JMIR Publications

ABSTRACT

Background:

While clustering techniques are commonly used in bibliometric analysis to identify research themes, few studies systematically assign these themes back to individual articles. This gap limits the interpretability of findings and hinders granular, article-level longitudinal analysis.

Objective:

This study introduces the Theme Assignment Algorithm for Articles (TAAA), a data-driven framework designed to map clustered themes to individual publications. We demonstrate its utility by identifying dominant research patterns and thematic shifts within JMIR Aging.

Methods:

TAAA was applied to 434 JMIR Aging articles published between 2020 and 2025. Keywords were harvested from three sources: WoSCC Keywords Plus, author-provided keywords, and abstract-derived terms. These were grouped into thematic clusters using the Following Leader Clustering Algorithm (FLCA). The TAAA, implemented via R and a web-based application, determined each article’s primary theme using statistical mode to create a discrete article-level variable. Core themes were identified via h-index computation. Analytical visualization included Kano and Sankey diagrams, alongside volcano plots and heatmaps. The framework’s robustness was further tested by applying a "differentially expressed genes" (DEG) analogy to map "unknown" core metadata across "known" pre/post publication stages using Cohen’s Kappa as the extent of mapping power(MP).

Results:

Analysis across the three keyword sources yielded 9, 7, and 9 core themes, respectively. The most prominent themes identified were HEALTH (39.4%), OLDER ADULTS (43.1%), and DEMENTIA (25.3%). Notably, DEMENTIA emerged as a consistent core theme across all sources and visual layers, validating the TAAA’s ability to capture cross-source thematic coherence. The adaptation of dual heatmaps demonstrated the algorithm’s capacity for comparative bibliometric mapping in JMIR Aging. A mapping precision of 0.33 provided quantitative evidence of a two-stage publication pattern, though the separation between stages was not strictly confined to predefined time intervals.

Conclusions:

The TAAA framework provides a replicable, scalable, and interpretable method for article-level thematic assignment. Its ability to uncover consistent research patterns—specifically the dominance of dementia-related studies in JMIR Aging—demonstrates its value for bibliometricians and its potential adaptability to other domains, such as bioinformatics-inspired meta-analyses.