysis and accelerate reproducible, auditable thematic coding. The APC Trap paper (published July 1, 2026) shows how even well-resourced U.S. researchers feel constrained by article processing charges. This post maps the study’s methods and numbers to a 7-step pipeline you can run with Evidano to save hours on coding, surface co-occurrence networks, and produce stakeholder-ready visuals while preserving auditability and data security. Target audience: librarians, qualitative analysts, research managers, and policy teams who handle transcripts, survey free-text, or codebooks.

Key Takeaways

Evidano enables teams to reproduce the APC Trap thematic analysis quickly and securely using AI-assisted inductive coding, coder-agreement workflows, and exportable, auditable outputs. Cantrell et al. (published July 1, 2026) analyzed 154 substantive open-ended responses from 322 completed surveys and identified an APC Trap consisting of eight interacting qualitative codes. The study shows high prevalence of APC-related concerns among R1 authors and demonstrates that mixed sentiments require nuance beyond off-the-shelf sentiment tools.

154 substantive free-text responses analyzed from a 2023 survey (322 completed surveys total).
Eight APC Trap codes identified, with 139 responses (90%) showing at least one code.
Cantrell et al. published the study on July 1, 2026 in PLOS ONE; the dataset highlights mixed, co-occurring tensions such as Financial Concern and Restriction of Decisions.

Findings snapshot

Metric	Value	Source / Note
Survey period	Oct 31–Nov 21, 2023	Qualtrics recruitment window
Respondents (completed substantive portion)	322	Study sample across 4 R1 institutions
Substantive free-text responses analyzed	154 (47.8%)	Two open-ended questions aggregated per respondent
APC Trap codes identified	8 codes	Inductive thematic coding; consensus if 4+ coders agreed
Responses with ≥1 APC Trap code	139 (90%)	Shows prevalence among R1 authors
Common co-occurrence patterns	Financial Concern ↔ Restriction of Decisions (37%)	Cross-tab analysis in paper
Published	July 1, 2026	PLOS ONE (open access)

Fast take: what the paper found (and where to read it)

Cantrell et al. (published July 1, 2026) analyzed 154 substantive open-ended responses and identified an “APC Trap” composed of eight interacting qualitative codes. The APC Trap codes are financial concern, diversity/equity, threshold of reasonableness, restriction of decisions, grants don’t cut it, academic obligation/expectation, academic labor, and quality/rigor, and 90% of substantive responses showed at least one code (n = 139). Read the original paper on PLOS ONE: PLOS ONE.

Dataset highlights in the study: 322 completed surveys, 154 substantive free-text responses (47.8%).
Key tension reported: authors can be both price-sensitive and price-indifferent, creating cognitive dissonance and constrained publishing choices.

How the original study coded free text (methods in plain English)

The study used an inductive thematic approach to code two aggregated free-response questions per respondent (n = 154). Analysts built a Code Book from recurring patterns, then independently applied eight yes/no codes per respondent, requiring agreement by at least four researchers or group consensus for disputed cases. The team also tested automated sentiment tools (NVivo Auto Sentiment, SentiStrength) but found those tools unreliable for nuanced, ambivalent responses.

Unit of analysis was the whole response per participant, with multiple codes allowed.
Consensus procedure involved iterative codebook refinement and Excel-based agreement checks.
A limitation flagged by the authors was that leading question framing may have biased responses toward financial topics.

So what for qualitative teams: implications from the APC Trap

For librarians & policy analysts

Librarians and policy analysts should expect APCs to be both practical and ethical problems because even authors at R1 institutions report financial strain. The study notes that some institutions (UMass) had notably higher financial concerns, and advisors should anticipate mixed sentiments where authors who can pay still raise equity objections. Action item: prioritize cross-segment comparisons (discipline, institution, OA type) to show where subsidies or transformative agreements matter most.

For qualitative researchers & UX teams

Qualitative researchers and UX teams should treat open-ended survey questions as producing layered, mixed sentiments that demand human-informed methods. The authors found that off-the-shelf sentiment scores mislabel sarcasm or compound views, therefore manual thematic coding or AI tuned for qualitative nuance is necessary. Action item: combine inductive codebook development with cross-segment frequency and co-occurrence analyses to surface the most policy-relevant narratives.

For PIs & grant managers

Principal investigators and grant managers should expect grants often not to cover APCs fully and to experience timing mismatches that constrain publishing decisions. The study documents constraints under codes labeled 'Restriction of Decisions' and 'Grants Don’t Cut It'. Action item: quantify APC burden per PI/year and present explicit trade-offs (for example, trainee support versus APCs) in budget requests or institutional subsidy proposals.

Do more, faster with Evidano (map to the APC Trap workflow)

Problem: messy, multilayered free text

Messy, multilayered free text requires preserving respondent metadata and running AI-assisted inductive code discovery; Evidano ingests the full spreadsheet of responses, preserves respondent IDs and metadata (institution, discipline, OA type), and runs AI-assisted inductive code discovery to suggest initial codes while keeping humans in the loop.

Problem: inconsistent coding & long reconciliation

Inconsistent coding and long reconciliation slow projects, so Evidano lets teams import a master Code Book, run AI-assisted coding across all responses, then review and lock codes; Evidano shows coder agreement and highlights low-consensus items for focused human adjudication.

Problem: weak sentiment and context-blind tools

Weak sentiment and context-blind tools mislabel rhetorical structures, so Evidano uses LLMs tuned for qualitative research to avoid naive polarity errors and to annotate rhetorical structures like irony or conditional statements.

Problem: need to compare segments (HSS vs NSE, Gold vs Hybrid)

Comparing segments requires structured cross-tab and co-occurrence analyses, and Evidano runs cross-segment frequency and co-occurrence analyses, exports network visuals (co-occurrence networks, hierarchical code to subcode maps), and produces stakeholder-ready dashboards.

Security & governance

Security and governance require enterprise-grade controls, and Evidano provides enterprise-grade encryption and a strict policy that your data is never used to train third-party models, so teams can analyze grant-funded or sensitive research safely.

7-step workflow: reproduce the APC Trap analysis in Evidano

This 7-step checklist shows how to reproduce the APC Trap analysis in Evidano and move from raw survey export to a leadership brief.

1) Import: upload spreadsheet (responses plus metadata) and tag institutions, discipline, OA type.
2) Preprocess: normalize text, preserve punctuation for rhetorical cues, optionally run transcription or translation if needed.
3) Exploratory pass: run AI-assisted code suggestion to surface candidate themes, seeding with codes such as 'financial concern' and 'diversity'.
4) Build Code Book: merge AI suggestions with manual codes; import or export the Code Book as CSV for versioning.
5) Apply & adjudicate: auto-code the corpus, review low-consensus items, and finalize coding with inter-coder agreement metrics.
6) Analyze: produce frequency tables, co-occurrence networks, and cross-segment comparisons (for example, HSS vs NSE; Gold vs Hybrid).
7) Report: generate visuals and an executive brief including redacted quotes with respondent IDs removed, and export reproducible outputs for audit.

FAQ: qualitative analysis of APCs

Can automated sentiment be trusted for APC debate?

No, automated sentiment should not be trusted alone for APC debates because Cantrell et al. found NVivo Auto Sentiment and SentiStrength unreliable for nuanced, ambivalent responses. Use sentiment as an initial guide but validate with thematic coding and context-aware AI tuned for qualitative nuance.

How do I compare Gold versus Hybrid authors?

Tagging by OA type at import allows direct comparison, so tag by OA type and run cross-segment frequency and co-occurrence analyses to surface distinct patterns; the paper found Hybrid authors showed more 'Grants Don’t Cut It' codes.

Is this analysis reproducible?

Yes, reproducibility requires versioned codebooks and exportable assignments and metrics; keep codebook versions, export code assignments and agreement metrics, and store raw and processed outputs to ensure reproducibility. Evidano supports exportable reports and reproducible pipelines.

Wrapping up: next steps

To turn the PLOS ONE APC Trap dataset or your own survey/transcript corpus into transparent, reproducible themes and stakeholder-ready visuals, follow the 7-step workflow above and use Evidano to cut weeks of manual work. The study is available to download for methods and figures on PLOS ONE: PLOS ONE.

Download the Cantrell et al. study and inspect methods and figures on PLOS ONE: PLOS ONE.
Try the 7-step workflow on a pilot subset and measure time-to-insight versus manual coding.

Ready to test this on your corpus? Get a demo and a hands-on trial at Evidano and sign up here: Try Evidano for free.

Topics

qualitative analysis of APCs
APC Trap
APC open-ended responses
thematic coding
Evidano

Keep reading

Browse all articles