Assessment of Non-Institutional AI Tool Usage Among Clinicians
Sarah Pungitore, Jarrod Mosier
physics.soc-ph · cs.AI
Quick verdict
Survey of 44 clinicians at one center; 97.7% reported non-institutional AI use, but self-selection and internal numeric inconsistencies warrant caution.
Summary
This descriptive cross-sectional survey examines clinicians' use of non-institutional AI tools — those not provided, managed, or governed by their institution. A Qualtrics survey was distributed once via listserv to roughly 2,000 clinicians at the University of Arizona College of Medicine–Tucson and Banner University Medical Center–Tucson between May 20 and June 26, 2026; 44 completed responses (2.2% response rate) were analyzed descriptively. Respondents were mostly APPs and faculty from Emergency Medicine and Medicine. Forty-three of 44 (97.7%) reported non-institutional AI use for at least one work-related task in the prior six months. Conversational AI and clinical decision support/diagnostic AI led usage (28 respondents each), followed by question answering (20) and OS-integrated and clinical documentation tools (18 each). Administrative and clinical tasks dominated: over 60% used AI for writing emails and drafting letters or reports, 59% for literature searching, 55% for finding educational materials, with heavy use of drafting patient notes and transcription plus diagnostic assistance. The authors add a risk-tiered table of practical considerations. Key limitations: severe self-selection from one-contact recruitment, single-center design, self-reported use, and internal numeric inconsistencies in the text (one tool count conflicts with Table 1's percentages; the Conclusion reverses a count), with neither survey instrument nor data available.
Review complete
Assessment at a glance
4
risk flags identified
● high 0
● medium 3
● low 1
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- 97.7% (43/44) of surveyed clinicians reported using non-institutional AI tools for at least one work-related task in the preceding six months.
- Conversational AI and clinical decision support/diagnostic AI were the most-used categories (28 respondents each); AI use appeared in every assessed task across all five workload categories.
- Administrative and clinical tasks showed the most frequent use, including higher-risk activities such as diagnostic assistance and clinical decision support.
- Usage patterns varied little by position, career stage, or department, and non-institutional AI adoption warrants dedicated institutional governance frameworks.
Strengths
Novel construct, transparently operationalized
First survey centered specifically on non-institutional (ungoverned) AI tools; the Methods section discloses that the only institutional tool (Oracle Clinical AI Agent) was left in the survey to avoid confusion, and that its tool category plus the 'Drafting patient notes' and 'Transcribing patient visits' tasks were excluded from non-institutional-use analyses — 43/44 usage persists after exclusion.
Triangulable sample tables
Table 1 stratifies all 10 tool categories by position, career stage, and department; the position and department blocks each sum to 44 and the narrative marginals (28/18/2/5/3/18/28/1/3/20/1) reconcile with the position-block percentages across every column.
Granular task-frequency measurement
Actionable risk framework
Weaknesses
Severe self-selection and non-response
Internal numeric inconsistencies
Blurry institutional/non-institutional boundary
Fragile descriptive stratification
Single-center, self-reported
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Preliminary Review: Assessment of Non-Institutional AI Tool Usage Among Clinicians
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What did this study find about clinicians using non-institutional AI tools?
Among 44 clinicians at one academic medical center who completed a 2026 survey, 43 (97.7%) reported using AI tools not provided or governed by their institution for at least one work-related task in the prior six months. Use spanned all surveyed tasks, with administrative and clinical documentation tasks most frequent, and included higher-risk activities such as diagnostic assistance and clinical decision support.
Which AI tools were most used by clinicians in the survey?
Conversational AI (e.g., ChatGPT, Claude, Gemini, Perplexity) and clinical decision support/diagnostic AI (e.g., Aidoc, OpenEvidence, Tempus AI) were most reported — 28 of 44 respondents each — followed by question-answering tools (20), and OS-integrated assistants plus clinical documentation tools (18 each).
What does 'non-institutional AI' mean?
AI tools that are not provided, managed, or governed by the clinician's healthcare institution — for example, a personal ChatGPT account used for work tasks, as opposed to a hospital-vetted scribe or documentation system. These tools raise privacy, security, and accuracy concerns because they lack institutional oversight and data agreements.
What are the study's main limitations?
Severe self-selection (44 responses from ~2,000 contacted, a 2.2% response rate from a single listserv email), a single urban academic center, fully self-reported usage, and descriptive-only analysis with small subgroups. The review also identified internal numeric inconsistencies (one tool count conflicting with Table 1's percentages and a reversed count in the Conclusion), so figures should be read as provisional until audited.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figure 1 and the supplementary materials (full Qualtrics survey, Table S1 task list) are referenced in the text but not contained in the provided extraction, so their content could not be assessed.
- Table 1 was extracted as dense percentage runs; the reported column-sum contradictions were re-derived arithmetically from the extracted percentages, but original table formatting could not be inspected.
- The abstract as supplied in the paper metadata cuts off mid-sentence ('including tas'); the full abstract within the paper body text is complete.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.