A Knowledge Graph Framework for Linking Health Assessment Scales and Scientific Literature in Low-Annotation Settings: Development and Evaluation Study

Health assessment scales are essential to measurement-based care, yet their adoption is hindered by difficulties in accessing relevant knowledge dispersed across unstructured scientific literature. This study explores a knowledge graph (KG) approach for the structured representation and semantic integration of health assessment scales and their associated literature metadata, aiming to propose a simple yet effective framework for this task. We defined a fine-grained KG schema that incorporates scale component entities and their relations, cross-scale mappings, and contextual associations. A construction pipeline was developed that combines prompt-based extraction of scale components and relations, type-specific entity normalization and relation fusion, cross-scale semantic mapping, and contextual linkage. To evaluate the pipeline, we built a prototype KG from 55 scales and 234 scientific articles, supported by a manually annotated dataset for cross-sentence relation extraction. Experiments with GLM-4-plus achieved F1 scores of 64.90% and 65.81% for component entity and relation extraction, respectively. Normalization yielded ARI and FMI scores above 0.54 for AssessmentScales and above 0.63 for MeasurementConcepts. Cross-scale mapping achieved F1 scores of 0.80 for AssessmentScales, 0.77 for MeasurementConcepts, and 0.85 for Items. The resulting prototype KG contains 6,468 concept-level entities and 32,096 triples, accessible through a public web interface. The scale-specific KG framework represents a practical method for structuring scale-related knowledge under low-annotation conditions, broadening medical KG research and supporting the advancement of measurement-based care.

The online version contains supplementary material available at https://doi.org/10.1007/s41666-026-00247-7.