{ "abstract": "Introduction: Large language models (LLMs) can render structured questionnaire variables as natural-language profiles, but their value in mental-health screening cannot be judged from a single fixed-threshold score. We evaluated LLM prompt conditions in a cross-questionnaire benchmark for predicting screen-positive status on the Self-Rating Depression Scale (SDS) among undergraduates. The intended application was an offline methodological benchmark using previously collected questionnaire data, not a shorter alternative to direct SDS administration.\n\nMethods: De-identified records were assigned to Development (n=1,951), Test (n=2,186), and Holdout (n=2,258) grade cohorts. We evaluated logistic regression, XGBoost, and 12 model-prompt conditions at the fixed 0.5 threshold and at thresholds selected only in Development-validation. We also examined paired bootstrap differences, ranking stability, scale correlations, grouped ablations, cohort shift, calibration intercepts and slopes, alternative expected calibration error specifications, representative repeated calls at temperatures 0 and 0.1, sex-stratified performance, and per-1,000 screening scenarios.\n\nResults: At the fixed 0.5 threshold on Test, GPT E1, Gemini E2, and XGBoost achieved macro-F1 values of 0.732, 0.730, and 0.674. At Development-derived thresholds of 0.38, 0.50, and 0.174, the corresponding macro-F1 values were 0.702, 0.730, and 0.660. At the Development-selected specificity target, XGBoost achieved a Test macro-F1 of 0.738. Pairwise intervals showed no clear fixed-0.5 macro-F1 difference among the three highest-ranked LLM conditions. XGBoost yielded the lowest Brier scores. The combined-sample Spearman correlation between SDS and SAS was 0.60. Together with the ablation results, this correlation indicated predictive information distributed across overlapping questionnaire groups rather than dependence on a single measured group. In repeated calls across the complete Holdout cohort, inconsistent-case rates were 23.6% for GPT E1 and 1.5% for Gemini E2. Temperature 0 did not clearly eliminate GPT E1 instability in the representative 400-case subset.\n\nDiscussion: Model conclusions depended on both the threshold and the metric. Certain LLM prompt conditions were competitive at specific operating points, but the evidence did not establish a stable or general advantage over structured-data models. Nor did it establish external validity, clinical diagnostic utility, or reduced questionnaire burden. Threshold choice, calibration, output validity, and repeated-call stability therefore require separate evaluation before prospective use.", "abstract_shown": true, "affiliations": [ "School of Educational Sciences, Huangshan University", "Faculty of Health Sciences, Universiti Teknologi MARA", "School of Economics and Management, Huangshan University", "College of Education, Zhejiang Normal University", "School of Computer Science and Technology, Zhejiang Normal University", "Security Department, Zhejiang Normal University", "School of Psychology, Zhejiang Normal University", "College of Mathematical Medicine, Zhejiang Normal University", "AIFORIA (Hangzhou) Intelligent Technology Co., Ltd" ], "article_type": "research article", "arxiv": null, "author_total": 14, "authors": [ { "affs": [ 0, 1 ], "name": "Xiao Yang" }, { "affs": [ 0 ], "name": "Ying Cao" }, { "affs": [ 0 ], "name": "Haibin Wang" }, { "affs": [ 2 ], "name": "Huan Li" }, { "affs": [ 3 ], "name": "Xin Jiang" }, { "affs": [ 4 ], "name": "Huiying Xu" }, { "affs": [ 5 ], "name": "Tao Yao" }, { "affs": [ 3 ], "name": "Xiaolei Zhang" }, { "affs": [ 6 ], "name": "Yekai Chen" }, { "affs": [ 4 ], "name": "Jiaming Li" }, { "affs": [ 7 ], "name": "Bin Zhou" }, { "affs": [ 8 ], "name": "Lingwei Xu" }, { "affs": [ 4 ], "name": "Zhendong Chen" }, { "affs": [ 4 ], "name": "Xinzhong Zhu" } ], "bibtex": "https://research.draconitic.com/paper/encr0rj4fsrjc.bib", "cited_by": 0, "doi": "10.3389/fpsyt.2026.1942164", "fulltext_url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13724331/?tool=EBI", "id": "encr0rj4fsrjc", "issns": [ "1664-0640" ], "landing_url": "https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1942164/full", "language": "eng", "licence": "CC-BY-4.0", "licence_label": "CC BY 4.0", "licence_source": "article", "licence_url": "https://creativecommons.org/licenses/by/4.0/", "metadata_licence": null, "native_subjects": [ [ "heading", "Original Research" ], [ "openalex", "3605|Artificial Intelligence in Healthcare" ] ], "openalex": "W7222986200", "pdf_url": "https://www.frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2026.1942164/pdf", "pmcid": "PMC13724331", "pmid": "42973452", "published": "2026-10-19", "publisher": "Frontiers Media SA", "ris": "https://research.draconitic.com/paper/encr0rj4fsrjc.ris", "sources": [ { "first_seen": "2026-10-30T12:20:18Z", "native_id": "10.3389/fpsyt.2026.1942164", "source": "crossref", "url": "https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1942164/full" }, { "first_seen": "2026-11-03T15:20:51Z", "native_id": "PMC:PMC13724331", "source": "europepmc", "url": "https://europepmc.org/article/PMC/PMC13724331" }, { "first_seen": "2026-11-04T12:22:55Z", "native_id": "PMC13724331", "source": "pmc", "url": "https://doi.org/10.3389/fpsyt.2026.1942164" }, { "first_seen": "2026-11-04T15:20:48Z", "native_id": "MED:42973452", "source": "europepmc", "url": "https://europepmc.org/article/MED/42973452" } ], "subjects": [ { "field": "medicine", "origin": "issn:1664-0640", "subfield": "psychiatry" } ], "title": "SDS screen-positive prediction from structured questionnaires in university students: a comparative evaluation of large language model prompting strategies", "title_text": "SDS screen-positive prediction from structured questionnaires in university students: a comparative evaluation of large language model prompting strategies", "type": "peer-reviewed", "updated_at": "2026-11-06T12:20:51Z", "url": "https://research.draconitic.com/paper/encr0rj4fsrjc", "venue": "Frontiers in Psychiatry", "versions": [] }