Predictive learning analytics (PLA) is gaining prominence in higher education as a data-driven approach to improve student retention and academic success. By leveraging machine learning models to identify at-risk students early, institutions can implement targeted interventions to reduce dropout rates. Although some studies explore PLA’s implications for educational equity, most evidence concentrates on its effectiveness in predicting and preventing academic attrition. This systematic review analyzed empirical research on PLA interventions in higher education, following PRISMA 2020 guidelines. Searches were conducted in Web of Science, Scopus, and PubMed, focusing on predictive analytics, dropout prevention, and student persistence. Of 141 records retrieved, 28 studies met inclusion criteria. Eligible studies reported outcomes related to retention, academic performance, or equity indicators. Risk of bias was assessed using ROBINS-I and Cochrane tools. A narrative synthesis was carried out, supported by subgroup and sensitivity analyses. Findings indicate that predictive models such as Random Forest, Support Vector Machines, and deep learning architectures achieve high accuracy in identifying students at risk of dropping out. Several studies went beyond prediction to implement interventions—such as motivational support or academic guidance—based on PLA insights, showing measurable improvements in student persistence. While some studies considered demographic and socioeconomic variables, evidence directly supporting PLA’s impact on reducing equity gaps remains limited. The review concludes that predictive learning analytics is an effective tool for enhancing student retention through timely identification and support. However, more research is needed to standardize methods, validate models across diverse educational contexts, and address ethical concerns related to transparency, bias, and data use.