Enhancing the prediction of publications' long-term impact using early citations, readerships, and non-scientific factors
This study aims to improve the accuracy of long-term citation impact prediction by integrating early citation counts, Mendeley readership, and various non-scientific factors, such as journal impact factor, authorship and reference list characteristics, funding and open-access status. Traditional citation-based models often fall short by relying solely on early citations, which may not capture broader indicators of a publication's potential influence. By incorporating non-scientific predictors, this model provides a more nuanced and comprehensive framework that outperforms existing models in predicting long-term impact. Using a dataset of Italian-authored publications from the Web of Science, regression models were developed to evaluate the impact of these predictors over time. Results indicate that early citations and Mendeley readership are significant predictors of long-term impact, with additional contributions from factors like authorship diversity and journal impact factor. The study finds that open-access status and funding have diminishing predictive power over time, suggesting their influence is primarily short-term. This model benefits various stakeholders, including funders and policymakers, by offering timely and more accurate assessments of emerging research. Future research could extend this model by incorporating broader altmetrics and expanding its application to other disciplines and regions. The study concludes that integrating non-citation-based factors with early citations captures a more complex view of scholarly impact, aligning better with real-world research influence.