Symbiosis in Health: The Powerful Alliance of AI and Propensity Score Matching in Real World Medical Data Analysis

Abstract

Background: The rapid expansion of real-world data in medicine is driving the adoption of advanced methods like Artificial Intelligence (AI) and Propensity Score Matching (PSM). AI is widely applied across diagnostics, prediction, and treatment planning, while PSM is a crucial statistical technique used in quasi-experimental studies to mitigate confounding bias and approximate the reliability of randomized controlled trials. There is a growing research interest in combining these two methods to leverage their symbiotic strengths, but this association has not been holistically explored. Methodology: This study employed Synthetic Thematic Analysis (STA), derived from synthetic knowledge synthesis, to systematically review the existing literature on AI and PSM in medicine. Publications were harvested from the Scopus database using a comprehensive search string limited to the Medical subject area. The resulting corpus (N=433 documents) was analyzed using bibliometric tools (Bibliometrix and VOSViewer) to map the research landscape, identify thematic clusters based on author keywords, analyze collaboration patterns, and synthesize findings from highly prolific publications. Results: The field is young and rapidly accelerating, showing an exponential increase from 2020 to 2024. China and the USA dominate research production and citation impact. The symbiotic relationship is published in high-impact medical journals and health informatics journals. STA identified four main thematic clusters: Prediction, Cancer Management, Diagnosing, and Deep Learning. AI and PSM are combined in two primary ways: AI used in PSM and PSM used in AI. Conclusion: The symbiotic association between AI and PSM is a global and rapidly developing trend in medical research, driven by major international contributors. This convergence is enhancing methodological rigor in observational studies, primarily by improving prediction models and refining causal inference in complex areas like cardiovascular disease, cancer, and diagnostics.

Ähnliche Arbeiten