Repeated measures are often collected due to longitudinal follow up or repeated sampling. Motivated by the Assessment, Serial Evaluation, and Subsequent Sequelae of Acute Kidney Injury (ASSESS-AKI) study, recurrent events of AKI may occur more frequently for the subject with deteriorated kidney function; however, some healthy subjects may not experience any AKI events. Therefore, the number of associated biomarker measures varies across subjects, which could be correlated with the overall kidney outcome, resulting in informative cluster size. Meanwhile, a high percentage of zero AKI recurrences are detected from this study, thus, we propose a joint modeling approach of longitudinal data and informative zero-inflated cluster size by considering a terminal event as a covariate. The parameter estimation is obtained by using a three-stage semi-parametric likelihood-based approach. Extensive simulations are conducted, and a real data application on the ASSESS-AKI is provided in the end.