Abstract: [Purpose/Significance] As an important altmetric indicator, Mendeley readership is often compared with citation counts. Examining the relationship between the early accumulation of Mendeley readership and subsequent citation counts of scientific publications, while controlling for confounding factors, can help validate their potential to predict the future scholarly impact of scientific literature and provide empirical support for evaluating the influence of scientific papers in their early stages of publication. [Design/Methodology] Based on a dataset of 629,234 scientific publications, this study employs Coarsened Exact Matching（CEM）to control key confounding variables, and uses negative binomial regression to estimate the association between Mendeley reader accumulation within six months after publication and citation counts in the subsequent five years. Furthermore, sub-disciplinary regression models are constructed for ten broad disciplinary categories to analyze disciplinary differences. [Findings/Conclusion] Early accumulation of Mendeley readership shows a significant positive association with citation counts of scientific papers. Compared with papers with zero Mendeley readership counts within six months after publication, those that did have Mendeley readership counts exhibited about 34.7% more citations over five years. This association is observed across all disciplinary fields, but with some variation in magnitude: Engineering & Materials Science shows the largest increase（about 42.8%）, while Social Sciences yields a relatively smaller increase（about 22.4%）. [Originality/Value] By introducing a causal inference-oriented approach that combines CEM and negative binomial regression into altmetric research, this study provides a more robust estimate of the relationship between Mendeley indicators and citation indicators, offering empirical support for the potential value of altmetric indicators in complementing and improving research evaluation systems.

Keywords: Scientometrics, Altmetrics, Mendeley, Research evaluation, Citation analysis