What Nursing Students Say: AI Thematic Analysis

According to the PLOS One study published on November 5, 2026, a cross-sectional survey of 699 third-year nursing students combined importance-performance ratings, exploratory stress analyses, and supplementary open-ended responses.

The study’s qualitative component produced 217 non-mutually-exclusive segment-to-theme assignments, showing how thematic analysis can add context to nursing education survey results while not establishing prevalence or causality.

• According to Tsai and Chen’s 2026 study, all 699 analyzed students answered both open-ended practicum questions, but only substantively meaningful response segments were assigned to themes.
• The researchers reported 128 coding assignments for perceived practicum gains and 89 assignments for practicum difficulties, for 217 assignments in total.
• The researchers recorded 92% initial intercoder percentage agreement before consensus discussion, but they did not calculate Cohen’s kappa.
• The study found that nursing-process ratings had an importance-performance gap of 0.229, with a 95% confidence interval from 0.179 to 0.280.
• The study authors cautioned that the adapted stress questionnaire lacked support for its prespecified two-factor structure, so stress findings should be treated as exploratory.

The 2026 study used open-ended survey responses as supplementary qualitative evidence alongside quantitative measures of nursing competency, performance satisfaction, and learning stress.

Tsai and Chen asked students, “What was your greatest gain during the practicum? ” and “What was your greatest difficulty during the practicum? ” The authors analyzed responses written in Chinese through thematic content analysis, using meaningful response segments as the coding unit.

The researchers independently reviewed responses, developed preliminary codes and a codebook, compared coding decisions, discussed disagreements, consulted experts when needed, and retained an audit trail. The researchers also examined negative and divergent cases rather than forcing every response into a dominant theme.

The qualitative results were not prevalence estimates because one student could contribute multiple segments and one segment could receive multiple thematic assignments. The study authors explicitly distinguished the 217 coding assignments from the number of unique respondents.

The study connected written comments with quantitative findings through a joint display. According to Tsai and Chen, students described gains involving communication and nursing skills, while reported difficulties included nursing-process assignments, theory-practice integration, and stress.

Nursing education researchers should use AI thematic analysis to organize student feedback quickly, then preserve human judgment about context, ambiguity, and evidentiary limits.

The study shows why researchers should connect qualitative themes to, but not substitute them for, quantitative results. Students’ comments about care-plan preparation and theory-practice integration can help explain survey patterns, while the cross-sectional design cannot show whether those difficulties caused stress or lower academic performance.

Researchers should report the unit of analysis before reporting theme counts. In Tsai and Chen’s study, the unit was a meaningful response segment, and the resulting counts were non-mutually-exclusive coding assignments rather than percentages of students.

Researchers should also retain uncertainty during interpretation. The study reported that communication and biomedical-science quadrant classifications remained stable in only 51.3% and 52.1% of bootstrap samples, respectively. Those findings show why automated theme summaries and quantitative classifications should be reviewed against the method, sample, and measurement quality.

For health and clinical education research, qualitative analysis is non-diagnostic and research-focused. Student comments about stress should not be interpreted as clinical diagnoses or as proof that a counseling intervention worked or failed.

The 2026 nursing practicum study shows that open-ended student feedback can add important context to survey-based evaluations, especially around communication, nursing skills, theory-practice integration, and stress.

The study also demonstrates why AI thematic analysis must preserve coding-unit definitions, uncertainty, measurement limitations, and the difference between association and causality. Evidano can help nursing researchers analyze open-ended surveys and documents faster while keeping themes, segments, and cross-group comparisons organized for human review.

Researchers who want a practical starting point can try Evidano for free and explore an AI-assisted workflow for nursing education and qualitative research data.