Work

Retrospective harmonization of neuropsychiatric questionnaires in a large multi-site research consortium using expert mappings and semantic textual similarity

Year: 2026
Type: preprint
Abstract: Differences between clinical questionnaires limit the pooling of symptom data across studies, particularly when participants have completed different instruments. We developed a framework that uses questionnaire item content to harmonize neuropsychiatric measures without requiring overlapping response data. Experts mapped items to shared symptom dimensions, which informed score transformation and the training of models to map new items automatically. We evaluated the framework in a multisite Parkinson’s disease sample and an independent depression cohort. Transformation improved score alignment across questionnaires, while additional analyses supported clinical relevance and cross-questionnaire convergence. Our fine-tuned embedding models performing the semantic mapping task generalize to unseen questionnaires with a mean accuracy of 85.3%. This semantic, dimensional harmonization approach supports retrospective pooling of heterogeneous questionnaire data and offers a practical and systematic path toward mega-analyses on retrospective datasets.
Language: English
Open Access status: green
OpenAlex ID: W7223224185
Cites:
Cited by:
Related to: 10