Evolution of Risk Stratification and Prognostic Prediction in Coronary Revascularization for Acute Coronary Syndromes: A Changepoint-Enhanced Science-Mapping Study

Background/Objectives: Risk stratification and prognostic prediction are embedded in revascularization practice for acute coronary syndromes, but the longitudinal evolution and scientific architecture of this literature remain poorly characterized. We mapped growth, contributors, conceptual structure, and named instruments. Methods: English-language articles and reviews indexed in the Web of Science Core Collection from 1975 to 2025 were retrieved with a three-block Topic search (n = 10,390; 9608 articles, 782 reviews). Bibliometric performance, collaboration, source impact, keyword co-occurrence (validated thesaurus), named instrument and clinical setting terms, and metadata completeness were analyzed with VOSviewer, Bibliometrix/Biblioshiny, and Python. Annual output was modeled by changepoint detection with bootstrap localization. Results: Annual output rose from 8 records in 1988 to 843 in 2025; the principal transition was a slowdown in proportional growth around 2009, from approximately 16.5% to 6.9% annually, reproduced on the 1991–2025 series. The United States led output, citations, and collaboration strength; China ranked second in volume, with the youngest output among the leading countries (median year 2022) and year-normalized citation impact of 0.78 versus 1.43 for the United States (corpus average 1.0). A 92-keyword network formed five clusters centered on percutaneous coronary intervention (PCI). SYNTAX, TIMI, and GRACE were the most frequently named scores, and machine learning terminology grew fastest (53 records in 2025). Records naming ST-segment elevation myocardial infarction alone (33.9%) outnumbered those naming non-ST-elevation presentations alone (13.7%), and 75.7% mentioned PCI without bypass surgery. Conclusions: These patterns suggest that prognostic assessment has developed primarily within revascularization practice rather than as an independent methodological field; machine learning and high bleeding risk terminology are emerging below the keyword map threshold.