Primary keyword: physician perceptions of LLMs. According to the PLOS One protocol by Pokharel et al., published on August 7, 2026, researchers designed a qualitative study to map how family physicians perceive adoption of large language models in clinical care. The introduction below summarizes the protocol’s methods, key numbers, and immediate implications for qualitative teams who want to use AI-enabled tools to accelerate interview collection, transcription, coding, and synthesis.
Key Takeaways
Answer: The PLOS One protocol (Pokharel et al., 2026) lays out a descriptive qualitative study using the Technology Adoption Behavior (TAB) framework to interview family physicians in British Columbia about LLM adoption and to produce thematic findings that will inform policy, practice, and training.
- Published on August 7, 2026, the protocol plans semi-structured interviews with purposively sampled family physicians in British Columbia, according to PLOS One.
- The protocol defines LLM usage groups as frequent (50+ uses), occasional (5–50 uses), and nearly no use (<5 uses) in the context of health care, as stated in the PLOS One methods section.
- The study expected recruitment and data collection between August 1, 2025 and January 31, 2026, and notes timelines for completion in the protocol (Pokharel et al., 2026).
- The protocol highlights system-level context, for example that the number of Canadians without a family doctor rose from 4.5 million in 2019 to about 6.5 million in 2023, which the authors cite as motivation for focusing on primary care.
What Happened: PLOS One protocol and its methods
Answer: Pokharel et al. published a protocol in PLOS One on August 7, 2026 that specifies a descriptive qualitative study using semi-structured interviews and the Technology Adoption Behavior (TAB) framework.
According to the PLOS One protocol (Pokharel et al., 2026), the study will purposively sample practicing family physicians in British Columbia and stratify by self-reported LLM use frequency (50+, 5–50, <5 uses).
According to the PLOS One protocol (Pokharel et al., 2026), interviews will be recorded, transcribed verbatim, deidentified, and analyzed using thematic analysis with both deductive TAB-informed codes and inductive codes.
According to the PLOS One protocol (Pokharel et al., 2026), the multidisciplinary team will use reflexive memos and monthly codebook review, and they plan to analyze data in NVivo.
Findings snapshot from the protocol
| Date / Timeline | Metric / Item | Value / Detail | Implication |
|---|---|---|---|
| August 7, 2026 | Protocol published | PLOS One | Publicly available study design to inform similar qualitative projects |
| August 1, 2025 to January 31, 2026 | Planned recruitment window | 6 months anticipated (Pokharel et al., 2026) | Useable timeline for project planning and resource allocation |
| 2019 and 2023 | Primary care access context | 4.5 million without a family doctor in 2019; about 6.5 million in 2023 (Pokharel et al., 2026) | Motivates focus on primary care workflow and access implications |
| Usage thresholds (no date) | LLM experience strata | Frequent: 50+; Occasional: 5–50; Nearly none: <5 (uses) | Enables purposive sampling across experience levels |
Implications for qualitative researchers studying LLM adoption
How should teams choose participants for an LLM perceptions study?
Answer: Teams should purposively sample across LLM experience and years in practice to capture diverse perspectives, as recommended in the PLOS One protocol (Pokharel et al., 2026).
According to Pokharel et al. (2026), the protocol’s strata (frequent (50+ uses), occasional (5–50 uses), and nearly none (<5 uses)) provide a clear recruitment heuristic to ensure contrasting experiences in interviews.
According to the PLOS One protocol (Pokharel et al., 2026), recruiting through professional organizations and snowball sampling helps reach clinicians who both use and avoid LLMs.
What interview approach yields the richest data on adoption behavior?
Answer: Semi-structured interviews guided by the TAB framework yield both breadth and depth, according to the PLOS One protocol (Pokharel et al., 2026).
According to Pokharel et al. (2026), the TAB framework anchors questions in perceived usefulness, perceived ease of use, and task-technology fit, which the authors expect will clarify why clinicians adopt or reject LLMs.
How Evidano Helps: AI-enabled qualitative workflows for physician LLM perception studies
Problem: Long transcription and deidentification steps
Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents.
According to the PLOS One protocol (Pokharel et al., 2026), interviews will be recorded and transcribed verbatim; Evidano’s transcription features can accelerate that step and support custom dictionaries and PII redaction via the speech-to-text tool.
Problem: Managing both deductive TAB codes and inductive themes
Answer: Evidano’s thematic coding and hierarchical code management reduces manual overhead when mixing deductive and inductive approaches.
According to Pokharel et al. (2026), the study plans deductive codes based on the TAB framework and iterative inductive coding; Evidano’s thematic and cross-segment analyses map directly to that workflow and let teams compare code frequencies and co-occurrence across LLM experience strata.
Problem: Rapid synthesis for policymakers and clinician audiences
Answer: Evidano’s AI chat over documents and visualization outputs speed synthesis into policy and practice briefs.
According to Pokharel et al. (2026), the authors plan dissemination to professional organizations and policymakers; Evidano’s exportable visualizations and AI summary features can generate tailored policy-ready summaries aligned to the study’s thematic outputs and timelines, reducing time-to-briefing.
Data security and ethics
Answer: Research teams must follow ethics-approved workflows for deidentification and data handling as described in the PLOS One protocol (Pokharel et al., 2026).
According to Pokharel et al. (2026), the study will deidentify transcripts and destroy original audio; Evidano documents its security practices and non-use-for-training commitments on the data security page to support compliance.
FAQ: physician perceptions of LLMs
What is the Technology Adoption Behavior (TAB) framework and why use it?
Answer: The TAB framework is a unified theoretical model combining perceived usefulness, perceived ease of use, and task-technology fit, and it guides both question design and deductive coding in the PLOS One study (Pokharel et al., 2026).
According to Pokharel et al. (2026), TAB integrates classic models such as TAM and task-technology fit and adds individual and social considerations to better explain complex uptake decisions.
How many interviews should a qualitative LLM perceptions study plan for?
Answer: Plan for purposive sampling until data sufficiency, using strata for LLM experience and years in practice as the PLOS One protocol recommends (Pokharel et al., 2026).
According to Vasileiou et al. as cited in the protocol, sample size sufficiency is justified by depth of conceptual understanding rather than a fixed numeric threshold (Pokharel et al., 2026).
Can AI tools bias thematic analysis of clinician interviews?
Answer: AI tools can introduce bias if applied without human oversight; the PLOS One protocol (Pokharel et al., 2026) emphasizes reflexivity and team-based coding to reduce interpretive bias.
According to Pokharel et al. (2026), reflexive memos and multidisciplinary coding discussions are central to preserving nuanced interpretation when using automated aids.
Conclusion & Next Steps
Recap: According to the PLOS One protocol (Pokharel et al., 2026), a TAB-informed descriptive qualitative study will interview family physicians in British Columbia to map perceptions of LLM adoption, using purposive sampling across clear LLM experience strata and reflexive team analysis.
Quote: Pokharel et al. (2026) write, "We plan to use semi-structured interviews with purposively sampled primary care physicians from British Columbia, Canada."
Quote: Pokharel et al. (2026) also state, "We expect participant recruitment and data collection to be completed within six months (anticipated timeline from August 1, 2025, to January 31, 2026)."
Next step: If you are a qualitative team running a similar protocol, combine rigorous reflexive practice with AI-enabled transcription, deidentification, and theme mapping to shorten time-to-insight without sacrificing interpretive depth.
Get started: To test AI-enabled qualitative workflows for your next LLM perceptions study, Try Evidano for free.
Topics
- physician perceptions of LLMs
- qualitative study LLM adoption
- LLM adoption in primary care
- technology adoption behavior framework
- AI qualitative analysis
Keep reading
- Commentary on NewsProtocol: Physician perceptions of LLMsHow primary care physicians view large language models, per a PLOS ONE protocol (Aug 7, 2026); learn methods, numbers, and how AI-enabled qualitative research accelerates insights.
- Commentary on NewsLLM Adoption in Primary Care: Qualitative ProtocolRead a PLOS One 2026 qualitative protocol on LLM adoption in primary care, methods and implications for researchers. Learn how AI-enabled qualitative tools can help.
- Commentary on NewsPhysician Perceptions of LLMs: Adoption in Primary CarePractical explainer of physician perceptions of LLMs from a PLOS One protocol (7 Aug 2026). Learn methods, timelines, and how AI-enabled qualitative research accelerates insight.