PROMIS CATs: AI Thematic Analysis in Healthcare

According to the November 13, 2026 Feasibility of patient-reported outcomes measurement information system (PROMIS) computerized adaptive tests in patients receiving physical therapy: A mixed-methods study, PROMIS CATs were feasible and generally acceptable in primary care physical therapy, but wider adoption depended on reducing duplicate questionnaires, improving score interpretation, and embedding the tools in clinical workflows.

• The 2026 PLOS One study covered 14 physical therapy practices in the Netherlands and combined a survey of 25 physical therapists with interviews involving 10 physical therapists and 9 patients.
• The study found that 100% of surveyed physical therapists considered PROMIS CATs easy to administer, while 96% found the CATs easy to locate in the electronic health record.
• Only 28% of surveyed physical therapists reported regular daily use of PROMIS CATs, and 56% wanted to increase their use, according to the 2026 study.
• The qualitative interviews showed that patients generally viewed the CATs as quick and relevant when results were used in consultations, while therapists raised concerns about score interpretation, portal problems, and duplicate legacy PROMs.
• AI thematic analysis can help healthcare researchers compare clinician and patient perspectives, identify implementation barriers, and connect qualitative themes with survey frequencies.

The 2026 PLOS One study evaluated PROMIS CAT feasibility through acceptability, demand, practicality, and implementation, using a concurrent mixed-methods design. The researchers integrated Dutch-Flemish PROMIS CATs into electronic health records at 14 Dutch primary care physical therapy practices and examined routine-care experiences.

The study collected survey data from 25 physical therapists between May 2 and June 20, 2022. The researchers also conducted interviews with 10 physical therapists and 9 patients between March 21 and November 23, 2022; interviews lasted 32 to 62 minutes, with a mean duration of 47 minutes.

The researchers transcribed 19 interviews, using automatic transcription software for 12 interviews and a professional transcription service for 7 interviews. Two researchers independently open-coded transcripts, compared codes in consensus meetings, and mapped inductively developed themes onto Bowen's feasibility domains.

PROMIS CATs selected questions based on previous answers and stopped after at least 4 items when the precision threshold was reached, or after a maximum of 12 items. The study therefore examined both the efficiency promise of adaptive questionnaires and the human factors that determine whether clinicians and patients trust the results.

The qualitative findings added context that survey percentages alone could not provide. For example, the study reported that 80% of therapists were interested in using PROMIS CATs, but interviews explained that interest did not translate into routine adoption because CATs were administered alongside legacy PROMs.

AI thematic analysis in healthcare adds value by organizing large volumes of qualitative evidence while preserving the differences between stakeholder groups. The PROMIS CAT study demonstrates why this matters: therapists and patients did not always describe the same experience, and interview accounts explained the reasons behind survey discrepancies.

The study reported that only 24% of therapists believed patients were satisfied with PROMIS CATs, yet interviewed patients generally described the CATs as quick, simple, and relevant. The authors interpreted this divergence as a difference between therapists' perceptions of patient experience and patients' direct accounts.

The interviews also revealed implementation issues that were not reducible to questionnaire length. A physical therapist said, "I can’t always link the questionnaire scores directly to my treatment choices" (PT#3). Another therapist said, "If the CAT can replace NPRS, PSFS, and Quebec while giving the same information, that’s a win-win" (PT#6). These quotations show that adoption depended on clinical relevance and replacement of existing instruments, not simply on adding a new digital tool.

The study identified a second important theme: results were most useful when they supported a conversation. One patient said, "Answering the questions made me reflect on my symptoms and what I want help with" (Patient#5). AI-assisted analysis can help researchers locate recurring ideas such as reflection, relevance, score uncertainty, questionnaire fatigue, and workflow fit, then compare their frequency across patient and clinician segments.

Researchers should treat AI-generated themes as an analytic aid rather than an automatic substitute for methodological judgment. The PROMIS CAT researchers used independent coding, consensus meetings, a disagreement log, and peer debriefing, providing a useful standard for reviewing AI-supported outputs.

The 2026 PLOS One study shows that PROMIS CATs can support efficient, patient-centered conversations, but feasibility depends on workflow alignment, interpretation skills, and meaningful use of results. The study also shows why qualitative evidence is essential: interviews explained the gap between high interest and low routine adoption, as well as the different experiences reported by therapists and patients.

AI thematic analysis in healthcare can help research teams process transcripts and open-ended survey data, compare stakeholder segments, and connect themes with quantitative findings. Researchers should use AI to accelerate organization and discovery while retaining human review, transparent coding decisions, and appropriate privacy safeguards.

Evidano supports this workflow by combining document and spreadsheet analysis with thematic, frequency, and cross-segment outputs. Start your next healthcare qualitative research project with Try Evidano for free.